Essays on the Namibian Economy Michael N. Humavindu Umeå 2008

Essays on the Namibian Economy Michael N. Humavindu Umeå 2008
Essays on the Namibian Economy
Michael N. Humavindu
Umeå 2008
Copyright © 2008 by Michael N. Humavindu
Umeå Economic Studies No. 745
From the Department of Economics
Umeå University, Umeå, Sweden
ISBN: 978-91-7264-633-9
ISSN: 0348-1019
Printed in Sweden by Arkitektkopia, Umeå, 2008.
Abstract
This thesis consists of an introduction and four papers exploring various aspects
of the Namibian economy. These aspects cover shadow pricing, environmental
valuation and capital market development in Namibia.
Paper I estimates the shadow prices of capital, labour and foreign
exchange for the Namibian economy. The results suggest that the shadow price
of capital for Namibia is 7.2%. The economic costs of Namibian labour, as a
share of financial costs, are 32% for urban semi- and unskilled labour, and 54%
for rural semi- and unskilled labour. The economic cost of foreign labour as a
share of financial costs is 59%. The estimated shadow exchange rate factor is
4% for the Namibian economy.
Paper II derives a set of accounting price ratios (APRs) for the various
economic sectors of Namibia by using the Semi-Input–Output (SIO) Technique.
An APR is the ratio between the market or financial price and the efficiency or
economic value of a specific commodity or sector, which is useful for the
economic analysis of investment or development initiatives. This larger set of
APRs, derived on the basis of information contained in a Namibian Social
Accounting Matrix (SAM), should be useful in improving the effective
appraisal of development projects and other major investment programmes in
Namibia.
Paper III analyses returns and volatility on the Namibian and South
African stock markets, using the daily closing indices of the Namibian Stock
Exchange (NSX) and the Johannesburg Stock Exchange (JSE). The sample
covers the period from 4 January 1999 to 20 March 2003. The methodology has
three main parts: (i) unit root tests, (ii) cointegration analysis, and (iii) volatility
modelling. The results show that the two markets exhibit very low correlations,
and there is no evidence of a linear relationship between the markets.
Furthermore, a volatility analysis shows evidence of no spillover effects. These
results suggest that the NSX could be an attractive risk diversification tool for
regional portfolio diversification in southern Africa
Paper IV studies the determinants of property prices in the township areas
of Windhoek, the capital of Namibia. The work’s major finding is that
properties located close to an environmental bad (e.g. garbage dump) sell at
considerable discounts. On the other hand, properties located near an
environmental good (e.g. a recreational open space) sell at a premium. These
results provide evidence of the importance of environmental quality in lowerincome property markets in developing countries. It is important, therefore, for
Namibian urban planners to incorporate environmental quality into the planning
framework for lower-income areas.
Acknowledgements
I would like to thank the following people and institutions who, informally or formally, have
contributed to this work and have helped me during my studies.
Jesper Stage, my supervisor, mentor and co-author of one of the papers in this thesis: words
of gratitude could never be enough to thank you for what you have done, and continue doing,
in helping me become a competent economist working for the development of my country.
Hopefully, one day, your faith in me will be justified.
To my co-supervisor, Tomas Sjögren, many thanks for your valuable comments during the
thesis writing phase.
To my co-authors on Papers III and IV, I am grateful for the academically enriching
experience of working with such brilliant minds.
My gratitude is also due to Kirk Hamilton and Sylke von Thadden for their constant
encouragement.
I also owe thanks to Karl-Gustaf Löfgren, David Potts, Kenneth Backlund, Jon Barnes,
Tomas Sjögren, Magnus Wikström and Kurt Brännäs, for their comments on earlier versions
of the papers in this thesis. Elio Londero provided me with some useful material for the
shadow pricing papers.
Thank you to Sandie Fitchat for her consistent language editing.
To the Jan Wallander and Tom Hedelius Foundation, I am extremely grateful for the research
grant that made all this possible.
To my wife, Veripi Humavindu, thank you for your constant encouragement and appreciation
of the life that we share.
To my colleagues at the Development Bank of Namibia, I value the wonderful working
relationship that we have built up over the years. John Mbango, Valentine Schaneck, Martin
Inkumbi, David Nuyoma and Gottlieb Hinda deserve special mention. Erastus Hoveka, now
at Nedbank Namibia, was and still is a valuable mentor to me.
Peter Muteyauli, Olympio Nhuleipo, Gerson Kadhikwa, Daniel Motinga and Mihe Gaomab II
have provided inspiration to me as committed Namibian economists.
Finally, many thanks go to my friends and family for their constant encouragement. Uahatjiri
Ngaujake, Gotti Riruako, Naphataline Ndivanga, Selina Meroro, Kaitaa Meroro, Jacklyn
Hambuindja, Edson Humavindu, Matjiua Humavindu and Rudolph Humavindu: your belief
in my ability is greatly appreciated.
The work is dedicated to my grandmother, Kauripondua Humavindu; my mother, Yahepa
Humavindu; my son, Tjarirove Mbajoroka Humavindu; and especially my niece, Charmaine
Kuverua, who faced personal adversity during the summer of 2006 to early Autumn 2007.
Thank you, Charmaine, for teaching me the virtues of tenacity and the commitment to survive
even if all odds are set against you.
Michael Nokokure Humavindu
Umeå, 19 August 2008
This thesis consists of an introduction and the following four self-contained papers:
I
Humavindu, MN (2008): “Estimating national economic parameters for Namibia”.
Umeå Economic Studies 744. A shorter version of this paper has been resubmitted to
the South African Journal of Economics.
II
Humavindu, MN (2008): “Estimating Namibian shadow prices within a semi-input–
output framework”. Forthcoming in the Journal for Studies in Economics and
Econometrics.
III
Humavindu, MN & C Floros (2006): “Integration and volatility spillovers in African
equity markets: Evidence from Namibia and South Africa”. African Finance Journal,
Vol. 8(2):31–50. Reprinted with permission from the Africagrowth Academy. The
version published in the African Finance Journal contained typographical errors and a
graphical omission; these have been corrected in the version published here.
IV
Humavindu, MN & J Stage (2003): “Hedonic pricing in Windhoek townships”.
Environment and Development Economics, 8(2):391–404. Reprinted with permission
from the Cambridge University Press.
Introduction and summary
INTRODUCTION AND SUMMARY
1.
Introduction
This thesis consists of four self-contained papers analysing shadow pricing, capital market
development, and aspects of environmental economics in Namibia. Paper I estimates shadow
prices of capital, labour and foreign exchange for the Namibian economy for use in a social
cost-benefit analysis. Paper II extends the analysis set out in Paper I by estimating sectoral
accounting price ratios (APRs) for the Namibian economy. Paper III analyses the integration
of the Namibian and South African equity markets. The ultimate aim of this analysis is to
examine the scope for diversification for investment managers. Paper IV analyses the
determinants of property prices in a low-income area of Namibia’s capital, Windhoek. The
focus is on the implicit valuation of environmental advantages and disadvantages in lowincome property pricing markets.
Although these papers cover different aspects of economics, a common thread linking them is
that effective project/programme evaluation, be it for developmental projects or capitalmarket development initiatives, can enhance decision-making. Another link between the
papers is that missing or distorted market prices can lead to suboptimal investment decisions
in a wide range of circumstances and for a wide range of agents, national planning agencies,
local government, and private investors. This is becoming more important in the light of
recent Namibian Government efforts to restrict the huge outflows of capital to neighbouring
South Africa (IMF 2006, 2008; Bank of Namibia 2003). If successful, these efforts will lead
to increased investment in Namibia, increasing the risk of suboptimal investments if markets
remain missing or distorted.
1
Michael N Humavindu
Economic tools for project appraisal in developing countries have been well established since
the early 1970s. The appropriate appraisal of public investment projects underlines the need
to determine the social value of costs and benefits accruing from these investments. In
developing countries in particular, social values may diverge from market prices and values.
These price distortions may be caused by market imperfections as a result of both
government interventions in product and factor markets, structural disequilibria in labour
markets, and thin or missing markets. As a result of these distortions, market prices can be
unreliable indicators of the real net worth of goods and services (Adhikari 1986). Official
trade policy, such as the adoption of various tariff and non-tariff trade barriers, may lead to a
distorted market value of foreign exchange. The result is a distortion in the domestic price of
all tradables, but also of non-tradables which use tradables in their production. In labour
markets, the equilibrium wage may be higher than the market clearing wage as a result of
minimum wage laws and a union bargaining presence. In capital markets, the market interest
rate may diverge from the marginal productivity of capital. For environmental externalities,
there may not be any prices at all, potentially creating biases against decisions that benefit the
environment, and in favour of decisions that harm the environment.
In project appraisal, therefore, modifications to market values are essential. A modification is
determined by estimating a set of national parameters and conversion factors. These
parameters are termed shadow prices. Conversion factors give the ratio between the price to
be used in evaluating an input or output of a project (the shadow price) and the market price
of that input or output. In the valuation of inputs used in production, the inherent assumption
is that the price of any input should represent the opportunity cost of that input. The
opportunity cost reflects the value of output forgone on one project when used on another.
Thus, shadow prices are useful when the market price for an input or output is unavailable or
2
Introduction and summary
does not reflect its opportunity cost. For example, labour is an important input in many
investment projects and, therefore, should be valued at its economic cost.
Shadow prices are a crucial link between the macro level and the project level of economic
planning, and an important component of the overall process of development planning in
developing countries. Only when the costs and benefits of all potential projects are valued at
their shadow prices may those projects that most efficiently use scarce resources be selected.
Following this strategy allows a developing country to maximise the potential net economic
benefits accruing from its public investments, thereby improving its potential to pursue
broader social, political and other non-economic objectives (Saerbeck 1989).
In general, national parameters to be estimated for economic analysis are divided into five
categories (Potts et al. 1998): primary factors, traded goods, non-traded goods, average
estimates, and the discount rate.
Primary factors relate to different categories of labour, the value of domestic resources, and
foreign exchange.
Traded goods are goods for which the economic cost or benefit derived from their use is
determined by their international prices. Shadow price estimation is essential where there is a
significant difference between the border price and the local market price. Deriving the
shadow price is also a necessity in situations where a benefit is likely to feature prominently
as an input or output for a number of projects.
3
Michael N Humavindu
Non-traded goods are items that, by their nature, cannot be traded across borders, or may not
be economically viable for trade. The estimation of a shadow price is prompted by a situation
where there is a significant difference between the local market price of a resource and its
economic value, or where, as for traded goods, a benefit is likely to feature prominently as an
input or output for a number of projects.
Average estimates relate to sectors where cost data do not allow further breakdown. The most
important of such estimates, the standard conversion factor, describes the value of a unit of
domestic resources in terms of a unit of foreign exchange. The standard conversion factor, in
the case of average estimates, is derived indirectly through conversion factors for traded
goods.
Discount rates quantify the effect of time on a project’s cost and benefit values.
At the national level, classic shadow pricing estimation would involve deriving a general
equilibrium economic optimisation model with the following specific features (UNIDO
2003):
x
An objective function, describing the effects of the use and generation of resources on
a measure of economic value such as the gross domestic product (GDP)
x
Constraints on the use of resources (technological coefficients for each economic
activity and a limit for the resource as a whole), and
x
Non-zero constraints for the value of resources, and non-negativity constraints for
resources.
4
Introduction and summary
The shadow price is then the effect on the value of the objective function resulting from an
increase or decrease by one unit in the availability of a scarce resource.
At the microeconomic level, numerous studies have been made estimating shadow prices for
objective functions with one or a few non-market constraints. The distance function
methodology, for example, is used in deducing shadow prices for pollutants (Lee et al. 2003).
Namibian examples of micro-level shadow pricing include the shadow pricing of
environmental goods (e.g. Humavindu & Masirembu [2001]; see Humavindu [2002] for an
overview of other examples) and shadow pricing of fishing quotas (Stage & Kirchner 2005).
However, the estimation of nationwide shadow prices in this way is fraught with complexities
and numerous constraints, and is usually infeasible in practice (Little & Mirrlees 1974). This
has led to the adoption of 'second-best' approaches to shadow price estimations. These
methods were developed in the late 1960s and early 1970s by the United Nations Industrial
Development Organisation (UNIDO 1972, 1978, and 1980), and by Little and Mirrlees
(1974), and Squire and Van der Tak (1975). The departure point of these approaches is the
choice of unit of account.
In essence, the UNIDO approach uses a domestic resource as the unit of account, and it
estimates the scarcity value of foreign exchange using a shadow exchange rate. This
procedure is described as the use of a domestic price numeraire (Potts 2002). A second
approach, developed by Little and Mirrlees (ibid.) and Squire and Van der Tak (ibid.), uses
the unit of foreign exchange (expressed in local currency units) as the numeraire. The latter
method is described as the use of a world price numeraire.
5
Michael N Humavindu
A third approach is based on the premise that all shadow prices are interdependent because
their value depends on the value of inputs from other sectors (Potts 2002). These
interdependencies are accounted for through conversion factors that are derived by solving a
series of simultaneous equations using an input–output approach. Thus, this approach takes
into account all the sectoral interrelationships: it is called the semi-input–output (SIO)
analysis, and is useful for non-traded sectors where the output from each sector may appear
as inputs into others.
In estimating shadow prices, the choice of methodology is primarily determined by the nature
and extent of available data. Readily available data were a constraining factor in this study.
Therefore, an initial attempt is made in Paper I to estimate the three primary factors: capital,
labour, and the exchange rate. Paper II recounts how the SIO analysis was employed to
determine accounting price ratios for the various Namibian economic sectors.
Despite the clear importance of shadow pricing for a developing country such as Namibia, no
set of official national parameters exists; nor has any attempt been made to estimate them
until now. The country’s development path is guided by five-yearly National Development
Plans (NDPs). These NDPs stress the importance of investment/development projects to
alleviate chronic unemployment, low industrialisation, poverty, and income inequality. Under
such circumstances, it is vital that market signals provide an adequate guide for investment
planning and project appraisal. There is an apparent need for a consistent set of prices that
reflects the resource costs and social benefits of a proposed course of action. High
unemployment (36%), uneven income distribution, and an economy that exports most of its
capital are all strong motivations for the estimation of a set of national parameters. Recently
released national guidelines (Guidelines for preparing the Third National Development Plan
6
Introduction and summary
(NDP3): 2007/08 – 2011/12) reassert the importance of investment/development projects for
Namibia’s economic progress. Moreover, the government recently amended Regulation 28 of
the Pension Fund Act, 1956 and Regulation 15 of the Long-term Insurance Act, 1998 to
enhance the availability of funds for local investments and to deepen financial markets. The
availability of more funds for local and unlisted investments stresses the need for the
effective and prudent allocation of resources. In response, Paper I constitutes the first formal
attempt to estimate the shadow prices of capital, labour, and foreign exchange for the
Namibian economy. The estimation is based on data representing national averages and is to
assist in efficient and effective decision-making in investment allocation. Paper II extends the
shadow pricing analysis further, by employing an SIO analysis to arrive at sectoral APRs.
The availability of sectoral APRs is useful when investment decisions are to be driven by
sectoral allocations.
Cottoning on to the new initiatives for deepening Namibian financial markets, Paper III
examines whether the NSX offers regional portfolio diversification opportunities for
investors away from the JSE. This is essential to analyse, given the amounts of funds that
now need to be invested in the Namibian economy. Previous studies have created the
perception that the NSX tends to follow the JSE, implying that the scope for diversification
by investing in the NSX is limited. This has led to limited interest from investors and to thin
trading which, in turn, can lead to potentially misleading and volatile stock prices that may
cause underinvestment.
Although the Namibian and South African economies are closely linked, it does not
necessarily follow that the two countries’ stock markets are linked. The NSX is primarily
composed of dual-listed companies. The local firms listed on the exchange comprise only
7
Michael N Humavindu
0.3% of market capitalisation (IMF 2008). Companies having primary listings on the JSE and
the London Stock Exchange respectively represent 44% and 55% of NSX market
capitalisation. Moreover, the NSX overall index has always tracked the JSE, but the local
index tends to follow its own path (IMF 2008, Bank of Namibia 2007). This makes it
interesting to study whether the local firms provide more scope for diversification than
studies of the overall index have indicated.
The methodology used encompasses correlations, cointegration and volatility modelling
(Engle & Granger 1987, Engle 2001). Ceteris paribus, a low correlation between assets,
means lower portfolio risk and opportunities for portfolio diversification. However,
correlations induced by short-term trading can obscure long-run linkages among stock
markets (Chen et al. 1986). To circumvent the problems associated with correlations, unit
roots and cointegration analysis are employed.
Financial variables that have time-varying means and variances are termed non-stationary
and have unit roots (Harris & Sollis 2005). However, non-stationary variables may have
common trends, and may form stationary linear combinations (based on equilibrium long-run
relationships). Cointegration implies a long-run co-movement between trended economic
time series, meaning that there is a common equilibrium relation to which the time series
have a tendency to revert. Stock markets whose indices tend to follow each other are said to
be cointegrated. When they are, the equity markets move in tandem, and there are no longterm gains from international diversification.
Extending cointegration analysis a bit further, volatility modelling may be applied to further
examine equity market integration. Moreover, it is important to ascertain whether an adverse
8
Introduction and summary
situation in one equity market actually spills over into another equity market. Volatility refers
to the riskiness of stock prices and is an important determinant of the cost of capital for an
investment project underlying the stock or portfolio of stocks in question. The models of
conditional volatility commonly used in finance imply that there may be predictable patterns
in stock market volatility. Such models imply that investors can predict risk, thereby assisting
in investment decisions. Where an investor has forecast future prices to be volatile, they
might opt to leave the market or require a much higher premium.
Shadow prices based on national data averages have to be distinguished from sectoral,
regional or project-specific parameters (Saerbeck 1989). Ideally, project-specific parameters
should be estimated for each individual project because the opportunity costs of the resources
used or produced may differ from project to project, due to the specific characteristics of each
project. This can be applied, for example, to aspects of urban planning. The economic value
of an urban housing project for lower-income residents may be higher if it is located near
environmentally beneficial features (such as parks) and public amenities (such as schools and
taxi ranks), compared with one located near environmental hazards or far from public
amenities.
Paper IV is an application of the hedonic pricing methodology (Rosen 1974) to study the
determinants of property prices in a low-income area of Namibia’s capital, Windhoek. The
methodology uses property prices to estimate buyers’ implicit valuation of a property’s
attributes (such as access to public services, proximity to environmentally beneficial or
detrimental features) when trading takes place. Local authorities in Namibia are responsible
for the provision, operation and maintenance of most municipal infrastructure and services.
Although this simplifies the planning, design, financing and implementation of initiatives for
9
Michael N Humavindu
upgrading poor settlements as well as the development of low-cost housing schemes, for
example, it places considerable responsibility on local authorities to ensure efficient urban
planning.
Frayne and Pendleton (2001) allude to the high rates of internal migration and urbanisation in
Namibia. This puts enormous responsibility on local authorities to ensure efficient urban
planning and that the investments made are prudent. Failure to account for this might lead to
developing residential areas or serviced plots alongside environmental ‘bads’ such as garbage
dumps, which could prove detrimental to social welfare if households attach importance to
such issues.
The valuation of environmental assets and services would underline their economic
importance and make a case for their conservation. The incorporation of shadow prices of
environmental costs and benefits in planning falls within the ambit of non-market valuation
in the environmental economics discipline. Non-market valuation is a measure of the
willingness to pay for the value of unpriced environmental goods and services.
Generally, for non-marketable items (those that cannot be sold or bought), two groups of
valuation methods are employed (see Hufschmidt et al. 1983). The first group is the revealed
preference approach, in which consumer behaviour towards environmental goods is analysed
and values are inferred. Peoples’ preferences are revealed by their choices. The second group
of methods is applicable when consumer behaviour towards environmental goods cannot be
observed. The solution then is to apply what is termed as the stated preference approach. The
approach rests on the simple premise of putting hypothetical questions to consumers.
10
Introduction and summary
The hedonic pricing method used here is an example of a revealed preference approach,
which postulates that the price of a commodity is related to its characteristics. Therefore,
variations in demand for a commodity (such as a house) can be statistically related to its
attributes (e.g. local air quality, amenities). The hedonic pricing method is used to estimate
the value of environmental amenities that affect the prices of marketed goods. Most
applications use residential housing prices to estimate the value of environmental amenities.
This method is based on the assumption that people value the characteristics of a commodity,
or the services it provides, rather than the commodity itself. Thus, prices will reflect the value
of a set of characteristics, including environmental characteristics, that people consider
important when purchasing the commodity. Property prices can, therefore, be used to
estimate local shadow prices for environmental characteristics even though those
characteristics are not traded directly.
2.
Summary of the papers
2.1
Paper I: Estimating national economic parameters for Namibia
In the first paper of this thesis, shadow prices of capital, labour and foreign exchange for the
Namibian economy are estimated. Although the use of shadow prices is essential for sound
developmental planning, the application of shadow pricing in Namibia has been limited or
virtually non-existent. The interest in deriving Namibian shadow prices arises from both
practical and academic points of view.
In practical terms, recognising the need for large-scale investments to drive economic growth
prompts the need to apply shadow prices, in order to ensure scarce resources are optimally
11
Michael N Humavindu
allocated. From an academic point of view, the Namibian economy exhibits special features
that support the need to estimate national parameters. A highly uneven income distribution, a
large informal economy, and minimum wages in certain sectors all validate the necessity of
estimating shadow wage rates. Unlike most other developing countries, Namibia is a net
capital exporter. Although the economy has high domestic savings, the lack of domestic
investment opportunities leads to a capital outflow amounting to 10% of GDP annually. The
shadow price of capital can then be reasonably expected to be low. Finally, Namibia’s
membership of the Southern African Customs Union (SACU) and the Common Monetary
Area (CMA) might affect the estimation of the shadow price of foreign exchange in Namibia.
SACU groups together Botswana, Lesotho, Namibia, Swaziland (known as the BLNS
countries) and South Africa, and applies a common external tariff. The SACU Agreement has
recently been renegotiated, with key elements revised and given a new focus, in the light of
the need to allow BLNS countries greater say in the determination and administration of
SACU tariffs. The CMA comprises SACU countries, excluding Botswana, and is a monetary
area with a centralised monetary policy aimed at achieving greater financial stability for the
southern African region. The monetary policy is controlled by South Africa, and all other
CMA currencies are pegged to the South African Rand.
In principle, there should be two Shadow Exchange Rates (SERs): one for convertible
currency external to the CMA, and one for Rand-based currencies, which would have a
shadow exchange rate of 1 since there are no trade restrictions between CMA countries. The
SER calculated in this paper is applicable to transactions with countries outside the CMA, but
not to the foreign content of goods purchased from South Africa.
12
Introduction and summary
The results suggest that the economic opportunity cost of capital is 7.2% in Namibia. The
economic costs of Namibian labour, as a share of financial costs, are 32% for urban semi- and
unskilled labour, and 54% for rural semi- and unskilled labour. The economic cost of foreign
labour, as a share of financial costs, is 59%. The estimated shadow exchange rate factor is 4%
for the Namibian economy.
2.2
Paper II: Estimating Namibian shadow prices within a semi-input–output
framework
The purpose of Paper II is to estimate the sectoral shadow prices (Accounting Price Ratios, or
APRs) at the national level, using the Semi-Input–Output (SIO) Technique. In contrast to
estimates of shadow prices in Paper I, which are limited to a few aggregate shadow prices for
capital, labour and foreign exchange, the application of an SIO analysis in Paper II permits
the calculation of more shadow prices for the Namibian economy. Utilising the SIO analysis,
one is able to (Schohl 1979) –
x
readily derive shadow prices for many different sectors of the economy, and
x
include the direct and indirect effects of protection on the conversion factors of
typically non-traded goods and services.
This larger set of APRs is beneficial for project analysis within sectoral projects and, at the
same time, should improve overall appraisal results. In general, the APRs for tradables and
non-tradables are expected to fall within the range close to unity or less than unity,
respectively. The following table summarises results from the estimations:
13
Michael N Humavindu
Table 1: APR estimates for economic sectors, Namibia
Economic sectors
1. Tradables
APR
2. Non-tradables
APR
Commercial Agriculture – Cereal
0.87
Traditional agriculture
0.66
Commercial Agriculture – Other crops
0.91
Electricity
0.88
Commercial Agriculture – Animal products
0.97
Water
1.13
Fishing
1.00
Trade and repairs
0.53
Mining
1.00
Hotels and restaurants
0.50
Meat processing
1.00
Communications
0.95
Fish processing
1.00
Finance and insurance
0.62
Grain milling
0.91
Other private services
0.84
Beverages and other food processing
0.85
Government services
0.95
Textiles
0.85
Light manufacturing
0.95
Heavy manufacturing
0.87
Construction
1.00
Transport
1.00
Market – Real estate and business services
1.00
Tourism – Non-residents
1.00
Petroleum products
0.90
The results shows that most tradable sectors such as fishing and mining have APRs equal to
or closer to 1, with deviations explained by import tariffs. The APRs for non-tradable sectors
exhibit greater variation, with the water sector having the highest – reflecting the scarcity of
water.
14
Introduction and summary
2.3
Paper III: Integration and volatility spillovers in African equity markets:
Evidence from Namibia and South Africa
The third paper examines the integration between the Namibian Stock Exchange (NSX) and
the Johannesburg Stock Exchange (JSE). The study uses daily stock data to analyse returns
and volatility between the two equity markets. The methodology employed consists of unit
root tests, cointegration analysis, and volatility modelling. The strong economic and historical
ties between South Africa and Namibia from the apartheid era suggest that there should be
strong integration. Indeed, previous empirical work reports strong integration between
Namibian and South African equity markets.
The paper differs from previous empirical work in that it focuses on the local Namibian
index, which does not contain dual-listed stocks. Dual-listed stocks are listed on both the JSE
and NSX, where they will expectedly have the same returns and volatility on both exchanges.
However, it is not necessarily the case that stocks that are only listed on the NSX will also be
highly correlated with stocks on the JSE. Thus, the paper examines integration between the
local NSX index and the JSE index. The sample covers the period from 4 January 1999 to 20
March 2003.
The results show that, when dual-listed stocks are excluded, the two markets exhibit very low
correlations, and no evidence of a linear relationship could be found between the two equity
markets. Moreover, volatility analysis does not provide any evidence of volatility spillover
from the JSE to the local NSX. The results suggest that the Namibian local equity index can
be a risk diversification tool for regional portfolio diversification in southern Africa.
15
Michael N Humavindu
The constraints of not having within-day-trading data from the NSX hamper the further
investigations of aspects of simultaneity in the returns. The availability of within-day-trading
data would have permitted the analysis of whether there are unidirectional causations within
the day between the two stock markets. Brännäs et al (2007) mention that simultaneity is
most likely to arise in closely related markets due to geographic proximity, common
institutional set up and the presence of large common traders. The presence of dual listed
stocks on both the JSE and NSX is an additional reason to expect simultaneity. Further
extension of this work might, in addition to investigating simultaneity, specify alternative
models to the one applied here to investigate volatility in the two markets returns.
2.4
Paper IV: Hedonic pricing in Windhoek townships
The fourth paper attempts to determine whether property prices in several low-income areas
of Namibia’s capital, Windhoek, are affected either by positive or negative attributes, and
applies the hedonic pricing method in this analysis. Hedonic pricing, as previously stated,
involves the implicit price of attributes or characteristics of a commodity rather than the price
of the commodity itself. Hedonic pricing models are used to infer the demand for attributes of
environmental quality, through the analysis of marketed goods whose value partly depends
on these attributes. The methodology is generally applied for the valuation of environmental
goods, property and water, and the implicit price of attributes and characteristics of marketed
goods in general. The general assumptions of such a model are that all the goods or services
brought to the market should be clearly visible, and that property values and the implicit price
of attributes or characteristics should be treated as a single market. Under these assumptions,
the price of any residence can be described as a function of the environmental, structural, and
neighbourhood characteristics of the location of the residence in question. The hedonic model
16
Introduction and summary
can, thus, give a realistic estimation of the environmental values attached by households to
attributes, as model estimates are based on market information.
In this paper, we use property sales data obtained from the City of Windhoek municipality,
and apply the hedonic pricing model. Our findings are that – apart from housing quality,
access to the central business district, access to marketplaces, and access to transportation –
environmental quality has a large impact on property prices. Properties located close to a
garbage dump sell at considerable discounts, while properties located close to a combined
conservation and recreation area sell at premium prices. The results suggest, therefore, that
the hedonic pricing method can be usefully applied when studying township areas in
developing countries, and that this can clarify and emphasise the importance of
environmental factors that are otherwise frequently neglected in town planning for such
settlements.
17
Michael N Humavindu
REFERENCES
Adhikari, R (1986): “National economic parameters for Nepal”. Occasional Paper No. 9.
Bradford: Project Planning Centre for Developing Countries, University of Bradford.
Bank of Namibia (2003): Review of the domestic asset requirements (Regulation 28 and 34).
Windhoek: Research Department, Bank of Namibia.
Bank of Namibia (Various): Quarterly Bulletin. Windhoek: Bank of Namibia.
Brännäs, K, JG De Gooijer, C Lönnbark & A Soultanaeva (2007): “Simultaneity and
asymmetry of returns in the emerging Baltic state of stock exchanges”. Umeå
Economic Studies 725. Umeå: Umeå University.
Chen, N, R Roll & SA Ross (1986): “Economic forces and the stock market”. The Journal of
Business, 59(3):383–403.
Engle, RF & CWJ Granger (1987): “Co-integration and error correction: Representation,
estimation and testing”. Econometrica, 55(2):251–276.
Engle, R (2001): “GARCH 101: The use of ARCH/GARCH models in applied
econometrics”. Journal of Economic Perspectives, 15(4):157–168.
Frayne, B & W Pendleton (2001): “Migration in Namibia: Combining macro and micro
approaches to research design and analysis”. International Migration Review,
35(4):1054–1085.
Harberger, A (1978): “On the use of distributional weights in social cost-benefit analysis”.
Journal of Political Economy, 86(2):S87-S120.
Harris, R & R Sollis (2005): Applied time series modelling and forecasting. Chichester: John
Wiley & Sons.
18
Introduction and summary
Hufschmidt, MM, DE James, AD Meister, BT Bower & JA Dixon (1983): Environment,
natural systems, and development: An economic valuation guide. Baltimore, MD:
Johns Hopkins University Press.
Humavindu, MN (2002): “Economics without markets: Policy inferences from nature-based
tourism studies in Namibia”. DEA Research Discussion Paper 49. Windhoek:
Directorate of Environmental Affairs, Ministry of Environment and Tourism.
Humavindu, MN & S Masirembu (2001): “The economic value of urban recreation in
Namibia: Results of a pilot study at Avis and Goreangab Dams”. Windhoek:
Directorate of Environmental Affairs, Ministry of Environment and Tourism.
IMF/International Monetary Fund (2006): Financial Sector Assessment Program: Republic of
Namibia. Aide-Memoire. Washington, DC: IMF.
IMF/International Monetary Fund (2008): “Namibia: Selected issues and statistical
appendix”. IMF Country Report No. 08/82. Washington, DC: IMF.
Lee, JD, JB Park & TY Kim (2002): “Estimation of the shadow prices of pollutants with
production/environment inefficiency taken into account: A nonparametric directional
distance function approach”. Journal of Environmental Management, 64(4):365–375.
Little, I & J Mirrlees (1974): Project appraisal and planning for developing countries.
London: Heinemann.
National Planning Commission (2007): Guidelines for preparing the Third National
Development Plan (NDP3): 2007/08–2011/12. Windhoek: NPC.
Potts, D (2002): Project planning and analysis for development. London: Lynne Rienner
Publishers.
Potts, D, J Weiss, M Beyene, F Guta, H Kinfu & R Assegid (1998): National economic
parameters and economic analysis for the Public Investment Programme in Ethiopia.
Addis Ababa: Ministry of Economic Development and Co-operation.
19
Michael N Humavindu
Rosen, S (1974): “Hedonic prices and implicit markets: Product differentiation in pure
competition”. Journal of Political Economy, 82(1):34–55.
Saerbeck, R (1989): “National economic parameters for Botswana”. Research Monograph 1.
Bradford: Project Planning Centre for Developing Countries, University of Bradford.
Schohl, WW (1979): “Estimating shadow prices for Colombia in an input–output table
framework”. Staff Working Paper No. 357. Washington, DC: World Bank.
Squire, L & H van der Tak (1975): Economic analysis of projects. Baltimore, MD: Johns
Hopkins University Press.
Stage, J & CH Kirchner (2005): “An economic comparison of the commercial and
recreational linefisheries in Namibia”. African Journal of Marine Science, 27(3):577–
584.
UNIDO/United Nations Industrial Development Organisation (1972): Guidelines for project
evaluation. New York: UNIDO.
UNIDO/United Nations Industrial Development Organisation (1978): Guide to practical
project appraisal. New York: UNIDO.
UNIDO/United Nations Industrial Development Organisation (1980): Practical appraisal of
industrial projects: Application of social cost-benefit analysis in Pakistan. New York:
UNIDO.
UNIDO/United Nations Industrial Development Organisation (2003): Investment Project
Preparation and Appraisal. Vienna: UNIDO.
20
I
Estimating national economic parameters
for Namibia
Michael N Humavindu1
Development Bank of Namibia, PO Box 235, Windhoek, Namibia
[email protected]
Abstract
This paper estimates national economic parameters to be used for project appraisal in
Namibia. The shadow prices of capital, labour and foreign exchange are derived. The results
suggest that the economic opportunity cost of capital is 7.2%. The economic costs of
Namibian labour as a share of financial costs are 32% for urban semi- and unskilled labour,
and are 54% for rural semi- and unskilled labour. The economic costs of foreign labour as a
share of financial costs are 59%. The shadow exchange rate factor is estimated to be 4% for
the Namibian economy.
Keywords: shadow prices; discount rates; Namibia
1
The author wishes to acknowledge a research grant from the Jan Wallander and Tom Hedelius Foundation.
The author would also like to thank Jesper Stage, Karl-Gustaf Löfgren, David Potts, Kenneth Backlund, Jon
Barnes, Glenn Jenkins, Tomas Sjögren, Fredrik Carlsson and an anonymous reviewer for comments on earlier
versions of the paper. The usual disclaimer applies.
Estimating national economic parameters for Namibia
1. INTRODUCTION
The purpose of this paper is to estimate shadow prices of capital, labour and foreign
exchange for the Namibian economy. Shadow prices are defined as the opportunity costs of
inputs and outputs consumed or produced by a project (Potts 2002). The value that a
resource could have generated elsewhere in the economy is lost if the resource is moved to
a project. Therefore, shadow prices are calculated to take into account the true opportunity
costs of resources, inputs and any externalities resulting from a developing programme or
project.
In many markets, especially in developing countries, financial or market values differ from
their real economic values due to distortions brought about by imperfect or underdeveloped
markets, government protection policies, and other externalities (Behrman 1986). The most
emphasised distortions are with regard to unskilled labour, the cost of foreign exchange,
and the cost of financial capital.
Shadow pricing is then used to account for these distortions and value resources to
approximate their actual value. The use of unadjusted market prices for labour and capital
might lead to underestimating the real costs of capital-intensive projects and tend to
promote these at the expense of socially less costly labour-intensive projects. The existence
of high levels of nominal and effective tariff protection, in combination with import quotas
and overvalued exchange rates, discriminates against the agricultural sector in favour of the
import-substituting manufacturing sector. In addition to reflecting – incorrectly – the real
terms of trade between, for example, agriculture and industry, such distorted domestic
1
Michael N Humavindu
product prices tend to favour upper-income groups disproportionately in relation to
society’s lower-income groups.
Thus, the estimation of shadow prices is essential for the practical application of the
economic analysis of project evaluation. By way of cost-benefit analysis, project evaluation
aims to induce allocation efficiency in the use of a country’s resources (Campbell & Brown
2003).
Despite its importance for sound developmental planning, the application of shadow pricing
in Namibia has been limited or virtually non-existent. This is unfortunate as Namibia’s
development strategy, as encapsulated in the five-yearly National Development Plans and
in its Vision 2030, underpins the importance of development/investment programmes in
addressing the challenges of poverty, high unemployment and inequality, and low
industrialisation.2 Moreover, the launch in 2004 of the Development Bank of Namibia to
fund long-term infrastructure projects increases the need to understand the economic costs
and benefits of its funded projects. Potential large-scale projects such as the development of
the Kudu gas fields, transfrontier tourism parks, and other infrastructure projects would
need to be assessed on both financial and economic grounds. Thus, the practical application
of shadow pricing in the economic analyses of Namibia’s development projects would help
ensure that its scarce resources are optimally utilised, and would help attain the country’s
targets as set out in its development strategy.
2
Shadow prices are consistently used in the Ministry of Environment and Tourism (MET), but these are based
on educated guesswork rather than real estimates. This framework assumes the economic opportunity cost of
capital at 8%, an adjustment (up by 6%) to the value of tradable goods to reflect excess demand for foreign
exchange, and an adjustment (down by 65%) to unskilled labour costs to reflect unemployment (Barnes 1994).
2
Estimating national economic parameters for Namibia
Estimating shadow prices for the Namibian economy is also interesting from an academic
point of view. Namibia has special features that are not commonly found among other
developing countries. Namibia’s gross domestic product (GDP) per capita of US$3,100, at
2005 market exchange rates, is relatively high for a developing country. However,
according to the World Bank’s World Development Indicators for 2006, Namibia has the
world’s highest Gini index (74.3, compared with Botswana’s 63 and South Africa’s 57.8).3
This implies an uneven income distribution that amplifies the interest to estimate shadow
wage rates.
The Namibian economy has a large service sector (around 58.7% of GDP), which is
unusual for a developing country. In addition, independence in 1990 brought considerable
changes to the economy’s external and internal migration patterns, especially in relation to
the labour market. According to Frayne and Pendleton (2001), internal migration and
urbanisation in Namibia is growing rapidly, and is driven largely by employment
opportunities in urban centres. In the 1990s, the population of Windhoek, the capital city,
grew at an average annual rate of 5.4%. Overall, no substantial research has been done in
Namibia on either the scale or the possible consequences of skills emigration. However,
according to preliminary analyses by Frayne and Pendleton (2001, 2002) and the Migration
Dialogue for Southern Africa (MIDSA 2006), Namibian migratory labour (both skilled and
unskilled) to South Africa (SA) and other neighbouring countries is very limited: the
overall net migration is estimated at 0.47 per 1,000 members of the population.
3
The Gini index is a measure of the degree of income inequality.
3
Michael N Humavindu
Furthermore, unlike most other developing countries, Namibia is a net capital exporter.
Although the economy has high domestic savings, these flow out mostly to SA to seek
higher returns. The lack of domestic investment opportunities is cited as one reason for
persistent capital outflows (Fitch Ratings 2005). These capital outflows amount to 10% of
GDP annually, and continue unabated.4
The linkages to SA are not only restricted to Namibia’s capital outflows. The two
economies are members in regional groupings such as Southern African Customs Union
(SACU), the Southern African Development Community (SADC), and the Common
Monetary Area (CMA). Namibia’s currency is pegged to the SA Rand, while 82% of her
total imports are from SA. Some 26% of Namibia’s total exports go to SA. The SA
economy, being the regional economic powerhouse, is approximately 30 times the size of
Namibia’s.
Recent work by Harberger et al. (2003), Kuo et al. (2003) and Bicak et al. (2004) has
estimated shadow prices for the SA economy for labour, capital and foreign exchange.
Since these two countries share similar historical political ties and a current close economic
relationship, it would be interesting to compare the results from this work with those from
4
An anonymous reviewer suggests that this could be a symptom of ‘Dutch disease’. However, unlike most
other primary product exporters, the Dutch disease phenomenon appears to be a limited risk to Namibia (IMF
2008). This is because Namibian mineral exports have a relatively modest and decreasing share of GDP (20–
25%). This share actually overstates domestic expenditures by the mineral sector, as it imports most of its
capital equipment and its labour costs are very low (it contributes 2% of national employment). In addition,
fiscal revenues from the sector average around 2–3% of GDP. Thus, domestic pricing pressures from the
sector are relatively modest, and wage pressures are unlikely to be large.
4
Estimating national economic parameters for Namibia
the SA studies. However, given Namibia’s special features not common to a developing
economy, it can reasonably be expected for estimates of the two economies’ national
parameters to be different.
This work will be the first formal exercise to estimate shadow prices for the Namibian
economy. The paper is structured as follows: Section 2 discusses economic features
pertinent to the estimation of Namibian shadow prices; Section 3 treats approaches to
shadow pricing as well as the methodology to be employed; Section 4 describes the data
employed as well as the assumptions used for each estimate; Section 5 presents the results;
and Section 6 concludes the discussion.
2.
FEATURES OF THE ECONOMY PERTINENT TO AN ESTIMATION OF
SHADOW PRICES
2.1
Capital market dynamics
The Namibian financial markets exhibit special features that will affect the estimation of a
shadow price of capital. As mentioned earlier, overall limited investment opportunities in
domestic financial markets have led to sizable outflows of Namibian savings into the liquid
and relatively developed South African markets. Membership in the CMA also allows for
free capital flows, and requires Namibia to conform to South African exchange control
practices for countries outside the CMA. These outflows averaged around N$1.8 billion per
year from 1990–1994, and accelerated to about N$2.4 billion per year from 1995 to 2007.
Net outflows in both portfolio and other investments drive the capital outflows.
5
Michael N Humavindu
The Namibian economy is primarily resource-based and, thus, has some investments that
are highly profitable owing to resource rents. Resource rents are economic profits that are
obtained by utilising natural resources. These rents exist due to the scarcity of the natural
resources in question. Such rents can be an important source of development finance, and
countries like Botswana and Malaysia have successfully leveraged natural resources this
way. However, in sectors that do not have resource rents, the marginal product of capital
appears to drop sharply since many funds are invested outside Namibia.
To stem capital outflows, the Namibian authorities have followed a two-pronged strategy:
firstly, imposing regulatory controls to restrict capital outflows, and secondly, developing
domestic markets to provide institutional investors with assets denominated in Namibia
Dollars. The latter strategy is still in its infancy and has not been developed. In the mid
1990s, the Namibian authorities raised regulatory requirements for both the insurance and
pension fund industries (Regulations 15 and 28, respectively), so that 35% of the assets
under their management had to be domestic assets (up from an earlier 10%). This action
contributed to the growth of the Namibian Stock Exchange due to an increase in dual
listings by South African companies. However, even investments in such dual-listed
companies were unable to contain capital outflows, and the regulation may not have had
much impact on the real economy. As a result, government has proposed further changes to
tighten the domestic asset requirements. A 5% minimum for unlisted investments and a
10% maximum on dual-listed shares were among the new proposals gazetted on 4 February
2008.
6
Estimating national economic parameters for Namibia
2.2
Labour market dynamics
The Namibian labour market is governed by a policy framework that includes a Labour
Act, a Social Security Act, an Employment Policy, an Affirmative Action (Employment)
Act, and incentives for investment and training. However, on balance, unemployment and
underemployment remain high. According to the latest Labour Force Survey, conducted in
2004, unemployment was estimated at 36%. The Bank of Namibia Annual Report for 2004
states that underemployment was estimated at 15% of the employed population. Motinga
and Tutalife (2006) indicate that Namibia created a mere 22,000 formal jobs between 1991
and 2001. Unemployment falls disproportionately on the youth and the unskilled
workforce, while the duration of unemployment is longer in rural areas, and can vary
between six months and two years (ibid).
There is also evidence of wage inequality between the skilled and unskilled. Motinga and
Mohammed (2002) calculated that the average unskilled person earns 3% of the wages and
salaries of top management, and less than 50% of what the average skilled person earns.
Westergaard-Nielsen et al. (2003) confirm the huge wage differentials between skilled and
unskilled labour. Although there is no formal minimum wage legislation, some industryspecific wage agreements do contain stipulations for minimum wages, namely the
construction, agriculture, and security industries. There is also a large informal economy
employing at least 133,000 people, of whom 64% are young people. Remuneration in this
sector is very low, and there are no benefits such as social security or medical aid.
The presence of a large informal economy and minimum wages, both of which lead to
Namibian wages being set higher than the economic opportunity cost of labour, justifies the
7
Michael N Humavindu
case for such an economic adjustment on the grounds of imperfections in the labour
markets. The informal economy, which consists of large numbers of small-scale businesses,
can be reasonably assumed to be a sector with market-clearing wages. In the formal sector,
however, the presence of minimum wages and collective bargaining – and, possibly,
efficiency wage issues – leads to wages above the market-clearing levels that exist in the
informal economy. As a result, a portion of the 36% unemployed Namibians would prefer
formal jobs, but cannot get them due to the presence of these distortions.
2.3
Issues in estimating the foreign exchange premium
Namibia’s participation in SACU affects the estimation of the shadow price of foreign
exchange (Shadow Exchange Rates, or SERs). SACU groups Botswana, Lesotho, Namibia,
SA and Swaziland together under a common external tariff. All customs and excise duties
collected by the five SACU members are combined in a Common Revenue Pool (CRP),
and distributed to them according to a Revenue Sharing Formula (RSF). The sharing of the
revenue from customs duties is determined on the basis of each country’s percentage share
of total intra-SACU imports, excluding re-exports, and not on the basis of SACU imports
from the rest of the world (Flatters & Stern 2005; Kirk & Stern 2005).
Some 82% of Namibian imports are from SA, which increases Namibia’s share of revenue
from the SACU system (due to the RSF’s intra-SACU imports rule). Namibian imports
from outside SACU (the remaining 20% of her total imports) are subject to SACU tariffs,
but generate very little extra SACU revenue for Namibia: tariff revenues are paid into the
SACU system, and Namibia only gets a small portion of that. Most Namibian exports are to
countries outside SACU, which therefore do not affect her revenue share from SACU.
8
Estimating national economic parameters for Namibia
Thus, since SACU revenue for Namibia is effectively not linked to the country’s out-ofSACU imports, it can be argued that SACU receipts are not relevant to the determination of
the shadow exchange rate since they are essentially intergovernmental transfers and do not
directly affect the relationship between prices of traded goods at world prices and their
domestic prices. Moreover, the SACU revenue pool is gradually declining due to
continuing trade negotiations at multilateral and regional levels.
Namibia is part of the CMA, which also includes Lesotho, SA and Swaziland.5 Apart from
Botswana, the CMA has four of the same member countries as SACU; thus, there should be
two SERs: one for convertible currency external to the CMA, and one for Rand-based
currencies that would have an SER of 1 since there are no trade restrictions between CMA
members. The SER to be calculated in this work, therefore, is applicable to transactions
with countries outside the CMA, but not to the foreign content of goods purchased from
SA. In principle, one would expect the SER – in relation to external economies – to be
similar for all members of the CMA because they all use the same tariff structure. However,
there might be some variation due to differences in the structure of imports.
5
The CMA is described as an area of coordination between the monetary and exchange rate policies of its
members under the Multilateral Monetary Agreement of 1992. Under the CMA, the Namibian currency is
linked one-to-one to the South African Rand, which is also legal tender. The CMA also guarantees free capital
flows among member countries, and guarantees access for Namibian government and financial institutions to
South Africa’s financial markets. See also Tjirongo (1995) and Vollan (2000).
9
Michael N Humavindu
3.
ANALYTICAL FRAMEWORK (METHODOLOGY)
This section describes the analytical framework to be used in estimating shadow prices in
Namibia. Generally, there are two approaches to shadow pricing that hinge on the
assumption of the existence of market distortions (Medalla 1982). The first approach may
be generalised as an attempt to estimate shadow prices associated with a first-best optimum.
In this approach, if market and shadow prices diverge due to policy failures, then the
appropriate shadow prices would be the equilibrium prices that would prevail if the
distortions were removed. However, if the divergence is caused by market rather than
policy failures, then the absence of first-best corrective measures is itself the essence of the
problem of non-optimality. The work by Tinbergen (1958) and Bacha and Taylor (1971) in
the case of shadow pricing of foreign exchange is associated with this first approach. As
Medalla (1982) states, however, this approach is not yet feasible for shadow pricing
primary factors such as capital and labour due to inadequate techniques and data.
The second approach treats present distortions as given and assumes that they might persist
over the long run (Medalla 1982). Shadow pricing is then a problem of deriving dual
solutions to the welfare optimisation problem, while the distortions are treated as
constraints. Under this approach the optimisation problem is usually not formally specified,
but it forms the conceptual framework for shadow pricing rules. The resulting shadow
prices are referred to as second-best shadow prices, representing social costs and benefits of
inputs at the second-best optimum. This approach is associated with the work of Little and
Mirrlees (1969), Harberger (1972), and Dasgupta et al. (1972).
10
Estimating national economic parameters for Namibia
In this paper we follow Harberger’s (1972) approach for two principal reasons. Firstly,
according to Khan (1979), this is the correct method of estimating the shadow discount rate,
namely where the marginal social value is not equal to the marginal social cost of funds at
the market equilibrium due to the presence of various distortions. Finally, and most
importantly, utilising this approach will enable comparison of the results with those of
Harberger et al. (2003), Kuo et al. (2003). and Bicak et al. (2004) for South Africa.
3.1
The discount rate
The economic literature advances four main methods of computing the discount rate .These
are the Social Rate of Time Preference (SRTP), the Weighted Opportunity Cost of Capital
(SOC), the Shadow Price of Capital (SPC), and the Economic Opportunity Cost of Capital
(EOCK).6 In terms of applicability, only the SRTP and the EOCK are feasible for the
Namibian estimations. However, a brief review of first three methods is presented, with a
more substantial review of the EOCK method, which allows for comparison with the South
African work.
3.1.1 The Social Rate of Time Preference approach
The SRTP approach is where the discount rate is composed of two factors: the first is a pure
rate of time preference based on people’s desire to gain short-term gratification, and the
6
See Boardman et al. (2001), Boscolo et al. (1998), Percoco & Nijkamp (2006), Powers (2003), and Zhuang
et al. (2007). These sources offer an excellent and detailed review of the major methods on estimating the
discount rate.
11
Michael N Humavindu
second an assumption that per capita consumption will grow over time. The formula for the
SRTP is given by the following equation:
r
U Tg
(1)
where U is the utility discount rate, T is the absolute value of the elasticity of marginal
utility of consumption, and g is the projected long-run annual growth of real consumption
per capita. The advantage of the SRTP approach is its applicability to the Namibian work
on discount rates.
3.1.2 The Weighted Social Opportunity Cost of Capital approach
The SOC approach is grounded on the notion that public investment crowds out private
investment, thus producing the need to account for the opportunity cost of the use of
resources used in the public project, and which could be used by the private sector. The
SOC could be approximated by the marginal pre-tax rate of return on riskless private
investments.
Zhuang et al. (2007) mention that a good proxy to be used is the real pre-tax rate of toprated corporate bonds. The application of the SOC is still contentious, however, both on
practical and theoretical grounds. A practical difficulty arises since the computation of the
SOC relies on a vast array of possible private sector interest rates which may not be readily
available. Some theoretical objections to the SOC follow the argument that the private
sector return may reflect individual rather than societal premium for risk. This argument is
based on the perspective that people may be more willing to accept risks as a group than as
12
Estimating national economic parameters for Namibia
individuals. Thus, a rate based purely on the pre-tax return in investment may overestimate
the discount rate: thereby making it more difficult to obtain a benefit-cost ratio of greater
than 1, particularly for projects of a longer tenure (Powers 2003).
3.1.3 The Shadow Price of Capital approach
This SPC approach postulates that, while the costs of a public project can displace private
investments, its benefits can also be reinvested in the private sector. Thus, it proposes to
convert the gains or losses from an investment project into consumption equivalents. The
proper conversion rate is then the shadow price of capital (Percoco & Nijkamp 2006).
Estimating the SPC is relatively simple if it is assumed that each dollar invested today
yields a perpetual return that is entirely consumed (Boscolo et al. 1998). Thus, the present
vale of the annual flow of consumption is given by /i, where i is the SRTP. By implication,
/i is the shadow price of investments in terms of consumption. A simple formula that
applies when investment returns are perpetual but a proportion of the annual return is
reinvested is derived as –
SPC
1 s J
(2)
i sJ
where =(1+)/(1+i), s is the marginal propensity to save, and s<1. The shadow price
increases with the fraction of invested. The SPC is conceptually correct as it allows the
use of the SRTP as the social discount rate without ignoring the opportunity cost of
displaced investment. However, its practical applicability is constrained due to its stringent
information requirements.
13
Michael N Humavindu
3.1.4 The Economic Opportunity Cost of Capital
Finally, the EOCK approach postulates that in a small, open, developing economy like
Namibia’s, there are three alternative sources of public funds. The first is from individual
savers who take resources that would have been spent on private consumption and instead
then lead to an increase in domestic savings. The second source is from additional foreign
capital inflows. The third is from resources whose investment has either been displaced or
postponed by the project’s extraction of funds from the capital market (Harberger 1972).
Based on these three alternative sources of public funds, the economic cost of capital can be
estimated as a weighted average of the rate of time preference applicable to –
x
additional savings
x
the marginal cost of additional foreign inflows, and
x
the rate of return on displaced or postponed investments.
In general, various distortions are associated with each of the three alternative sources of
funds.
If the weights of these three sources are expressed in terms of elasticities of demand and
supply of funds with respect to changes in interest rates, the economic opportunity cost of
capital can be calculated as follows (a derivation of this is given in Appendix 1):
EOCK
H s ( S r / S t ) x J H f ( S f / S t ) x MC f K x S
H s ( Sr / St ) H f ( S f / St ) K
14
(3)
Estimating national economic parameters for Namibia
For a country such as Namibia, with a fixed exchange rate, high capital mobility, and a
highly elastic supply of foreign funds, Zerbe and Dively (1994) point out that the social
discount rate will be equal to the international borrowing rate. For Namibia, where the
foreign funds are domestic savings, this will be the foreign lending rate (approximately
equivalent to South African bond returns, or the returns on other South African financial
instruments in which surplus Namibian assets are placed). Thus, in the standard EOCK
formula (Equation 3 above), the elasticity of foreign funds becomes extremely high
compared with the other elasticities. Equation 3 can, therefore, be simplified as follows:
EOCK
H f ( S f / S t ) x MC f
H f (S f / S t )
MC f
(4)
The EOCK in (4) essentially equals the real rate of return from investing Namibian funds in
South African long-term financial instruments. South African assets constitute
approximately 80% of both total and portfolio investments from Namibia (IMF 2008).
Therefore, a good proxy for the amended EOCK will be the average rate of return on longterm investments in South African bond instruments.
3.2
Economic Opportunity Cost of Labour
The EOCL reflects the value to the economy of the set of activities given up by the
workers, including the non-market costs (or benefits) associated when they change
employment from one project to another (Harberger & Jenkins 2002). Two approaches are
generally applied in estimating the EOCL: the value of marginal product of labour
foregone, and the supply price of labour (Bicak et al. 2004).
15
Michael N Humavindu
Under the value of marginal product of labour forgone approach, the EOCL is estimated by
starting with the gross-of-tax alternative wage earned in previous employment by the labour
hired for the new project (marginal product foregone), and then adjusting for differences in
other costs and benefits. Under the supply price of labour approach, the EOCL is
determined by starting with the gross-of-tax market wage (the supply price) required to
attract sufficient workers to the project, and then adjusting for distortions such as taxes and
subsidies. The two approaches have different data requirements, levels of computational
complexity, and hence, different degrees of operational usefulness (Bicak et al. 2004).
However, it can be shown that, theoretically, the two approaches will produce the same
result in estimating the EOCL. Since the supply price of labour is more straightforward and
easier to use under a wide variety of conditions in the labour market, and the two
approaches are equivalent when data are available, the supply price approach is used.
Bicak et al. (2004) also use the supply price of labour approach, making it easy to compare
the South African and Namibian results. It appears that the Namibian labour market does
not feature any special characteristics other than those of high wage inequality between
skilled and unskilled labour, and the large informal economy. There is little international
migration, although interregional migration to urban areas is high.7 Thus, a new project is
most likely to attract workers from both the formal and informal sectors, as well as some
foreign labour, if needed; but it may also attract some skilled Namibians currently working
in South Africa.
7
Harris and Todaro (1970) postulate that high urban unemployment rates could be explained by rationally
behaving unskilled rural migrants seeking to maximise expected income. According to this model, more than
one rural worker is likely to migrate for each new job created in the urban sector. The effect of this is that the
opportunity cost of the new urban job is greater than the marginal product of one rural worker.
16
Estimating national economic parameters for Namibia
It appears that skilled labour is in scarce supply in Namibia, with very little – if any –
unemployment experienced in this sector (LARRI 2005, 2006a; Marope 2005). Managers
and professionals earn annual remunerations of between N$250,000 to N$400,000. Around
4% unemployment is found among those with a university education. In such a case, the
opportunity cost of skilled labour is assumed to equal the domestic market wage (Potts
2002). In this paper, therefore, we concentrate on the shadow prices for unskilled labour
and for foreign labour.
The presence of a large informal economy presents the opportunity to determine the free
market wage at which everyone could work. From the Labour Resource and Research
Institute (LARRI 2006b), the free market wage can be estimated at N$175 per month.
Certain Namibian industries, as alluded to in Section 2 above, have minimum wages, and
many wages are determined by collective bargaining agreements. In such markets, the wage
rates are above their market clearing rates (Bicak et al. 2004). Because of the minimum
wage rates, chronic unemployment exists in this segment of the labour market.
The
illustration in Figure 1 shows how the EOCL for protected jobs can be determined under
the conditions of a linear supply curve and a perfectly elastic demand for labour in the
informal economy:
17
Michael N Humavindu
Wage
Rate
W
S
i
p
Di
Wi
W
0
O
Li
Labour
Li+Lq
Figure 1: Estimating the Economic Opportunity Cost of Labour for protected sector jobs
Let Wp be the protected sector wage, and let the supply curve of labour for those who are
not formally employed be given by WoSi. Let Li be the people who are willing to work at
Wi, and let Lq be quasi-unemployed willing to work at Wp but not at Wi. To simplify the
analysis, Wi is assumed to be the free market at which everyone could work if they wished.
The intersection of this supply curve and the free market wage rate of Wi determine the
number of people willing to work at this wage, or Li in Figure 1. In the Namibian case, Wi
would be the informal market wage.
When a project creates a demand for protected workers, such demand will be met partly by
those working in the free market (i.e. the informal economy in our case), and partly by
quasi-voluntarily unemployed workers (Bicak et al. 2004). If it were assumed that workers
are recruited randomly from among all those willing to work for the protected sector wage,
the economic cost of these jobs would be measured by the weighted average of the free
market wage and the supply price of the quasi-voluntarily unemployed. The EOCL will
18
Estimating national economic parameters for Namibia
then fall between the free market wage (i.e. the informal economy wage in our case) and the
protected wage rate. In the case of linear supply curves, the average supply price of the
quasi-voluntarily unemployed is measured by (Wi + Wp)/2. If any tax adjustments are
ignored, then the EOCL for protected sector jobs can be expressed as follows:
EOCLp = f1 Wi + f2 (Wi + Wp)/2
(5)
where f1 and f2, respectively, represent the proportions of the project jobs being filled by
those now working in the informal economy and those filled by unemployed individuals
who were waiting for new protected project jobs to become available.
The EOCL for skilled foreign labour will be measured by the net-of-tax wage that the
worker receives in Namibia, plus an adjustment for the foreign exchange premium that is an
additional cost on the share of wages the foreign worker remits back home. A second
adjustment is related to the goods and services that foreign workers consume in Namibia. If
foreign workers pay any excise or value added taxes on the goods they purchase, these
taxes should be deducted from the cost of foreign labour, as they do not represent a cost to
the Namibian economy. In some cases, temporary foreign workers might receive subsidised
housing or health benefits, for example. These should be added to the EOCL. Combining
these factors, the economic opportunity cost of labour for foreign workers (EOCLF) can be
estimated as follows:
EOCLF=WF(1-tF)+WF(1-tF)R[(Ee/Em)-1]-WF(1-tF)(1-R)tVAT
19
(6)
Michael N Humavindu
where WF is the gross-of-tax wage of foreign labour, tF is the rate of personal income tax
levied by the host country on foreign wages and salaries, R is the proportion of the net-oftax income repatriated by foreign labour, Ee is the economic exchange rate, Em is the market
exchange rate, and tVAT is the average rate of value added tax paid.
For labour from South Africa – the main source of skilled foreign labour – coming to work
in Namibia, Equation 6 can be rewritten as follows (since Ee/Em=1):
EOCLFRSA=WF(1-tF)+WFRSA(1-tF)(1-R)tVAT
(7)
Similarly, for Namibian skilled labour attracted back home by the project from out-ofcountry employment, the EOCL will need to adjust for a loss of remittances:
EOCLskilled Nam labour=WN(1-tRSA)R
3.3
(8)
Economic Opportunity Cost of Foreign Exchange
The wedge between the Shadow Exchange Rate (SER) and the Official Exchange Rate
(OER) can be attributed to a combination of two factors: disequilibria in the balance of
payments (BOP) and in the protection structure (Medalla & Powers 1984). Namibia does
not suffer from a BOP disequilibrium, but does have trade restrictions through SACU. An
SER higher than the OER reflects the premium placed on foreign exchange (used or
produced) when evaluating projects to correct the distorted relative prices between traded
and non-traded commodities. A higher SER does not suggest devaluation but rather
revaluation – to the exact degree of the SER estimate. This distortion in relative prices
20
Estimating national economic parameters for Namibia
arises from the protection system (and BOP disequilibrium) and not only affects price
relationships among tradable goods, but also distorts the prices of tradables relative to nontradables. Among tradable commodities, relative price distortion may be corrected in
project evaluation by using their relative border prices. However, further correction is
needed for the price distortion between tradables and non-tradables. This, in essence, is the
role of the SER in project evaluation. It serves as the conversion factor for non-tradables,
making their prices consistent with the border prices of tradables. One would ideally prefer
to compute a specific conversion factor for each non-tradable rather than use a standard
conversion factor such as the SER, but due to the practicalities involved in decomposing
non-tradables into their tradable and primary factor components, the SER is easier to
compute.
Lagman-Martin (2004) mentions three alternative approaches to estimating the SER. These
approaches are generally based on converting the OER to the SER through a conversion
factor known as the SER factor (SERF). The first approach is employed where an economy
enjoys balanced trade. The formula applied involves calculating the SER based on the
tariff-adjusted OER, weighted according to import–export shares. A second approach takes
into account the sustainability of the country’s trade imbalance through an assessment of
the Equilibrium Exchange Rate (EER). The use of the EER rather than the OER emphasises
the long-term stability of the exchange rate because of its significant effect on project
performance. Finally, in the third approach, when tariffs represent the only distortion to
trade and there are no distortions in factor or commodity prices, the SERF can be
approximated by 1 plus the weighted average tariff rate. This approach is consistent with
the accepted definition of the SER as the weighted average of the demand price for foreign
exchange paid by importers and the supply price of foreign exchange received by exporters.
21
Michael N Humavindu
This simple trade-weighted formula can be represented as (Potts 2002, Lagman-Martin
2004) –
SERF
M ( cif ) X ( fob )
(M Mt Ms ) (X Xt Xs )
TT
TT NTt
(9)
where M is the total value of imports (cif – cost, insurance, freight), X is the total value of
exports (fob – free on board), Mt is total value of import taxes, Xt is the total value of export
taxes, TT is the total value of trade, NTt is the total value of net trade taxes, and Ms and Xs
represent import and export subsidies, respectively.
Other, more complex, formulas for the SER can be derived if data are available to indicate
the types of imports or exports that change with a concomitant change in the availability of
foreign exchange. Such formulas use the elasticity of demand for imports and exports with
respect to changes in foreign exchange availability to provide weights for different export
and import categories. It is usually very difficult to obtain reliable information on these
elasticities, so the simple weighted formulas are commonly used. Harberger et al. (2003)
employ a general equilibrium model to estimate the SER for South Africa. Their approach
illustrates how the foreign exchange premium could be estimated in an economy where the
funds used to finance the purchase of tradable and non-tradable goods are obtained via the
capital markets. This framework ensures that all repercussions in the economy due to the
purchase of tradable goods for a project are taken into account in a consistent manner. Due
to data limitations, this work will employ the simple weighted trade formula presented
above. Other methods include using semi-input–output models in order to use the weighted
22
Estimating national economic parameters for Namibia
average of the conversion factors for traded goods. The question as to which formula to use
is essentially an empirical one (Potts 2002).
4.
DATA
The data are derived from various sources. For the shadow price of capital estimations, the
inflation data are derived from the Central Bureau of Statistics’ National Accounts from
1996 to 2006, and from the Bank of Namibia’s quarterly and annual reports. The rate of
return from investing Namibian assets in South African long-term bond instruments was
obtained from a local consulting firm, Jacques Malan Consultant and Actuaries.
For the SRTP calculations, we follow Evans and Sezer (2004), where the rate of pure time
preference U is assumed to be 1.5%, the elasticity of marginal utility of consumption T is
assumed to be 1.3, and the average growth rate of per capita real consumption g is the
average annual growth rate per capital real GDP from 1996 to 2006, derived from the
National Accounts data. The g was 2.87% over the 1996–2006 period.
The labour estimations used LARRI’s Actual Wage Rate Database, the results of the
LARRI labour force survey conducted in 2004, the Ministry of Labour’s survey on
Namibia’s informal economy in 2001, and LARRI’s study on that economy in 2006. In
terms of unskilled labour, we will use the minimum wages determined by LARRI (2005,
2006a) for the various economic sectors in Namibia. The database is derived from wage
agreements entered into between various trade unions and corporate entities between 2000
and 2005. This database will represent the urban semi- and unskilled labour pools. We will
23
Michael N Humavindu
also look at special categories such as farm workers and security guards, who are formally
paid a minimum wage as set out by legislation.
LARRI (2006b) shows that, on average, the majority of informal workers get paid N$175
per month. The estimated number of people working in the informal sector is 133,000.
Unfortunately, there are no disaggregated data available on rural and urban wages.
Therefore, wages for the informal sector as well as for farm workers are used as a proxy for
rural semi-skilled and unskilled labour. The labour force survey of 2004 estimates that
108,119 people are unemployed. Using these data, we obtain f1 at 0.55, f2 at 0.45, and Wf at
N$175 to estimate the EOCL equation. For urban semi-and unskilled labour, Wp is the
LARRI database’s average national wage, namely N$1,475 per month. For rural workers,
the Wp is the farm workers’ minimum wage of N$428 per month. For estimating the EOCL
of foreign labour, tF is 35%, with tVAT at 15%, and Ee/Em being the SERF calculated in this
study. Finally, we assume R (the proportion of the net-of-tax income repatriated by foreign
labour) at 40%.
Namibian trade statistics to estimate the forex premium were obtained directly from the
Central Bureau of Statistics and the Bank of Namibia reports.
5.
RESULTS
5.1
Discount rate estimations
The discount rate estimations using the amended EOCK formula yielded the following:
24
Estimating national economic parameters for Namibia
Table 1: Discount rate calculations: Results of estimations
Method
Discount rate
Amended EOCK
7.2%
SRTP
5.3%
The amended EOCK is 7.2%, whilst the results of the SRTP are 5.3%. Thus, the estimates
are slightly lower than the informal estimate of 8% from Barnes (1994). The work by Kuo
et al. (2003) estimates the South African EOCK at 11%, which is higher than our estimates.
Zhuang et al. (2007) mention that a major criticism of using SRTP is that it is purely a
measure of the social opportunity cost in terms of foregone consumption, and that it ignores
the fact that public projects could also crowd out private sector investments if they cause
the market interest rate to rise. Therefore, it is necessary to reflect what society could have
gained from the displaced private investment that can be measured by the marginal social
rate of return on private sector investment. As the SRTP is generally low, if it is exclusively
used as the social discount rate it may lead to too many low-return investments being
undertaken in the public sector.
5.2
EOCL estimation results
The results of the EOCL estimations are presented in Tables 2 and 3 below:
25
Michael N Humavindu
Table 2: EOCL estimations
Namibian minimum wage, by sector
Three-year
average
2003–2005
EOCL
Wi assumed at
N$175
Economic costs
as share of
financial costs
(%)
Agriculture / hunting / fishing / forestry
1,256
417
33
Community services / social services / personal
services
1,676
511
31
Construction
1,415
452
32
Manufacturing
1,366
441
32
Mining and quarrying
1,812
542
30
Transport and storage
1,693
515
30
Wholesale and retail trade
1,104
383
35
National average
1,410
452
32
Economic costs as share of financial costs
32%
Table 3: EOCL estimations of special categories
Special categories
Protected
wages
EOCL
Wi assumed at N$175
Economic costs
as share of
financial costs
(%)
Farm workers
428
231
54
Security guards
588
267
46
EOCL of foreign labour factor
n/a
n/a
59
EOCL of Namibian expatriates
n/a
n/a
28
The EOCL estimations show that, as a share of financial costs, economic costs are 32% for
Namibian urban semi- and unskilled labour, and around 54% for rural semi- and unskilled
labour. The economic costs of foreign labour and Namibian expatriates are 59% and 28% of
financial costs, respectively. In comparison, the informal estimate in Barnes (1994) was that
26
Estimating national economic parameters for Namibia
the economic cost was 35% of the financial cost for all unskilled labour. The estimations by
Bicak et al. (2004) show the South African accounting price of unskilled labour at 60%,
whilst their Namibian counterpart is at 32%. The South African accounting price for foreign
labour is 73%, whereas the Namibian estimations yielded an accounting price of 59%.
5.3
SERF estimation results
The results of the SERF estimation are presented in Table 4 below:
Table 4: SERF estimations (imports and exports for trade with countries outside SACU)
SERF estimations
2004
2003
2002
2001
2000
1999
Imports (cif) (N$)
48,494,118
39,233,302
21,112,383
19,137,768
29,366,773
22,431,392
Exports (fob) (N$)
81,766,641
117,776,370
64,175,570
55,621,342
52,919,528
27,512,564
Import taxes (N$)
7,185,927
6,362,725
2,314,213
2,503,542
3,714,767
3,103,544
Export taxes (N$)
29,733
25,727
9,044
9,832
10,279
7,241
7,156,193
6,336,998
2,305,169
2,493,710
3,704,488
3,096,303
130,260,759
157,009,672
85,287,953
74,759,110
82,286,301
49,943,956
SERF
1.05
1.04
1.03
1.03
1.05
1.06
SERF, six-year
average
1.04
Net trade taxes (N$)
Total trade (N$)
The SERF estimations indicate a value of 1.04. A more general point is that the SER is not
a precise figure since it will be used in projections into an uncertain future. Therefore, there
are grounds for using a central approximation (or best estimate) and doing some sensitivity
tests around the central value. Thus, in appraising projects, it is best to apply a sensitivity
analysis using a range of values around the 4% central value. Harberger et al. (2003)
27
Michael N Humavindu
estimate a value of 6.2% for the South African economy, which is higher than this work’s
estimate. As mentioned earlier in the paper, these estimates are for out-of-SACU trades as
the SERF for SACU is 1.
6.
CONCLUSIONS
This has been the first formal attempt at estimating national economic parameters for the
Namibian economy.
In terms of the amended EOCK, the lower value of 7.2% – compared with SA’s 11% –
clearly reflects the Namibian net saver position. The estimate is also close to the Barnes
(1994) guesstimate, which has been used for the last 14 years. The SRTP low value of 5.3%
is best used for public projects that are unlikely to displace private investments, such as
food-for-work programmes and other non-profit public sector initiatives. On the other hand,
the EOCL estimations for farm workers, which are used as a proxy for semi- and unskilled
rural labour, are much higher than the Barnes (1994) guesstimates. The SER estimate, while
lower than the Barnes (1994) guesstimates of 6%, is for out-of-SACU trades which the
latter work did not realise or incorporate.
The results should be useful for efficient and sustainable development planning in Namibia.
Further extensions and enhancements of this work should entail estimating shadow prices
using input–output analyses in order to estimate conversion factors for the various sectors
of the Namibian economy.
28
Estimating national economic parameters for Namibia
REFERENCES
Bacha, E & L Taylor (1971): “Foreign exchange shadow prices: A critical review of current
theses”. Quarterly Journal of Economics, 85(1):197–224.
Bank of Namibia (Various): Annual Reports. Windhoek: Bank of Namibia.
Bank of Namibia (Various): Quarterly Bulletins. Windhoek: Bank of Namibia.
Barnes, JI (1994): “Suggested criteria for shadow pricing in the cost–benefit analysis of
projects in Namibia”. Mimeo. Windhoek: Environmental Economics Unit, Ministry
of Environment and Tourism.
Barreix, A (2003): “Rates of return, taxation and the economic cost of capital in Uruguay”.
Unpublished Ph D dissertation, Harvard University, Cambridge, Massachusetts,
United States.
Berman, JR (1986): “Shadow prices and subsidies in Botswana”. Journal of Development
Economics, 22(2):351–392.
Bicak, BA, GP Jenkins, C Kuo & MB Mphahlele (2004): “An operational guide to the
estimation of the economic opportunity cost of labour in South Africa”. South
African Journal of Economics, 72(5):1056–1067.
Boardman, A, D Greenberg, A Vining & D Weimer (2001): Cost-benefit analysis:
Concepts and practice (2nd edition). Upper Saddle River, NJ: Prentice Hall.
Boscolo, M, J Vincent, & T Panayotou (1998): “Discounting costs and benefits in Carbon
Sequestration Projects”. Harvard Environment Discussion Paper No. 41.
Cambridge: Harvard Institute for International Development.
Campbell, H & R Brown (2003): Benefit-cost analysis. Cambridge: Cambridge University
Press.
29
Michael N Humavindu
Central Bureau of Statistics (2006): Namibia Household Income and Expenditure Survey
2003/2004. Windhoek: National Planning Commission.
Dasgupta, P, S Marglin & AK Sen (1972): Guidelines for project evaluation. New York:
United Nations.
Evans, D & H Sezer (2004): “Social discount rates for six major countries”. Applied
Economic Letters, 11(9):557–560.
Fitch Ratings Agency (2005): Sovereign Report: Namibia. London: Fitch Ratings Agency.
Flatters, F & M Stern (2005): “Implementing the SACU revenue-sharing formula: Customs
Revenue”. Mimeo.
Frayne, B & W Pendleton (2001): “Migration in Namibia: Combining macro and micro
approaches to research design and analysis”. International Migration Review,
35(4):1054–1085.
Frayne, B & W Pendleton (2002): “Mobile Namibia: Migration trends and attitudes”.
Southern African Migration Project Migration Policy Series No. 27.
Harberger, AC (1972): “The opportunity costs of public investment financed by
borrowing”. In Layard, Richard (Ed.). Cost-benefit analysis. New York: Penguin
Books Ltd.
Harberger, AC & GP Jenkins (Eds) (2002): Cost–benefit analysis. London/New York:
Edward Elgar Publishing Limited.
Harberger, AC, GP Jenkins, CY Kuo & MB Mphahlele (2003): “The economic cost of
foreign exchange for South Africa”. South African Journal of Economics,
71(2):298–324.
Harris, JR & MP Todaro (1970): “Migration, unemployment and development: A twosector analysis”. American Economic Review, 60(1):126–142.
30
Estimating national economic parameters for Namibia
International Monetary Fund (2008): Namibia Selected Issues and Statistical Appendix.
Washington DC: IMF.
Jenkins GP & CY Kuo (1998): “Estimation of the national parameters for economic cost–
benefit analysis for the Philippines”. Development Discussion Papers No. 653.
Cambridge: Harvard Institute for International Development.
Khan, MZ (1979): “Estimation of shadow prices for project evaluation in Pakistan”. The
Pakistan Development Review, 18(2):129–145.
Kirk, R & M Stern (2005): “The new Southern African Customs Union Agreement”. World
Economy, 28(2):169–190.
Kuo, CY, GP Jenkins & MB Mphahlele (2003): “The economic cost of capital in South
Africa”. The South African Journal of Economics, 71(3):523–543.
Lagman-Martin, A (2004): “Shadow exchange rates for project economic analysis: Toward
improving practice at the Asian Development Bank”. Economic & Research
Department Technical Note No. 11. Manila: Asian Development Bank.
LARRI/Labour Resource and Research Institute (2005): The Namibian Wage Bargaining
Report 2004. Windhoek: LARRI.
LARRI/Labour Resource and Research Institute (2006a): The Namibian Wage Bargaining
Report 2005. Windhoek: LARRI.
LARRI/Labour Resource and Research Institute (2006b): Namibia’s informal economy:
Possibilities for trade union intervention. Windhoek: LARRI.
Little, IMD & JA Mirrlees (1969): Manual of Industrial Project Analysis in Developing
Countries. Paris: Organization for Economic Cooperation and Development.
Marope, MT (2005): “Namibian human capital and knowledge development for economic
growth with equity. Africa region human development”. Working Paper Series No.
84. Washington, DC: The World Bank.
31
Michael N Humavindu
Medalla, EM (1982): “Shadow prices of goods and resources in the Philippines: An
assessment”. Staff Paper Series No. 82–4. Manila: Philippine Institute for
Development Studies.
Medalla, EM & T Powers (1984): “Estimating the shadow exchange rate, the shadow wage
rate and the social rate of discount for the Philippines”. Staff Paper Series No. 84–
03. Manila: Philippine Institute for Development Studies.
MIDSA/Migration Dialogue for Southern Africa (2006): MIDSA Workshop on Migration
and Development/Poverty in the SADC Region: Final Report. Kingston: Queens
University.
Ministry of Finance (2003): 2003/04 Budget Speech. Windhoek: Ministry of Finance.
Ministry of Labour (2001): The Namibian Informal Economy Survey: Report of analysis.
Windhoek: Ministry of Labour.
Ministry of Labour (2002): The Namibia Labour Force Survey 2002: Final report.
Windhoek: Ministry of Labour.
Ministry of Labour (2005): The Namibia Labour Force Survey 2004: Final report.
Windhoek: Ministry of Labour.
Motinga, D & G Mohammed (2002): “On the causes and consequences of globalization:
What implications for the Namibian labour market?”. NEPRU Working Paper 85.
Windhoek: Namibian Economic Policy Research Unit.
Motinga, D & C Tutalife (2006): Joblessness and public policy: Evaluating the labour
market impact of fiscal and labour market policies in Namibia. Windhoek: Institute
of Public Policy Research.
Percoco, M & P Nijkamp (2006): “Individuals’ time preference and social discounting: A
survey and a meta-analysis”. Mimeo.
32
Estimating national economic parameters for Namibia
Potts, D (2002): Project planning and analysis for development. London: Lynne Rienner
Publishers.
Powers, A (2003): Benefit–cost analysis and the discount rate for the Corps of Engineers’
water resource projects: Theory and practice. Washington, DC: Congressional
Research Services, The Library of Congress.
South African Reserve Bank (Various): Quarterly Bulletin. Pretoria: SARB.
The Service Group (2005): The Namibia Investor Roadmap. Report prepared for the
Regional Center for Southern Africa. Windhoek: United States Agency for
International Development.
Tinbergen, J (1958): The design of development. Baltimore: The Johns Hopkins Press.
Tjirongo, M (1995): “Short-term stabilization versus long-term price stability: Evaluating
Namibia’s membership of the Common Monetary Area”. Centre for the Study of
African Economies Working Paper 95–18. Oxford: Oxford University.
Vollan, B (2000): “The development of financial markets in Namibia”. South African
Journal of Economics, 68(1):63–97.
Westergaard-Nielsen, N, D Hansohm & D Motinga (2003): “Analysis of the Namibian
labour market: The demand side”. Unpublished draft for World Bank. Windhoek.
World Bank (2006): World Development Indicators. Washington, DC: The World Bank
Group.
Zerbe, JR & D Dively (1994): Benefit–cost analysis: In theory and practice. New York:
Harper Collins.
Zhang, J, Z Liang, T Lin & F de Guzman (2007): “Theory and practice in the choice of
social discount rate for cost-benefit analysis: A survey”. Economic & Research
Department Working Paper No. 94. Manila: Asian Development Bank.
33
Michael N Humavindu
APPENDIX 1
The economic opportunity cost of capital
Theoretically, the social rate of return may be defined by applying national accounting
principles. In an open economy, real income can be different from real product because of
the servicing of national debt. Let us assume that s is the average interest rate on the stock
of foreign debt (D). Then income Y is given by –
Y = q - s .D
(1)
where q is the real product. If we then consider a new public project, –
'8 G .'I g U .'I p i f .'D
(2)
where q= G .'I g U .'I p . q is the permanent change in real product, Ig is the new public
project, is the rate of return of the project, Ip is the change in private investment caused
by the new project (Ip<0), U is the marginal rate of return that the postponed investment
would have generated, if is the marginal cost of additional foreign borrowing, and D
represents the change in the external debt stock.
The decision rule for accepting the project is that the discounted stream of extra income
(Y) must be higher than the consumption forgone now (change in savings S). Thus, the
project should be accepted if the following condition can be satisfied:
34
Estimating national economic parameters for Namibia
'Y
t 'S
r
(3)
This can then be rewritten as follows:
'Y t r 'S
(4)
Substituting (2) into (4) gives us –
G'I g G'I p i f 'D t r'S
(5)
G'I g t r'S i f 'D U'I p
(6)
Thus, for marginal public investment, we have –
G
where
r
'I p
'S
'D
if
U
'I g
'I g
'I g
'S 'D 'I p
;
;
'I g 'I g 'I g
(7)
represents shares of funds sourced from different parts of the
capital market. We can then solve the following:
'I g
'S 'D 'I p
§ wS wD wI p
¨¨
wr
© wr wr
wI p wr
wS wr
wD wr
'I g 'I g 'I g
wr wI g
wr wI g
wr wI g
· wr
¸¸
'I g
¹ wI g
(8))
35
Michael N Humavindu
where
wS wD wI p
,
,
represent shares of funds.
wr wr
wr
The weights of (8) can be written in terms of the aggregate elasticity of each source:
Hs
wS r
wS
Ÿ
wr S
wr
HsS
r
,
(9)
and similarly for D and Ip. Thus, we have –
wS / wr
wS / wr wD / wr wI P / wr
HsS / r
I
H s S / r H f D / r K
r
HsS
H s S H f D KI
(10)
which represents the share of increased savings (weight, f1). The other two weights can be
derived similarly:
wD / wr
wS / wr wD / wr wI P / wr
HfD
H s S H f D KI
(11)
which represents the share of increased foreign borrowing (weight, f2); and finally,
wI p / wr
wS / wr wD / wr wI P / wr
KI
H s S H f D KI
which represents the share of displaced private investment (weight f3).
36
(12)
Estimating national economic parameters for Namibia
Thus, where s is the supply elasticity of household savings, f is the supply elasticity of
foreign funds and is the elasticity of demand for domestic investment relative to changes
in the interest rates. St is the total savings available in the economy, of which Sr is the
contribution to the total savings by households, and Sf is the total contribution of net foreign
capital inflows.
Barreix (2003) mentions that only this market-driven opportunity cost approach is
sufficiently flexible to easily add a new source of financing to the analysis. This approach
also has another important advantage: it can be defined as a single value. Thus, no extra
adjustment on investment expenditures is required, and no classification of benefits and
costs are needed.
Barreix (2003) surveys the empirical literature on the estimation of the shadow price of
capital and finds that most studies – especially those relating to developing countries – have
used the EOCK approach. The standard method for estimating the EOCK for developing
countries is captured in the work of Jenkins and Kuo (1998), where it is measured as a
weighted average of the rate of time preference to savers (J), the cost of additional foreign
capital inflows (MC), and the rate of return on displaced investment (S). The weighted
average of these three costs can be expressed as follows:
EOCK = f1 x J + f2 x MCf + f3 x S
(13)
where J, MCf and S , respectively, equal the costs of the public sector funds obtained at the
expense of current consumption, the cost of additional foreign capital inflow to the
37
Michael N Humavindu
economy, and at the expense of other domestic investment. The cost of foreign borrowing
(MCf) is valued at its marginal cost. The weights (f1, f2, and f3) are the shares derived earlier,
and are equal to the proportion of funds diverted or sourced from each sector.
If the weights are expressed in terms of elasticities of demand and supply of funds with
respect to changes in interest rates, equation (13) can be rewritten as follows:
EOCK
H s ( S r / S t ) x J H f ( S f / S t ) x MC f K x S
H s ( Sr / St ) H f ( S f / St ) K
38
(14)
II
Estimating Namibian shadow prices
within a semi-input–output framework
Michael Humavindu1
Development Bank of Namibia, PO Box 235, Windhoek, Namibia
[email protected]
Abstract
The purpose of this paper is to derive, using the semi-input–output (SIO) technique, a set of
accounting price ratios (APRs) for the various economic sectors of Namibia. An APR is the
ratio between the market or financial price and the efficiency or economic value of a specific
commodity or sector. APRs are useful for the economic analyses of investment or
development initiatives. In contrast to previous estimates of shadow prices in Namibia, which
are limited to a few aggregate shadow prices for capital, labour and foreign exchange, the
SIO estimation technique applied here to the Namibian economy permits the easy calculation
of many more such prices. The results shows that most tradable sectors such as fishing and
mining have APRs equal to or closer to one, with deviations explained by import tariffs.
Non-tradable sectors’ APRs exhibit greater variation, with the water sector having the highest
reflecting its scarcity.
Keywords: semi-input–output; accounting price ratios; Namibia
1
The author wishes to acknowledge a research grant from the Jan Wallander and Tom Hedelius Foundation.
The author would also like to thank Magnus Wikström, Karl-Gustaf Löfgren, David Potts, Jon Barnes, Elio
Londero and two anonymous referees for comments on earlier versions of the paper. The usual disclaimer
applies.
Estimating Namibian shadow prices within a semi-input–output framework
1.
INTRODUCTION
The purpose of this paper is to estimate shadow prices for the Namibian economy using the
semi-input–output (SIO) method. Essentially, this work is an extension of Humavindu (2008)
which was limited to only a few aggregate shadow prices for capital, labour and foreign
exchange. Applying the SIO analysis would permit the estimation of many more shadow
prices for the Namibian economy. A larger set of national shadow prices would improve
appraisal results, especially where these are needed for goods and services that are inputs or
outputs of a project.
The need for shadow pricing arises due to the notion that the market prices of goods and
services and productive factors may not reflect their real worth due to several distortions. The
distortions can be of an economic nature, such as import duties, export taxes or subsidies, and
other indirect taxes. An economic adjustment is needed for a productive factor like labour
due to imperfect labour markets. The presence of minimum wage laws may set wages
substantially higher than the economic opportunity cost of labour. This necessitates an
adjustment to the prevailing labour market prices, and for those goods and services where
labour is an important factor of production.
Humavindu (2008) advances both practical and academic reasons for the need to estimate
shadow prices for the Namibian economy. From a practical viewpoint, there is no consistent
shadow pricing application at national planning level – despite challenges that include
poverty, high unemployment, and low industrialisation. This could lead to suboptimal
allocation of resources to ameliorate the enormous development challenges. The country is
1
Michael N Humavindu
contemplating large-scale projects such as the development of the Kudu gas fields, diamond
mining in current conservation areas such as the Sperrgebiet, hydropower plants, and a
nuclear power plant, and these would need to be assessed on both financial and economic
grounds. In the field of energy in particular, the decision whether to promote renewable or
nuclear energy to alleviate the imminent energy shortfall is crucial. Both options would be a
matter of imported capital equipment, but shadow pricing analysis would aid in assessing
which option accrues greater economic benefits to Namibia. Another relevant example is the
textile sector, which has grown as a result of government support. However, the sector has
caused environmental and labour-relations problems, suggesting that the government decision
to support it could have benefited from objective economic assessment.
Academic reasons for the interest in Namibian shadow prices include the country being a net
capital exporter while having a large service sector – which is unusual for a developing
economy. Also, Namibia’s participation in the Southern African Customs Union (SACU),
which shares customs revenue amongst its member countries, could affect the estimation of a
shadow price of foreign exchange. Humavindu (2008) estimates the economic opportunity
cost of capital at 7.2%. The economic costs of Namibian labour as a share of financial costs
are 32% for urban semi-skilled and unskilled labour, and are 54% for rural semi-skilled and
unskilled labour. The economic costs of foreign labour as a share of financial costs are 59%.
The shadow exchange-rate premium is 4%.
The current work extends Humavindu (2008) by employing the SIO analysis: it distinguishes
between international sectors that produce tradable goods, and national sectors that produce
non-tradable goods (Kuyvenhoven 1978). The aim is to estimate the ratio between the market
2
Estimating Namibian shadow prices within a semi-input–output framework
price of a resource or commodity and its value at efficiency prices. The ratio between the
market price and the efficiency value of a specific commodity or sector is known as the
accounting price ratio (APR; see MacArthur 1994). This ratio can be applied to the constant
price financial values in project analysis to derive the corresponding economic values.
The SIO approach is regarded as the most advanced among those applied in national
parameter studies (Saerbeck 1989). This is because it treats two major theoretical areas of
concern – the problem of interdependence in the estimation of key parameters, and the
valuation of non-traded goods and services – in the most accurate way. Thus, with the SIO
method, one may readily derive shadow prices for many different sectors of the economy and
include the direct and indirect effects of protection on the conversion factors of typically nontraded goods and services.
2.
METHODOLOGY
2.1
IO and SIO analysis
Generally, input–output analysis (IO) is well suited to assess how changes in one or more
economic sectors will impact on the total economy. Input–output analysis is generally applied
to assess the impact of a change in the demand conditions for a given economic sector. IO
analysis uses matrix algebra to find out how much output will be utilised in productive
activities to obtain a final net output and how much will be left over for consumption
(Baumol 1977). Thus, an IO model can be used to estimate the amount of income,
3
Michael N Humavindu
employment, and production that will be generated by a given level of demand (Leontief
1986).
In IO analysis, the structure of an economy can be represented by the value of transactions
between sectors (primary, manufacturing and services) in a matrix (the A matrix). Each sector
in the A matrix has both a row and column. The rows of this matrix are the sectors that a
given sector sells its output to (as intermediate inputs to those sectors), while the columns are
the sectors a given sector purchases its intermediate inputs from.
Completing the IO table is the addition of final demand (including demand from consumers
and exports), the destination of sales that do not go to other sectors, and primary inputs
(labour, land, capital and imports), i.e. the inputs that are not purchased from other sectors.
This is called the Y. Thus, a simplified IO model can be demonstrated as presented in
Equation 1 below:1
X ij
a ij X j
(1)
where Xij, the amount of sector i’s output required for the production of sector j’s output, is
assumed to be proportional to sector j’s output – Xj, and aij is the relevant input–output
coefficient. Summing over sectors and adding final demand Yi to Equation (1) produces the
following input–output relationship:
n
Xi
¦
aij X j Yi
(2)
j 1
4
Estimating Namibian shadow prices within a semi-input–output framework
The latter is assumed to hold in first-difference form (depicting changes in the variables).
Equation (2) in matrix form is X = AX + Y. Rearranging the equation gives the following:
X
(, $) 1 Y
(3)
In Equation (3), X is the vector of outputs, Y is the vector of final demands, A is the matrix of
input–output coefficients, and I is the identity matrix. In IO analysis, the most common
assumption is that all sectors produce output using fixed proportions of inputs and, hence, can
increase production with constant marginal cost. If this is the case, the Leontief Inverse (I-A)1
can be used to calculate overall changes in sectoral outputs caused by changes in final
demand. All sectors are assumed to be perfectly elastic in supply.
In the IO model, prices are determined by the costs of inputs and primary factors. We have –
P
PA Pf F
(4)
where P is the vector of prices, A is the matrix of input-output coefficients, Pf is the vector of
prices of primary factors and F is the matrix of coefficients of primary factor requirements in
different sectors. Rearranging (4), we have –
P ( I A)
P
Pf F
Pf F ( I A) 1
(5)
5
Michael N Humavindu
The SIO approach differs from its IO counterpart in that the assumption of perfectly elastic
supply in all sectors is rejected because the assumption leads to an overestimation of output
responses following from any intervention. In reality, most developing countries have several
sectors that face supply constraints (Haggblade et al. 1991).
The SIO approach retains the basic assumptions of the IO approach, but introduces supply
rigidities in some sectors. The SIO approach classifies all economic sectors as either supplyconstrained or perfectly elastic in supply. The approach assumes perfect substitutability
between domestic goods and imports. Thus, world prices ensure fixed prices for tradable
goods. Tradable goods are assumed to be supply-constrained; an increase in domestic
demand merely reduces net exports, which are endogenous to the system (Cornelisse &
Tilanus 1966; Diao et al. 2007; Dorosh et al. 1992; Kuyvenhoven 1978). For non-tradable
goods, on the other hand, supply is still assumed to be perfectly elastic, as in the IO model.
Figures 1 and 2 below provide a simple graphic example of how matrix subdivisions differ in
IO and SIO, respectively.2
6
Estimating Namibian shadow prices within a semi-input–output framework
Figure 1: Stylised representation of an input–output transactions table
Purchase by sector: Read down p
Sales by sector: Read across o
Quadrant 1
Intermediate
inputs to
production (A)
Quadrant 2
Final demand
Total demand
Quadrant 3
Primary inputs to
production
(F)
Total output
Quadrant 1 in Figure 1 represents the intermediate inputs matrix, which defines the two-way
links between industries and, through these links, the labour, fixed capital and natural
resource requirements of final demand. Input–output methods use this information to define a
two-way process showing the flow of goods and services between sectors and “to” and
“from” processes or entities (value added and final demand). Therefore, the system can be
interpreted as reflecting the technical relationship between the level of output and the
required quantities of inputs, and the balancing of supply and demand for each type of good
and service. For each sector, total outputs are absorbed as inputs of other industries or in final
use (consumption, investment and net exports). In conventional IO models, no distinction is
made between tradable and non-tradable goods and services.
7
Michael N Humavindu
Figure 2: Stylised representation of a semi-input–output transactions table
Purchase by sector: Read down p
Sales by sector: Read across o
Quadrant 1a
Intermediate
inputs to
tradables from
tradable sectors
(A11)
Quadrant 1b
Intermediate
inputs to nontradables from
tradable sectors
(A12)
Quadrant 2a
Final demand
tradables
Quadrant 1c
Intermediate
inputs to
tradables from
non-tradable
sectors (A21)
Quadrant 1d
Intermediate
inputs to nontradables from
non-tradable
sectors (A22)
Quadrant 2b
Final demand
non-tradables
Quadrant 3a
Primary inputs to
tradables
production
Quadrant 3b
Primary inputs to
non-tradables
production
Total
demand
Total output
However, in SIO, the transaction table is rearranged to delineate Quadrant 1 in the tradables
and non-tradables sectors of the economy. Instead of producing a full A matrix as in
traditional IO analysis, we produce an A22 matrix containing the non-tradable economic
sectors. Therefore, A22 in Figure 2 represents the matrix coefficients for non-traded inputs
that are needed to produce non-traded output. The demand for traded inputs per unit of nontraded output is given by the A12 matrix – essentially, the matrix coefficients for traded inputs
used to produce non-traded output. Thus, the global demand for each type of input per unit of
non-traded output is obtained by adjusting each item by the Leontief Inverse of the A22
matrix, (I-A22)-1.
8
Estimating Namibian shadow prices within a semi-input–output framework
The difference between IO and SIO assumptions has implications for the determination of
prices. In the SIO model, the prices of tradable goods are not determined endogenously; they
are determined by world market prices and by taxes and subsidies that distort the domestic
prices relative to the world market prices. We have –
APRtradables: w
1
1
(1 w )
(1 t m )(1 v m )
(1 t x )(1 v x )
(6)
where tm represents the import tariff, vm represents indirect taxes (net of subsidies) levied at
the point of entry, tx represents the export tax, vx represents total indirect taxes (net of
subsidies) levied on export sales, and w represents the weights of imports relative to total
trade.
The prices of non-tradable goods, on the other hand, are determined endogenously. We
have –
P2
P2 A22 P1 A12 Pf F2
P2 ( I A22 )
P2
P1 A12 p f F2
P1 A12 ( I A) 1 Pf F2 ( I A22 ) 1
(7)
where P2 represents the accounting price ratios of non-traded goods, P1 represents the
accounting price ratios of the traded inputs, A12 represents the matrix coefficients for traded
inputs that are used to produce non-traded output, A22 represents the matrix coefficients for
non-traded inputs that are needed to produce non-traded output, Pf represents the shadow
9
Michael N Humavindu
price of the primary factors, and F2 represents the matrix of the coefficients of primary factor
purchases and transfer payments per unit of non-traded output.
2.2
The SIO approach
The SIO approach used here may conveniently be broken down into six steps:3
x
STEP 1: The economic sectors are separated into traded and non-traded sectors.
x
STEP 2: APRs are calculated for all traded sectors, using data on import duties and
export taxes. The APRs for the traded sector are calculated using the formula in eq.
(ref to equation above):
x
STEP 3: National economic parameters for labour, foreign exchange and capital are
derived. In this case, they were derived from Humavindu (2008).
x
STEP 4: The SIO table is used to derive direct-input coefficients for each of the nontraded sectors.
x
STEP 5: The resulting sets of equations are solved to yield both the direct and the
indirect input coefficients of all sectors into the non-traded sectors. This is done by
taking the Leontief Inverse of the submatrix of input coefficients from the non-traded
into the non-traded sectors, and premultiplying it with the matrices of direct input
coefficients from traded and primary sectors into the non-traded sectors.
x
STEP 6: The resulting matrix of the direct and indirect coefficients is multiplied by
the vector of APRs of the traded and primary sectors, to yield the APRs of each of the
non-traded sectors.
10
Estimating Namibian shadow prices within a semi-input–output framework
Thus, the APRs for non-tradables are the sum of all the economic values of the traded and
non-traded material inputs and factors used to produce them.
Economic profits should have a shadow value of zero in SIO analysis. The usual assumption
is that increased demand for a non-traded good will lead to the sector producing that good to
expand at the current capital-output ratio, so that, in principle, a need arises to price the
additional capital investment at its opportunity costs. In practice, SIO studies frequently
assume a range of shadow values for the gross operating margin, or just use the figure
derived from a social accounting matrix (SAM). However, when capital stock data are
available (as they were for this study), the additional capital can be directly shadow-priced at
its opportunity cost.
In the absence of trade distortions, the tradable sectors are expected to have APR values of 1
or very close to unity. Where tradable APRs exceed 1, it is generally an indication of either
taxed exports or of subsidies for local consumption. If tradable APRs values are below unity,
it is usually an indication of export subsidies (MacArthur 1997).
APRs for non-tradables are expected to have values less than 1. APRs with less than 1
normally arise if the market price contains either significant non-resource cost transfers (for
taxes, excess profits, etc.) or large amounts of labour whose market wage is considerably
higher than the estimated opportunity costs. APRs with values above 1 imply a heavily
subsidised market price, or that the non-tradable contains a higher-than-average content of
traded items in the cost structure and is available only in restricted supply, so that one project
11
Michael N Humavindu
can obtain its needs only by denying the product to another willing project/purchaser
(MacArthur 1994).
3.
THE NAMIBIAN ECONOMY AND DATA
3.1
Overview of the Namibian economy
The Namibian economy is characterised by dualism: a modern market sector based on a
capital-intensive industry and farming, which produces most of the country’s wealth, coexists
alongside a traditional subsistence sector. The country’s gross domestic product (GDP) per
capita (US$3,100 at market exchange rates) is relatively high among developing countries,
but obscures one of the most unequal income distributions on the African continent. There is
also a high degree of openness in the economy, due to the sum of exports and imports having
a more than 90% share of GDP.
The economy is also highly integrated with that of South Africa. Three-fourths (75%) of
Namibia’s imports originate there, and transport and communications infrastructure are
strongly linked with South Africa. In addition, the high degree of labour mobility between
Namibia and South Africa means that many services (which are usually non-tradables)
become tradables for Namibia.
Namibia’s imports are mainly machinery and equipment (40%), light manufacturing (21%),
and petroleum products (10%). Namibia’s exports consist mainly of diamonds and other
minerals (44%), fish products (21%), and tourism (15%). Beef and meat products and grapes
12
Estimating Namibian shadow prices within a semi-input–output framework
are other important exports. The textile sector has grown under the United States of
America’s African Growth and Opportunity Act (AGOA), thereby increasing apparel exports.
The Namibian economy is largely driven by, in descending order, the tertiary sector, primary
sector, fisheries and beef processing. The tertiary sector is currently the biggest contributor to
GDP (over 55%), with government services accounting for a third of the tertiary sector’s
output, whilst the trade, transport and finance sectors contribute to half of this sector’s output.
Namibian agriculture contributes less than 5% of Namibia’s GDP, although 70% of the
Namibian population depends on agricultural activities for their livelihood. The
manufacturing sector contributes about 11% to GDP. The latter sector’s growth has
historically been limited by a small domestic market, dependence on imported goods, a
limited supply of local capital, a widely dispersed population, a small skilled labour force,
high relative wage rates, and competition from South Africa. Tourism is increasingly
becoming an important sector, with recent estimates of a 10% contribution to GDP (World
Travel and Tourism Council 2006).4
3.2
Data
Three sets of data were used for estimating the APRs. Firstly, trade statistics and National
Accounts data were obtained directly from the Namibian Bureau of Statistics as well as the
Bank of Namibia, the country’s central bank. Trade statistics were used to determine the
tariffs and subsidy values for tradable sectors, as these are useful for estimating tradable
APRs. The National Accounts were used to derive capital stock data.
13
Michael N Humavindu
The second set of data is the inter-industry flow of transactions among sectors of the
economy. For this the 2001/2 SAM was used, which is the most recent one available (Lange
et al. 2004). The SIO model used here is built around a 2001/2 Namibian SAM. The latter is a
72 x 72 matrix and contains an account for each of 30 production activities, 5 factors of
production, and 9 institutions. Following United Nations (1999), the SAM is aggregated into
an IO table with 26 tradable and non-tradable sectors. The classification of the 26 sectors is
given in Table 1 below:
Table 1: Classification of the sectors
Sector and classification
Tradables
Commercial agriculture – Cereal
Commercial agriculture – Other crops
Commercial agriculture – Animal products
Fishing
Mining
Meat processing
Fish processing
Grain milling
Beverages and other food processing
Textiles
Light manufacturing
Heavy manufacturing
Construction
Transport
Market – Real estate and business services
Tourism – Non-residents
Petroleum products
Non-tradables
Traditional agriculture
Electricity
Water
Trade and repairs
Hotels and restaurants
Communication
Finance and insurance
Other private services
Government services
14
Estimating Namibian shadow prices within a semi-input–output framework
The shadow prices for primary factors (Humavindu 2008) were used as the third set of data.
The shadow price of capital is 7.2%. This discount rate was derived by applying an amended
Economic Opportunity Cost of Capital (EOCK) approach. This was to account for the net
capital exporter status of the Namibian economy. Two shadow prices for labour are derived:
rural and urban unskilled labour. The shadow price for urban unskilled labour is 0.32 of its
financial values, while the shadow price of rural labour is 0.54 of its financial values. These
shadow prices for labour were derived by utilising the supply price of labour method. In this
method, the labour shadow prices are determined by adjusting the gross-of-tax market wage
(i.e. the supply price) for distortions in the labour market, such as minimum wage regulations.
An important factor defined in the Namibian SAM is the net operating margin. As noted in
Section 2.2, the net operating surplus from the SAM is not used here; instead, the sectoral
capital costs were calculated by using unpublished National Accounts data on such stocks in
individual sectors. The tariff values obtained are shown in Table 2 below:
Table 2: Tariff values for tradable sectors
Sector
Commercial agriculture – Cereal
Commercial agriculture – Other crops
Commercial agriculture – Animal
products
Fishing
Mining
Meat processing
Fish processing
Grain milling
Beverages and other food processing
Textiles
Light manufacturing
Heavy manufacturing
Tourism – Non-residents
Petroleum
Tm
14.54%
16.58%
Vm
0
0
Tx
0.00%
0.00%
Vx
0
0
w
0.97
0.63
1-w
0.03
0.37
16.80%
0.00%
0.00%
0.00%
0.00%
14.51%
30.93%
17.15%
15.57%
15.54%
0.00%
11.97%
0
0
0
0
0
0
0
0
0
0
0
0
0.66%
0.00%
0.00%
0.30%
0.00%
0.00%
0.00%
0.00%
0.70%
0.00%
0.00%
0.00%
0
0
0
0
0
0
0
0
0
0
0
0
0.23
0.32
0.07
0.20
0.05
0.74
0.62
0.77
0.42
0.70
0.00
0.82
0.77
0.68
0.93
0.80
0.95
0.26
0.38
0.23
0.58
0.30
1.00
0.18
15
Michael N Humavindu
4.
RESULTS
The main results of our analysis are summarised in Table 3 below:
Table 3: APR estimates for Namibian economic sectors
Economic sectors
1. Tradables
Commercial agriculture – Cereal
Commercial agriculture – Other crops
Commercial agriculture – Animal
products
Fishing
Mining
Meat processing
Fish processing
Grain milling
Beverages and other food processing
Textiles
Light manufacturing
Heavy manufacturing
Construction
Transport
Market – Real estate and business
services
Tourism – Non-residents
Petroleum products
APR
0.87
0.91
2. Non-tradables
Traditional agriculture
Electricity
APR
0.66
0.88
0.97
1.00
1.00
1.00
1.00
0.91
0.85
0.85
0.95
0.87
1.00
1.00
Water
Trade and repairs
Hotels and restaurants
Communications
Finance and insurance
Other private services
Government services
1.13
0.53
0.50
0.95
0.62
0.84
0.95
1.00
1.00
0.90
As expected, most of the tradables have APRs closer or equal to unity. Sectors such as
fishing, mining, meat processing, fish processing and tourism all have APRs of 1. These
results once again affirm the high degree of openness of the Namibian economy. However,
the beverages sector’s economic values show the largest divergence, namely 15%, from their
financial values or prices. This divergence is reflected in the high import tariff of 30%.
Similarly, all sectors (Commercial agriculture – Cereal, Commercial agriculture – Other
crops, Commercial agriculture – Animal products, Grain milling, Beverages and other food
16
Estimating Namibian shadow prices within a semi-input–output framework
processing, Textiles, Light manufacturing, Heavy manufacturing, and Petroleum) that have
import tariffs have APRs that are below unity.
The estimated non-tradable APRs show more sectoral variations. It is not surprising that the
highest APR (1.13) is for the water sector, reflecting the scarcity of this resource. The
communication sector’s APR is also high (0.95), reflecting the high capital goods imports
and investments that sector makes. Government services constitute 35% of the total tertiary
sector and 19% of total GDP. This sector thus plays a major role in the economy. The high
APR (0.95) of this sector may also reflect high capital goods imports as well as investments.
As expected, the traditional agriculture (0.66), trade and repairs (0.53), hotel and restaurants
(0.50), and finance and insurance (0.62) sectors have low APRs.
Macarthur (1994) and Ghani (1999) caution against the need to seek commonality in APR
values between economies for comparative analyses. Application of different methodologies
partly accounts for different results, whilst the levels of commodity or sectoral disaggregation
are another cause. However, the major driver of differences in APR values is the variation in
underlying pricing and fiscal approaches of the various economies. Macarthur’s (1994)
review of 13 studies found in general that tradables’ and non-tradables’ APRs fell within their
expected range close to unity or less than unity, respectively. However, some outliers did
occur, with the fuel oil sector of Egypt having an APR estimate of 15.299. This reflected the
high subsidisation of fuel in Egypt at the time. On the other hand, some low APR values were
evident in sectors such as banking and insurance in Ecuador and Egypt, with values of 0.293
and 0.328, respectively. Most of the APRs reviewed were above 0.50 and below 1. Thus, the
results for Namibia are in congruence with previous work on APRs.
17
Michael N Humavindu
Of the non-tradables, the government-involved sectors (electricity, water, communications
and government services) have high capital-to-output ratios so that expansion will require a
lot of capital, especially for water. This drives up the APRs for these sectors because the
current return to capital is less than the opportunity cost of the capital that would be needed
for expansion. On the other hand, the opposite is true for several of the privately controlled
sectors - presumably because of limited competition, profits (and hence returns to capital) are
high and these sectors could expand with relatively little additional capital. Hence, the
opportunity cost of the additional capital needed for expansion is less than the current return
to capital. At the same time, unskilled labour will tend to pull down the shadow price of all
the non-tradable sectors, so sectors for which unskilled labour is an important input will have
lower APRs as a result. However, unskilled labour is a more important input to production
for government sectors than for the privately managed sectors (apart from traditional
agriculture). Thus, it appears that the main reason why private non-tradable sectors have low
APRs (and, implicitly, the reason government spending on those sectors would have a lower
social than financial cost) is not mainly that they employ a lot of unskilled labour, but mainly
that they can expand production considerably with relatively little additional capital.
5.
CONCLUSIONS
The work provides the first detailed estimates of APRs for the various economic sectors in
Namibia and demonstrates how SIO analysis can be applied to Namibia. The work is
essentially an extension of the first formal shadow pricing analysis for the Namibian
economy (Humavindu 2008), which estimated national economic parameters for capital,
18
Estimating Namibian shadow prices within a semi-input–output framework
labour and foreign exchange. By estimating specific economic sectors’ APRs, the work is a
useful addition for efficient and sustainable development planning in the country.
Reflecting the high degree of openness of the Namibian economy, the efficiency values for
tradables show a slight deviation from their market values. However, the non-tradables’
efficiency values show a wide divergence from their market values. As expected, unskilled
labour-intensive sectors such as traditional agriculture and trade have very low APRs, which
reflects the wide divergence between the shadow wage rate and market wage rate. However,
for several of the non-tradable sectors, the main reason for the divergence of the efficiency
values from the market values is the difference between the current rate of return to capital in
the sector, and the opportunity cost of the capital needed for the sector to expand. Thus, for
instance, water is a highly scarce and capital-intensive good in Namibia, and large additional
capital investments would be required to expand water delivery. At the same time, the rate of
return to the capital currently invested in the sector is low. Water, therefore, has the highest
APR of all the economic sectors. On the other hand, several other sectors currently have high
returns to capital and the opportunity cost of additional capital is lower, so their APRs are
low.
Further extensions to the work will greatly enhance the estimated APRs. Due to data
limitations, the work is confined to aggregate economic sectors, with no attention paid to
various subsector APRs. Such an activity of including estimating APRs also requires
resources that are beyond the capacity of this study. With improved data and adequate
resources, future work could aim to estimate subsector APRs.
19
Michael N Humavindu
Furthermore, due to economic conditions changing over time, it is inevitable that the work
would require intermittent updating. This would be best done by the central planning
agencies of the country. But despite the ample scope for further research with regard to
Namibian shadow pricing applications, this work represents a considerable improvement on
the current situation.
20
Estimating Namibian shadow prices within a semi-input–output framework
REFERENCES
Bank of Namibia (Various years): Annual report. Windhoek: Bank of Namibia.
Baumol, WJ (1977): Economic theory and operations analysis (4th edition). Englewood
Cliffs, NJ: Prentice-Hall.
Beyene, M, F Guta, H Kinfu, R Assegid, D Potts & J Weiss (1998): National economic
parameters and conversion factors for Ethiopia. Addis Ababa: Ministry of Economic
Development and Cooperation.
Bulmer-Thomas, V (1982): Input–output analysis in developing countries. New York: John
Wiley and Sons.
Central Bureau of Statistics (Various years): National Accounts. Windhoek: Republic of
Namibia.
Cornelisse, PA & CB Tilanus (1966): “The semi-input–output method – with an application
to Turkish data”. De Economist, 114(9/10):521–533.
Diao, X, B Fekadu, S Haggblade, T Seyoum, K Wamisho & B Yu (2007): “Agricultural
growth linkages in Ethiopia: Estimates using fixed and flexible price models”. IFPRI
Discussion Paper 00695. Washington, DC: International Food Policy Research
Institute.
Dondur, N (1996): “National economic parameters for Hungary”. Project Appraisal, 11(1):
41–50.
Dorosh, P & S Haggblade (2003): “Growth linkages, price effects and income distribution in
sub-Saharan Africa”. Journal of African Economies, 12(2):207–235.
21
Michael N Humavindu
Dorosh, PA, S Haggblade & RE Bernier (1992): “Agricultural growth linkages in
Madagascar”. Working Paper No. 22. Ithaca, NY: Cornell Food and Nutrition Policy
Program.
Dorosh, P, MK Niazi & H Nazli (2003): “Distributional impacts of agricultural growth in
Pakistan: A multiplier analysis”. Pakistan Development Review, 42(3):249–275.
Ghani, N (1999): “Semi-input-output analysis: A practical approach to interdependencies of
sectors, factors of production and primary factors”. Pakistan Economic and Social
Review, 37(2):197–214.
Haggblade, S, J Hammer & PBR Hazell (1991): “Modeling agricultural growth multipliers”.
American Journal of Agricultural Economics, 73(2):361–374.
Humavindu, MN (2008): “Estimating national economic parameters for Namibia”. Umeå
Economic Studies 744. Umeå: Department of Economics, Umeå University.
Kuyvenhoven, A (1978): “Planning with the semi-input–output method: With empirical
applications to Nigeria”. Studies in Development and Planning, Vol. 10. Leiden:
Martinus Nijhoff Social Sciences Division.
Lange, GM, K Schade, J Ashipala & N Haimbodi (2004): “A social accounting matrix for
Namibia, 2002: A tool for analysing economic growth, income, distribution and
poverty”. NEPRU Working Paper No. 97. Windhoek: Namibian Economic Policy
Research Unit.
Leontief, W (1986). Input–output economics. New York: Oxford University Press.
Londero, E (1996): Benefits and beneficiaries. Washington, DC: Inter-American
Development Bank.
Lopes da Silva, A (1993): “Estimates of national parameters for the economic analysis of
projects in Brazil”. Project Appraisal, 8(4):231–239.
22
Estimating Namibian shadow prices within a semi-input–output framework
MacArthur, JD (1994): “Estimating efficiency prices through semi-input–output methods – A
review of practice in available studies”. Journal of International Development,
6(1):19–43.
MacArthur, JD (1997): “Shadow pricing simplified: Estimating acceptably accurate
economic rates of return using limited data”. Journal of International Development,
9(3):367–382.
Medalla, E, C Rosario, V Pineda, R Querubin & E Tan (1990): “Re-estimation of shadow
prices for the Philippines”. Working Paper Series, No. 90:16. Manila: Philippine
Institute for Development Studies.
Miller, R & Peter Blair (1985): Input–output analysis: Foundations and extensions.
Englewood Cliffs, NJ: Prentice-Hall.
Nouman, G (1991): “Semi-input-output analysis: A practical approach to interdependencies
of sectors, factors of production and primary factors”. Pakistan Economic and Social
Review, 37(2):197–214.
Potts, D (1996): “Estimating shadow prices in a transitional economy: The case of
Lithuania”. In Kirkpatrick, & J Weiss (Eds). Cost–benefit analysis and project
appraisal in developing countries. Cheltenham: Edward Elgar.
Potts, D (2002): Project planning and analysis for development. London: Lynne Rienner
Publishers.
Powers, TA (1981): Estimating accounting prices for project appraisal. Washington, DC:
Inter-American Development Bank.
Saerbeck, R (1989): “National economic parameters for Botswana”. Research Monograph
No. 1. Bradford: Development and Project Planning Centre, University of Bradford.
23
Michael N Humavindu
United Nations (1999): “Handbook of input–output table compilation and analysis”. Studies
in methods: Handbook of National Accounting, Series F, No. 74. New York:
Department for Economic and Social Affairs, United Nations.
Weiss, J (1988): “An introduction to shadow pricing in a semi-input–output approach”.
Project appraisal, 3(4):182–189.
Weiss, J (1999): National economic parameters and economic analysis for the public
investment programme in Jamaica. Report to the Planning Institute of Jamaica.
World Travel and Tourism Council (2006): Namibia: The impact of travel and tourism on
jobs and the economy. London: WTTC.
ENDNOTES
1
Treatment of IO theory and applications can be found in Bulmer-Thomas (1982), Leontief (1986) and Miller &
Blair (1985).
2
Weiss (1988) and Potts (2002) provide helpful introductions to the SIO approach. Powers (1981) also provides
an account of the theory backed by empirical examples.
3
See Medalla et al. (1990) and Londero (1996). Various empirical studies employ the SIO methodology for
estimating sectoral multipliers in general and sector-specific shadow prices in particular. Examples of such work
are Lopes da Silva (1993), Dondur (1996), Potts (1996), Beyene et al. (1998), Nouman (1999), Weiss (1999),
Dorosh et al. (2003) and Dorosh & Haggblade (2003) and Diao et al. (2007).
4
Tourism is not classified as a separate economic sector in the National Accounts; rather, it is a combination of
products from various other sectors.
24
III
Integration and volatility spillovers in
African equity markets:
Evidence from Namibia and South Africa
Michael N Humavindu1
Development Bank of Namibia, PO Box 235, Windhoek, Namibia
[email protected]
Christos Floros
Department of Economics, University of Portsmouth, Portsmouth, PO1 3DE, UK
[email protected]
Abstract
This paper analyses returns and volatility on the Namibian and South African stock markets.
We use daily closing indices of the Namibian Stock Exchange (NSX) and the Johannesburg
Stock Exchange (JSE). The sample covers the period from January 4, 1999 to March 20,
2003. Our methodology has three main parts: (i) unit root tests, (ii) cointegration analysis and
(iii) volatility modelling. The results show that both markets exhibit very low correlations,
while there is no evidence of linear relationship between the markets. Furthermore, volatility
analysis shows evidence of no spillover effects. Our results suggest that NSX is an attractive
risk diversification tool for regional portfolio diversification in Southern Africa.
Keywords: financial returns, volatility, GARCH, cointegration, NSX, JSE
1
The authors would like to thank the anonymous referees for helpful comments on an earlier draft of this paper.
We accept responsibility for any remaining errors.
Integration and volatility spillovers in African equity markets: Evidence from Namibia and South Africa
1.
INTRODUCTION
The purpose of this paper is to re-examine the integration between the Namibian Stock
Exchange (NSX) and Johannesburg Stock Exchange (JSE) using daily data. This paper also
investigates volatility spillover effects between the markets as a complement to cointegration
analysis2. The major difference between this work and previous studies is that we focus on
the local Namibian index (which does not contain dual-listed stocks). Previous empirical
evidence reports strong integration between the Namibian and South African equity markets
(see Piesse & Hearn 2002; Tyandela & Biekpe 2001). A possible explanation for this is the
fact that they use the overall Namibian index (which contains dual-listed South African
stocks). This assertion does not refute the notion that stock prices across countries reflect
economic integration through trade linkages and foreign direct investment. The dividend
discount model posits that current share prices equal the value of future cash flows. These
cash flows are normally dependent on the earnings growth of companies, which are in turn
dependent on conditions of the domestic economy and its major trading partners. If
macroeconomic variables (inflation, unemployment) exhibit co-movement patterns, then it’s
likely the same for stock prices in such economies. Namibia and South Africa share strong
economic and historical (Apartheid) ties and thus high stock market integration is reasonably
expected.
Research into the integration of African equity markets is increasing. A possible explanation
is that African equity markets are potential outlets for international portfolio diversification
(IPD). Portfolio theory asserts that if returns on securities in a portfolio have a correlation of
2
Similar papers are given by Booth & Yioman (1997), Chen et al. (2002), and Koutmos & Booth (1995).
1
Michael M Humavindu and Christos Floros
less than one, then diversification can reduce risk. IPD is driven by the notion that the returns
of securities from different countries exhibit low correlations. Hence, through diversification,
risk can be lowered without sacrificing return3. Solnik (2000) observes that African markets
are less integrated with the capital markets from developed countries, and hence offer
diversification options. Consequently, Kenny and Moss (1998) report an increasing trend
among foreign investors buying into African stocks to capitalize on high returns and
diversification benefits. The surge in capital inflows to African equity markets raises the
important question of stock market integration: Are there in fact any benefits in diversifying
into these markets, and what is the extent of their integration?
Recent empirical evidence on the African markets indicates equity integration in countries
with strong economic ties. Piesse and Hearn (2002) investigate equity market integration
between JSE, NSX and the Botswana stock market (BSE). They use monthly data for the
period August 1993 to January 2000, and find integration between the JSE and NSX markets.
Chatterjee et al. (1998) apply cointegration analysis using monthly data and investigate
linkages among the Nigerian, Zimbabwean and Ghanaian markets. Their results show no
evidence of cointegration and indicate low correlations. Tyandela and Biekpe (2001) perform
a correlation analysis between weekly indices in the Southern African Development
Community (SADC). They find high correlation (around 90%) between NSX and JSE. This
is also followed by a high correlation (around 88.3%) between the NSX and BSE stock
indices. Overall the stock markets of South Africa, Namibia and Botswana are highly
correlated, impliying little diversification benefits among them. Finally, Appiah-Kusi and
Pescetto (1998) investigate volatility spillover aspects among African equity markets. They
3
See Grubel (1968).
2
Integration and volatility spillovers in African equity markets: Evidence from Namibia and South Africa
employ the GARCH (p, q) technique and find very little volatility spillover effect between
the markets. However, some integrated markets show strong economic links (Botswana,
South Africa and Zimbabwe).
This paper gives a comprehensive analysis of the NSX return dynamics and the extent of its
integration with the JSE market. Understanding the linkages with the JSE is useful for trading
strategies of foreign investors who venture into ‘frontier’ markets of Africa. The paper is
structured as follows. Section 2 describes the data. Section 3 shows the econometric
methodology adopted, while Section 4 evaluates the empirical results. Discussion and
conclusions are presented in Section 5.
2.
DATA
The data employed in this paper consists of daily closing values of the NSX local index and
the JSE overall index. The sample covers the period from January 4, 1999 to March 20, 2003.
The data for JSE was obtained from the Datastream, while the NSX data was obtained from
the Namibian Stock Exchange4. Daily data usually introduces the problem of nonsynchronous trading periods. However the trading periods of both exchanges are highly
synchronous (NSX does not trade on the public days of the JSE). We also circumvent this
problem by including the trading days of both exchanges during this period.5
4
On balance, the NSX local index is defined by no more than 15 securities during any year of the studies. In
fiscal year 2002 there were 14 locally listed firms and in 2005 there are 8 firms.
3
Michael M Humavindu and Christos Floros
For the empirical analysis, all index values are converted to returns. Brooks (2002) mentions
that there are various statistical reasons to avoid analysing stocks prices directly. To convert
our prices into returns we employ the following expression:
ri = ln(Pt/Pt-1)
(1)
where Pt is the index value in the current period and Pt-1 denotes the lagged (previous period)
value of the index.
3.
ECONOMETRIC METHODOLOGY
3.1
Unit roots
Unit root tests (with and without a deterministic trend) are employed for the lagged data
series in both levels and first differences following the Augmented Dickey Fuller (ADF)
approach6. The approach has its origins in Dickey and Fuller (1979) method and accounts for
possible variations in autocorrelations and ensures random residuals. The initial equation is
augmented by a sequence of differenced terms, and the overall model is assessed according to
the Akaike information criterion (AIC) and the Schwarz Bayesian criterion. The ADF test for
a unit root incorporating a deterministic trend is defined as –
5
The JSE does not stop trading on Namibian public/holidays. There are 1128 days of trading, while 52 days
removed to circumvent non-sychronous trading.
6
See Piesse & Hearn (2002).
4
Integration and volatility spillovers in African equity markets: Evidence from Namibia and South Africa
p 1
yt
D 1 Ii G t Iy t 1 ¦ Ii y t 1H t ,
t=1,…, n
(2)
i 1
where yt is the data series, t is a deterministic time trend and is a constant. Under the null
hypothesis, H0: ø = 1, the coefficients øi are assumed to be stochastically generated. We can
also define the corresponding test in first differences7:
p
'yt D UG t U yt 1 ¦ J i'yt 1 H i,
t=1,…, n
(3)
i 1
Except for the terms (1-ø)t and UG t , the corresponding test equations for unit roots without
a deterministic trend are as above.
3.2
Cointegration
Before testing for co-integration, we employ the Vector Autoregressive (VAR) model8. The
general VAR model is given by –
yt
a A1 yt 1 .. A p yt p Bxt H t
(4)
where y is a k vector of endogenous variables, x is a d vector of exogenous variables, A1,...,Ap
and B are matrices of coefficients to be estimated, and H is a vector of innovations that may
7
With the null H0: = (1-ø)=0.
8
See Harris (1995, p. 77-124))
5
Michael M Humavindu and Christos Floros
be contemporaneously correlated with each other but are uncorrelated with their own lagged
values and uncorrelated with all of the right-hand side variables.
Furthermore, we test for cointegration between the series under consideration. Given a group
of non-stationary series, we can determine whether there is co-movement between the series,
and if there is, we can identify the long-run equilibrium relationship (based on the Johansen’s
cointegration test, see Johansen 1991, 1995). Johansen’s method tests the restrictions
imposed by cointegration on the unrestricted VAR. We use the Johansen approach (Johansen
1988; Johansen & Juselius 1990) because it has several advantages over the Engle-Granger
two-step approach. The method of Johansen is given by the equation of the form –
'y t
G 3y t 1 ¦ *' y t i H t
(5)
where the i determines the number of lags specified in the dynamic VAR relationship9. y t is
a column vector of the two indices. If 3 has zero rank the variables in y t are noncointegrated. However, if the rank is r, there will be r possible stationary linear combinations.
3.3
Volatility modelling
We model the dynamics of volatility on the NSX following the GARCH approach (Engle
1982; Bollerslev 1986; Taylor 1986; Bollerslev et al. 1992). GARCH is an alternative model
to the ARCH model and has a more flexible lag structure than its predecessor. The
9
The method of Johansen method is based on the optimal lag structure of a VAR model.
6
Integration and volatility spillovers in African equity markets: Evidence from Namibia and South Africa
conditional variance of the error term in the GARCH model is given as a linear function of
the lagged squared residuals and the lagged conditional variance. A GARCH (p, q) process is
defined as –
Rt = a+t,
ht
tÐ t-1~N (0, ht)
p
q
i 1
j 1
D 0 ¦ D i H t2i ¦ E j ht j
j 0, j 0, for i= 0,…,p ;
(6)
j= 1,…., q
where Rt is the return for day t, a is the unconditional mean of Rt, t is an error term, ht is the
conditional variance of Rt based on past available information included in the information set
t-1. A major problem of this model is that constraints are necessary on the coefficients j and
j to ensure non-negativity. Another major deficiency of GARCH models is that they do not
capture the asymmetric response of volatility to news, since the sign of returns plays no role
in the specification of the model.
Furthermore, we capture asymmetries in the volatility using the Exponential GARCH
(EGARCH) model (Nelson 1991). EGARCH models captures skewness and allows the
ARCH process to be asymmetrical. The conditional variance is formulated as an exponential
function of the previous conditional variances and excess returns. The EGARCH model is
defined as –
Loghi2,t
D D
0
1
H
i, t 1
hi , t 1
ª H i, t 1
º
D 2 «
(2S )0.5 » E Loghi2,t 1
i, t 1
h
¬
¼
7
(7)
Michael M Humavindu and Christos Floros
hi2,t is the conditional volatility at time t and / i ,t 1 / is the absolute value of standardized
residuals. The asymmetric response to the last period shocks on current volatility is measured
by D 1 . This implies that if D 1 <0, negative past errors have greater impact on current variance
than the analogous positive errors. Since the coefficient is typically negative, positive return
shocks generate less volatility than negative return shocks. Thus, hi ,t is a function of both the
magnitude and the sign of the lagged errors. In this paper we also investigate volatility spillovers using a univariate EGARCH(1,1) model (for the NSX), which is augmented by
including conditional volatility of JSE. The model is given by –
Loghi2,t
D 0 D1
H i ,t 1
hi ,t 1
º
ª H i ,t 1
D2 «
(2S ) 0.5 » E Loghi2,t 1 Th 2j ,t 1
¼»
¬« hi ,t 1
(8)
where i denotes the primary country and j refers to the secondary country, while hj is the
conditional volatility of country j. The coefficient measures the extent of volatility
spillovers across markets. A significant implies that innovations in country j spill over to
country i.
3.4
Adjusting for thin trading
Dimson (1979), Miller, Muthuswamy and Whaley (1994) show that thin trading can
potentially lead to serial correlation in the return series. Thin trading is the outcome of either
non-synchronous trading or non-trading. Non-synchronous trading occurs when stocks trade
at every consecutive interval, but not necessarily at the close of each interval. This form of
thin trading has been studied by Scholes and Williams (1977a, 1977b) and Muthuswamy
8
Integration and volatility spillovers in African equity markets: Evidence from Namibia and South Africa
(1990). Non-trading occurs over a much shorter trading interval than non-synchronous
trading and thus all stocks in the market are unlikely to be traded at least once over that
interval. Fisher (1966), Dimson (1979), Cohen et al. (1978, 1979), Lo and MacKinlay (1990),
and Stoll and Whaley (1990a) also explain non-trading. Non-synchronous trading is
distinguished from non-trading by the interval over which price changes or returns are
computed. As the trading interval shrinks, non-synchronous trading becomes non-trading.
A major effect of thin trading is that an asset value over a certain time cannot be directly
observed if the asset is not traded during that period. As most market indices are computed
based on recent stock transactions, the reported index becomes stale if thin trading is present
Thus the observed index does not reflect the true value of the underlying stock portfolio. Thin
trading induces spurious serial correlation in the observed index returns. Thus the observed
dependence should not be construed as evidence of predictability, but rather a statistical
illusion due to thin trading.
There are different approaches to correct for thin trading. Bassett, France, and Pliska (1991)
suggest the use of a Kalman filter to estimate the distribution of the true index. Stoll and
Whaley (1990b) use the residual from an ARMA (p,q) regression as a proxy of the true index.
Jokivuolle (1995) proposes a modified version of the Stoll and Whaley approach to estimate
the true unobserved index from the history of the observed index. The correction consists of
decomposing the log of the observed index in its random and stationary components, using
the Beveredge and Nelson (1981) methodology. For this, the random component can be
shown to be equal to the log of the true index.
9
Michael M Humavindu and Christos Floros
In this paper we follow Miller, Muthuswamy, and Whaley (1994) to correct for thin trading.
Their methodology proposes the estimation of a moving average model (MA) that reflects the
number of non-trading days and then adjusts returns accordingly. Yet, in practice it is
difficult to identify non-trading days, under which Miller et al. (1994) show that it is
equivalent to estimating an AR (1) model. Such a model involves estimating the following
equation (i.e. random walk model):
Rt
a1 a 2 Rt 1 H t
(9)
The residuals from the regression are then used to estimate the adjusted returns as follows
adj
Rt
Ht
(10)
(1 a 2)
where Rtadj is the return at time t adjusted for thin trading.
The model by Miller et al. (1994) assumes that non-trading adjustment is constant over time
(which is true for highly liquid markets but not for emerging markets). Hence, equation (9) is
estimated recursively.
10
Integration and volatility spillovers in African equity markets: Evidence from Namibia and South Africa
4.
EMPIRICAL RESULTS
4.1
Descriptive statistics
First, the descriptive statistics are displayed in Table 1, while the plots of the data are in
Figure 1 (Time-series) and Figure 2 (Histograms). Table 1 shows that only JSE shows a
positive (but insignificant) mean return. The mean return for the NSX is negative over the
sample period. In terms of market volatility, as measured by standard deviations, NSX has
the highest value. The coefficients of skewness indicate that none of the two markets have
positively skewed returns. Thus, in terms of series distribution properties normality is
rejected (the probability of Jarque-Bera10 is < 0.05). These small and negative values indicate
long lean left tails. The kurtosis statistics indicate the peakedness of the distributions (NSX
return has the highest value). To indicate lean tails, the excess kurtosis statistic should have a
value of 3. The NSX kurtosis for returns indicates ‘fat tails’ and a sharp peakedness. Holmes
and Wong (2001) argue that presence of fat tails in returns supports the application of ARCH
models for the variance processes of returns.
10
The Jarque-Bera statistic has a chi-squared distribution with two degrees of freedom under the null hypothesis
of normally distributed errors.
11
Michael M Humavindu and Christos Floros
Table 1: Descriptive Statistics for price and returns of the NSX & JSE indices
JSE
LOGJSE
LOGNSX
NSX
RJSE
RNSX
Mean
8490.572
9.030883
4.433984
90.83240
0.000474
-0.000611
Median
8512.745
9.049320
4.406719
82.00000
0.000000
0.000000
Maximum
11653.36
9.363350
5.123964
168.0000
0.051979
0.120144
Minimum
5207.400
8.517355
3.850148
47.00000
-0.076893
-0.179586
Std. Dev.
1458.194
0.178537
0.389700
34.85873
0.011745
0.014207
Skewness
-0.020144
-0.459647
0.056722
0.375768
-0.141163
-0.881318
Kurtosis
2.584055
2.874449
1.526222
1.747203
6.249677
45.57039
Jarque-Bera
7.814853
38.52373
97.77381
95.51028
476.1449
81236.68
Probability
0.020092
0.000000
0.000000
0.000000
0.000000
0.000000
1074
1074
1074
1074
1074
1074
Observation
The plots show interesting results. The NSX index declines for most part of the sample
period. Local Namibian equities hardly trade and most institutional investors who buy them
prefer to hold them (and not trade them)11. The JSE index shows persistently upwards trend,
but with some downturns across the sample period. This feature indicates the susceptibility to
world markets downturns (especially for the 2000-2002 trading period). The graphs for the
returns (Figure 1) are useful in depicting some volatility features. Volatility seems to be more
prevalent in the JSE than in the NSX market. Also, the histograms (Figure 2) show that price
series and returns are not normally distributed.
11
The NSX market exhibits significant jumps (often referred to as seasonality in microstructure).
12
Integration and volatility spillovers in African equity markets: Evidence from Namibia and South Africa
5.2
9 .4
5.0
9 .3
4.8
9 .2
9 .1
4.6
9 .0
4.4
8 .9
4.2
8 .8
8 .7
4.0
8 .6
3.8
250
500
750
8 .5
1000
250
500
LOGNSX
.15
.06
.10
.04
.05
.02
.00
.00
-.05
-.02
-.10
-.04
-.15
-.06
-.20
250
500
750
1000
L O G JS E
750
-.08
1000
250
500
R(NSX)
750
1000
R(JSE)
Figure 1: Time series plots of price and returns for NSX & JSE indices
Density
Density
LJSE
1.5
N(s=0.179)
LNSX
N(s=0.389)
2
1.0
1
50
0.5
8.50
Density
8.75
DLJSE
9.00
9.25
9.50 3.50 3.75 4.00
Density
200
N(s=0.0117)
DLNSX
40
4.25
4.50
4.75
5.00
5.25
N(s=0.0142)
150
30
100
20
50
10
-0.075
-0.050
-0.025
0.000
0.025
0.050
-0.15
-0.10
-0.05
0.00
Figure 2: Histograms of price and return series for the NSX and JSE
13
0.05
0.10
5.50
Michael M Humavindu and Christos Floros
4.2
Correlation results
Table 2 presents the results from the correlation analysis. The results indicate negative and
significant correlations for both logarithmic series. However for the returns, the relationship
is positive but not high significant. This may suggest that there are opportunities for risk
diversification between the markets. The results obtained from the correlation matrix are not
quite useful for our analysis. Therefore, other methods (cointegration, GARCH) need to be
used.
Table 2: Correlation Matrix: Log index prices and index returns
LOGJSE
LOG(JSE)
LOGNSX
RJSE
RNSX
1
-0.7121458
-0.0264246
-0.0402273
LOG(NSX)
-0.7121458
1
0.0488854
0.0291513
R(JSE)
-0.0264246
0.0488854
1
0.0499905
R(NSX)
-0.0402273
0.0291513
0.0499905
1
4.3
Unit roots and cointegration results
The results from the ADF tests on both series (daily index in logarithms) and the first
differences are presented in Table 3 (JSE) and Table 4 (NSX). The results show that the logprices of both markets contain unit roots. However, the presence of unit roots is rejected in
the returns series of the two markets. The results show that daily indices in both markets are
integrated of order one, I(1). Furthermore, we employ a VAR model in order to specify the
number of lags we need for Johansen test. The results from the selected VAR model (with 4
14
Integration and volatility spillovers in African equity markets: Evidence from Namibia and South Africa
lags) are presented in Table 5. We find that previous periods’ log-prices depend (and show a
positive/negative effect) on the current log-prices for both markets. Furthermore, the results
of the cointegration analysis based on the Johansen test are presented in Table 6. Both the
Trace and Max-eigenvalue tests indicate no cointegration at both the 5% and 1% significant
levels. So, we conclude that there is no co-movement and long-run relationship between the
(logarithmic) indices of JSE and NSX stock markets.
Table 3: Unit roots results for the JSE market
Null Hypothesis: LOGJSE has a unit root
Exogenous: Constant
Lag Length: 3 (Automatic based on AIC)
t-Statistic
Prob.*
Augmented Dickey-Fuller test statistic
-2.577347
0.0980
Test critical values:
1% level
-3.436238
5% level
-2.864028
10% level
-2.568146
*MacKinnon (1996) one-sided p-values.
Null Hypothesis: D(LOGJSE) has a unit root
Exogenous: Constant
Lag Length: 2 (Automatic based on AIC)
Augmented Dickey-Fuller test statistic
t-Statistic
Prob.*
-18.65232
0.0000
15
Michael M Humavindu and Christos Floros
Table 4:Unit roots results for the NSX market
Null Hypothesis: LOGNSX has a unit root
Exogenous: Constant
Lag Length: 3 (Automatic based on AIC)
t-Statistic
Prob.*
Augmented Dickey-Fuller test statistic
-0.097304
0.9478
Test critical values:
1% level
-3.436238
5% level
-2.864028
10% level
-2.568146
*MacKinnon (1996) one-sided p-values.
Null Hypothesis: D(LOGNSX) has a unit root
Exogenous: Constant
Lag Length: 2 (Automatic based on AIC)
Augmented Dickey-Fuller test statistic
t-Statistic
Prob.*
-20.53571
0.0000
16
Integration and volatility spillovers in African equity markets: Evidence from Namibia and South Africa
Table 5: VAR estimates for cointegration analysis
Vector Autoregression Estimates with 4 lags
Sample (adjusted): 6 1076
t-statistics in [ ]
LOGJSE
LOGJSE(t-1)
LOGJSE(t-2)
LOGJSE(t-3)
1.134741
0.026966
[ 37.0476]*
[ 0.71489]
-0.085892
-0.120242
[-1.86253]*
[-2.11727]*
-0.126232
0.106807
[-2.74062]*
[ 1.88298]*
0.070514
-0.022042
[ 2.31529]*
[-0.58768]
-0.036432
0.979411
[-1.44658]
[ 31.5782]*
0.044029
-0.006386
[ 1.24776]
[-0.14696]
-0.035612
-0.058540
[-1.00998]
[-1.34815]
0.027002
0.082693
[ 1.07522]
[ 2.67382]*
0.066847
0.088713
[ 2.22958]*
[ 2.40269]*
LOGJSE(t-4)
LOGNSX(t-1)
LOGNSX(t-2)
LOGNSX(t-3)
LOGNSX
LOGNSX(t-4)
C
Log likelihood
6314.230
Akaike information criterion
-11.75767
Schwarz criterion
-11.67403
* Significant at 5% level.
17
Michael M Humavindu and Christos Floros
Table 6: Cointegration results (Johansen method)
Trend assumption: Linear deterministic trend
Series: LOGJSE LOGNSX
Lags interval (in first differences): 4
Unrestricted Cointegration Rank Test (Trace)
Hypothesized
Trace
0.05
No. of CE(s) Eigenvalue
Statistic
Critical Value Prob.**
None
0.011167
12.97655
15.49471
0.1157
At most 1
0.000898
0.960772
3.841466
0.3270
Trace test indicates no cointegration at the 0.05 level
Unrestricted Cointegration Rank Test (Maximum Eigenvalue)
Hypothesized
Max-Eigen
0.05
No. of CE(s) Eigenvalue
Statistic
Critical Value Prob.**
None
0.011167
12.01577
14.26460
0.1101
At most 1
0.000898
0.960772
3.841466
0.3270
Max-eigenvalue test indicates no cointegration at the 0.05 level
* denotes rejection of the hypothesis at the 0.05 level
**MacKinnon-Haug-Michelis (1999) p-values
18
Integration and volatility spillovers in African equity markets: Evidence from Namibia and South Africa
4.4
Mean and volatility spillover effects
First we run AR(1) from equation 9 in order then to estimate adjusted returns. Table 7
presents the results from AR(1) for JSE (Part A) and NSX (Part B). Part A shows positive
and significant slope coefficient, while Part B shows negative but not significant coefficient.
In other words, previous’ period returns for JSE market depend (they have positive effect) on
the current returns. The estimated adjusted returns12 (using the residuals from AR(1) models)
are plotted in Figure 3. Both plots confirm that returns are stationary.
Table 7: AR (1) results for the JSE and NSX
PART A
Dependent Variable: R(JSE)
Method: Least Squares
Sample (adjusted): 4 1076
White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable
Coefficient
Std. Error
t-Statistic
Prob.
a1
0.000403
0.000360
1.121161
0.2625
a2
0.143736
0.045125
3.185289*
0.0015
12
The ADF test in adjusted returns suggest that both series are I(0). These results are available upon request.
19
Michael M Humavindu and Christos Floros
PART B
Dependent Variable: R(NSX)
Method: Least Squares
Sample (adjusted): 4 1076
White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable
Coefficient
Std. Error
t-Statistic
Prob.
a1
-0.000621
0.000434
-1.428626
0.1534
a2
-0.013584
0.024004
-0.565899
0.5716
*Significant.
.15
.08
.10
.04
.05
.00
.00
-.05
-.04
-.10
-.08
-.12
-.15
250
500
750
1000
-.20
250
500
750
Adjusted Returns (NSX)
Adjusted Returns (JSE)
Figure 3: Adjusted returns of the JSE and NSX indices
20
1000
Integration and volatility spillovers in African equity markets: Evidence from Namibia and South Africa
Furthermore, Table 8 presents the results from the Exponential-GARCH (EGARCH) model.
The results indicate significant EGARCH effects in the data (for both markets)
13
. The
leverage effect term (denoted as 2) is present only in the JSE market (2 is negative and
statistically different from zero).
Table 8: EGARCH (1, 1) for NSX and JSE
Equation and variables NSX returns JSE returns
Mean Equation
Coefficient Coefficient
0.000155
a
0.000530
(0.374703) (1.624306)
Variance equation
-1.200539
D0
(-1.675187) (-3.632725)*
0.048753
D1
-0.065734
(2.172064)* (-2.040622)*
0.861507
E
0.216813
(0.831296) (3.587454)*
0.150220
D2
-0.737571
0.936613
(10.28913)* (47.18847)*
Note: Figures in parentheses are t-statistics. * Significant at 5% level.
13
adj
The results of the simple EGARCH models with Rt
are available upon request.
21
Michael M Humavindu and Christos Floros
The conditional variance (continuous volatility) series from EGARCH models are plotted in
Figure 4 and Figure 5 for NSX and JSE, respectively. The variance series obtained from
EGARCH models confirm that variance for both JSE and NSX markets is changing over time
(i.e. volatility is time-varying). This characteristic is captured well from EGARCH models.
.0 0 1 2
.0 0 1 0
.0 0 0 8
.0 0 0 6
.0 0 0 4
.0 0 0 2
.0 0 0 0
2 5 0
E G A R C H
5 0 0
v a ria n c e
7 5 0
s e r ie s
1 0 0 0
(N S X )
Figure 4: NSX EGARCH volatility
.0 0 1 4
.0 0 1 2
.0 0 1 0
.0 0 0 8
.0 0 0 6
.0 0 0 4
.0 0 0 2
.0 0 0 0
2 5 0
E G A R C H
5 0 0
v a ria n c e
7 5 0
s e rie s ( J S E )
Figure 5: JSE EGARCH volatility
22
1 0 0 0
Integration and volatility spillovers in African equity markets: Evidence from Namibia and South Africa
Over the sample period, JSE is characterised by volatility clustering, with periods of high
volatility followed by less volatile periods14. Furthermore, NSX does exhibit volatility
clustering, but for most periods the plot seems to indicate a market that hardly trades15. The
volatility spillovers effects between the markets are reported in Table 9. The variables of
volatility spillover for both markets (indicated by ) are not significant. In other words, there
is no empirical evidence of volatility spillover between the markets under consideration16.
Our results confirm the previous findings from the correlations and cointegration analysis.
Table 9: Mean and Volatility spillovers estimated from EGARCH (1, 1) model
PART A. JSE
adj
Dependent Variable: R JSE
Method: ML - ARCH (Marquardt)
Sample(adjusted): 5 1076
Bollerslev-Wooldrige robust standard errors & covariance
Mean Equation
a
Coefficient
-9.41E-05
Std. Error
t-Statistic
0.000371 -0.253615
Prob.
0.7998
Variance Equation
a0
-0.629032
0.180506 -3.484822*
0.0005
14
The results are in line with Appiah-Kusi and Pescetto (1998) for the South African market.
15
Sherbourne and Stork (2004) indicate that the NSX local market is characterized by investors who ‘sit’ on the
stocks (to fulfil legislation requirements) and hardly trade in these local equities.
16
The results from AR(1)-EGARCH(1,1) model, not reported here, support this evidence.
23
Michael M Humavindu and Christos Floros
Mean Equation
Coefficient
Std. Error
t-Statistic
Prob.
a1
0.199992
0.061398 3.257317*
0.0011
a2
-0.068405
0.035607 -1.921086*
0.0547
E
0.945746
0.018040 52.42467*
0.0000
T
15.09191
131.9428
0.9089
0.114382
PART B. NSX
adj
Dependent Variable: R NSX
Method: ML - ARCH (Marquardt)
Sample(adjusted): 5 1076
Bollerslev-Wooldrige robust standard errors & covariance
Mean Equation
Coefficient
Std. Error
t-Statistic
Prob.
a
0.000372
0.000424
0.878391
0.3797
Variance Equation
a0
-2.107921
1.661831
-1.268433 0.2046
a1
0.138169
0.049074
2.815500* 0.0049
a2
0.159867
0.079372
2.014132* 0.0440
E
0.745396
0.202474
3.681448* 0.0002
T
-708.5497
559.2028
-1.267071 0.2051
* Significant at 5% level
24
Integration and volatility spillovers in African equity markets: Evidence from Namibia and South Africa
5.
DISCUSSION AND CONCLUSIONS
This paper reports evidence of portfolio diversification between the NSX and JSE markets for
the period January 1999 - March 2003. We investigate correlations, cointegration and
volatility spillover effects using daily data. Formal relationship between the two markets has
implications for investors who want to diversify their portfolio regionally.
The results show that returns exhibit very low correlations.17 In addition, cointegration
analysis shows no long-run relationship between the markets. Hence, the NSX local market
may offer an attractive opportunity for investors to diversify risk regionally. However the
results are not consistent with previous papers (Piesse & Hearn 2002; Tyandela & Biekpe
2001). They find high correlations and cointegration between the markets. A possible
explanation is that they consider the NSX overall index (which includes dual-listed stocks on
both exchanges).
In this paper we use the NSX local index (which is constituting of Namibian non-dual listed
shares only). The volatility analysis shows asymmetric leverage effects for both markets
under consideration. However there is no evidence of spillover effects from the JSE market to
the NSX market, while there are no volatility spillover effects from the NSX to the JSE. In
general, this paper provides little evidence of integration between the NSX local market and
the JSE market. The results suggest that NSX is an attractive risk diversification tool for
regional portfolio diversification in Southern Africa.
17
Our focus on the applicability of the findings is directed at portfolio composition.
25
Michael M Humavindu and Christos Floros
The fact that the NSX local market hardly trades diminishes the implications of the results.
The NSX should strive to adopt strategies that can ensure trading in local equities since they
do provide risk diversification potential. A good strategy, that should gain currency, is to
promote exchange-traded funds (ETFs) and invest in all of the stocks contained in the NSX
overall index. This will ensure that some trading in local equities does take place.
We strongly believe that our results are helpful to financial analysts and managers as well as
academics and investors. Certainly the findings can benefit those who want to trade
(assuming there is a liquid NSX market). Future work should incorporate practical modelling
realities. We should therefore employ monthly and weekly time series for comparisons of
results. We should also examine whether the asymmetries are in a way with the traditional
leverage effect explanation. We also need to estimate time-varying risk premia (using
Multivariate-GARCH models) to further investigate whether volatility changes are
accompanies by compensating changes in the estimated market risk premia. Finally, a
bivariate Markov switching model can be used to empirically investigate whether changes of
volatility in one market precedes the changes in volatility to the other.
26
Integration and volatility spillovers in African equity markets: Evidence from Namibia and South Africa
REFERENCES
Appiah-Kusi, J & G Pescetto (1998): “Volatility and volatility spillovers in emerging
markets: The case of African stock markets”. Ekonomia, 2(2):171–185.
Basset, GW, VG France & SR Pliska (1991): “Kalman filter estimation for valuing nontrading securities, with application to the MMI cash-futures spread on October 19 and
20, 1987”. Review of Quantitative Finance and Accounting, 1:135–151.
Beverdge, S & CR Nelson (1981): “A new approach to decomposition of economic time
series into permanent and transitory components with particular attention to
measurement of the business cycle”. Journal of Monetary Economics, 7:151–174.
Bollerslev, T (1986): “Generalized Autoregressive Conditional Heteroskedasticity”. Journal
of Econometrics, 31:307–327.
Bollerslev, T & JM Wooldridge (1992): “Quasi-maximum likelihood estimation and
inference in dynamic models with time varying covariances”. Econometric Reviews,
11:143–172.
Bollerslev, T, RY Chou & KF Kroner (1992): “ARCH modeling in finance: A review of the
theory and empirical evidence”. Journal of Econometrics, 52(1):5–59.
Booth, G & MT Yioman (1997): “Price volatility spillovers in Scandinavian stock markets”.
Journal of Banking and Finance, 21:811–823.
Brooks, C (2002): Introductory econometrics for finance. Cambridge: Cambridge University
Press.
Chatterjee, A, OF Ayadi & UB Dufrene (1998): “International portfolio diversification in the
African emerging equity markets: A cointegration analysis”. Journal of Emerging
Markets, 3(2):49–60.
27
Michael M Humavindu and Christos Floros
Cohen, KJ, SF Maier, RA Schwartz & DK Whitcomb (1978): “The returns generation
process, returns variance, and the effect of thinness in securities markets”. Journal of
Finance, 33:149–167.
Cohen, KJ, SF Maier, RA Schwartz & DK Whitcomb (1979): “On the existence of serial
correlation in an efficient securities market”. TIMS Studies in the Management
Sciences, 11:151–168.
Chen G, M Firth & Rui (2002): “Stock market linkages: Evidence from Latin America”.
Journal of Banking and Finance, 26:1113–1141.
Dickey, DA & WA Fuller (1979): “Distribution of the estimators for autoregressive time
series with a unit root”. Journal of the American Statistical Association, 74:427–431.
Dimson, E (1979): “Risk measurement when shares are subject to infrequent trading”.
Journal of Financial Economics, 7:197–226.
Engle, RF (1982): “Autoregressive Conditional Heteroskedasticity with estimates of the
variance of UK inflation”. Econometrica, 50:987–1008.
Fisher, L (1966): “Some new stock market indexes”. Journal of Business, 39:191–225.
Grubel, H (1968): “Internationally diversified portfolios: Welfare gains and capital flows”.
American Economic Review, 58:1299–1314.
Harris, R (1995): Using cointegration analysis in econometric modelling. Prentice
Hall/Harvester Wheatsheaf.
Holmes, P & MW Wong (2001): “Foreign investment, regulation and price volatility in
South-East Asian stock markets”. Emerging Markets Review, 2:371–386.
Johansen, S (1988): “Statistical analysis of cointegration vectors”. Journal of Economic
Dynamics and Control, 12:231–254.
Johansen, S (1991): “Estimation and hypothesis testing of cointegration vectors in Gaussian
Vector Autoregressive Models”. Econometrica, 59:1551–1180.
28
Integration and volatility spillovers in African equity markets: Evidence from Namibia and South Africa
Johansen, S (1995): Likelihood based inference on cointegration in the Vector Autoregressive
Model. Oxford: Oxford University Press.
Johansen, S & K Juselius (1990): “Maximum likelihood estimation and inferences on
cointegration – with applications to the demand for money”. Oxford Bulletin of
Economics and Statistics, 52:169–210.
Jokivuolle, E (1995): “Measuring true stock index value in the present of infrequent trading”.
The Journal of Financial and Quantitative Analysis, 30(3):455–464.
Kenny, C &T Moss (1998): “Stock markets in Africa: Emerging lions or white elephants?”.
World Development, 26:829–843.
Koutmos, G & G Booth (1995): “Asymmetric volatility transactions in international stock
markets”. Journal of International Money and Finance, 14(6):747–762.
Lo, A & A Mackinlay (1990): “An econometric analysis of infrequent trading”. Journal of
Econometrics, 45:181–211.
MacKinnon, JG (1996): “Critical values for cointegration tests”. In Engle, RF & CWJ
Granger (Eds). Long-run economic relationships: Readings in cointegration. Oxford
University Press.
Miller, MH, J Muthuswamy & RE Whaley (1994): “Mean reversion of standard and Poor 500
index basis changes: Arbitrage-induced or statistical illusion?”. Journal of Finance,
49:479–513.
Muthuswamy, J (1990): “Nonsynchronous trading and the index autocorrelation problem”.
PhD dissertation, Graduate School of Business, University of Chicago.
Nelson, D (1991): “Conditional heteroskedasticity in asset returns: A new approach”.
Econometrica, 59(2):347–370.
29
Michael M Humavindu and Christos Floros
Piesse, J & B Hearn (2002): “Equity market integration versus segmentation in three
dominant markets of the Southern African Customs Union”. Applied Economics,
14:1711–1722.
Scholes, M & J Williams (1977a): “Estimating betas from nonsynchronous data”. Journal of
Financial Economics, 5:309–327.
Scholes, M & J Williams (1977b): “Estimating betas from nonsynchronous data”. Working
Paper, Graduate School of Business. Chicago: University of Chicago.
Sherbourne, R & C Stork (2004): “The Namibian Stock Exchange and domestic asset
requirements: Options for the future”. IPPR Research Report No. 5. Windhoek:
Institute for Public Policy Research.
Solnik, B (2000): International investments (4th edition). Pearson Addison Wesley.
Stoll, R & E Whaley (1990a): “The dynamic of stock index and stock index futures returns”.
Journal of Financial and Quantitative Analysis, 25:441–468.
Stoll, R & E Whaley (1990b): “Stock market structure and volatility”. Review of Financial
Studies, 3:37–71.
Taylor, S (1986): Modelling financial time series. New York: Wiley.
Tyandela, L & N Biekpe (2001): “Construction of all SADC stock market indices”. African
Finance Journal, 3(1):59–76.
30
IV
Environment and Development Economics 8: 391–404 © 2003 Cambridge University Press
DOI:10.1017/S1355770X03000202 Printed in the United Kingdom
Hedonic pricing in Windhoek townships
MICHAEL NOKOKURE HUMAVINDU
Directorate of Environmental Affairs, Ministry of Environment and
Tourism, Windhoek, Namibia.
JESPER STAGE
Department of Economics, Umeå University, 901 87 Umeå, Sweden. Tel: 46
90 786 53 36. Fax: 46 90 77 23 02. Email [email protected]
ABSTRACT. This study applies the hedonic pricing model to property sales in the
township areas in Windhoek, the capital city of Namibia, where municipal authorities
have pursued a programme of selling plots of land to settlers in order to encourage
them into a formalized economic situation. We find that, apart from house quality,
access to the central business district, access to marketplaces and access to transportation, environmental quality also has a large impact on property prices. Properties
located close to a garbage dump sell at considerable discounts, while properties located
close to a combined conservation and recreation area sell at premium prices. The results
thus suggest that the hedonic pricing method can be useful for studying townships
in developing countries, and that this can help to clarify the importance of environmental factors which are otherwise frequently neglected in town planning for township
settlements.
1. Introduction
The objective of this paper is to study whether property prices in the township areas in Windhoek, the capital city of Namibia, are influenced by
attractive attributes in a similar fashion to prices in more developed property markets. This is analysed using the hedonic pricing method. This
method has been applied in developing countries only rarely (see
Malpezzi, 1999, for an overview), and hardly ever in township areas.
The authors would like to thank A. P. Brockerhoff, Wolf Gaerdes and Louis Lessing
of Windhoek Municipality; John Mendelsohn, Carole Roberts, and Tony Robertson
of the Directorate of Environmental Affairs, Ministry of Environment and Tourism;
and Peder Axensten, Niklas Hanes, Jörgen Hellström, Robert Sörensson, and Lars
Westin of Umeå University, for help with various stages in the data collection and
data analysis. We would also like to thank a number of people at Umeå University,
particularly Thomas Aronsson, Karl-Gustaf Löfgren, Niklas Rudholm, and Krister
Sandberg, as well as three anonymous referees and an associate editor, for constructive comments on earlier drafts of this paper. The usual disclaimer applies.
Finally, we would like to thank Jerome Kisting, then an economics student at the
University of Namibia, for arguing that this valuation method was completely irrelevant to most inhabitants in the developing world, thus giving us the idea for this
study in the first place.
392 Michael Nokokure Humavindu
It might seem obvious that households living in townships will spend
what little money they have on goods which are crucial for their shortterm survival, and are unlikely to be willing to pay extra for a property
with, for instance, a pleasing view. It might also seem obvious that even if
township dwellers are in fact willing to pay premia for slightly more
attractive dwellings, they are unlikely to have the necessary overview of
the property market to know when attractive properties are available and
be able to bid for them. However, if attributes such as environmental
quality do affect property prices, even among extremely poor households,
it is important to town planners to be aware of this, as these preferences
should be reflected in policy decisions.
Only a few hedonic pricing studies have been carried out in Africa,1 and
these did not look at property markets in townships; in most of the
informal and semi-formal settlements around large African cities the settlers do not have clear title to their land, and even when there are
permanent or near-permanent property rights, trade in these properties is
usually poorly documented.
In Windhoek, the municipality has pursued a policy of selling plots of
land by instalments to low-income households and ultra low-income
households – currently defined as households with monthly incomes of
less than 1860 N$ (160 US$) and less than 500 N$ (45 US$), respectively –
moving in from rural areas, in order to encourage them into a more formalized economic situation. Due to this policy, reliable information on
property prices is more easily available for township areas in Windhoek
than it is in most similar areas in developing countries. This means that
Windhoek is one of the few places where the hedonic pricing method can
be applied relatively easily to ultra low-income housing, in order to
examine which factors are considered important by township inhabitants
and what impact these factors have on property prices.
In the next section, the political and demographic background to the
current township policies in Windhoek is presented. This is followed by a
section which explains the theoretical framework for the study and discusses property attributes which might affect prices. The next section
describes the econometric model and the data set used, followed by a
section describing the results of the analysis. The concluding section discusses implications of the results.
2. The Windhoek townships
Windhoek lies in central Namibia. It was the colonial capital of what was
then called South West Africa during the German colonial period, and subsequently during South African rule, which ended in 1990. The city is now
the capital of independent Namibia and serves as the administrative, legislative, and judicial centre of the country.
The first of the present-day township areas in Windhoek was established
during the South African apartheid system. Before the 1960s, the black
population of Windhoek lived in the Old Location, a site west of the central
business district. Residential blocks were rented from the municipality and
1
Asabere, 1981a, 1981b; Megbolugbe, 1989; Arimah, 1992; Akpom, 1996.
Environment and Development Economics 393
inhabitants built their own houses. During the late 1950s through to the
1970s, the expansion of the ‘white section’ of Windhoek towards the Old
Location led to the forcible relocation of black residents to a site north-west
of the city centre. The new site was called Katutura, which means ‘the
place where we do not stay’ in one of the local languages.
The South African authorities adopted new housing policies as well, in
order to minimize the construction of urban dwellings and to focus more
on the provision of ‘temporary’ accommodation. Katutura initially consisted of 4,000 rental houses, a barrack-like single quarters area and a
walled compound to accommodate migrant labour. The rental units were
divided into ethnic group sections and were uniform in appearance,
quality, and size. There were also general regulations to stem the influx of
blacks into Windhoek (Pendleton, 1974).
The 1970s and 1980s witnessed a liberalization of the regulations governing black residents of Windhoek and other urban centres (Haines and
Tapscott, 1991). From the 1980s, plots of land – called ‘erven’ in Namibia
– and houses could be privately owned in Katutura. New areas west of
the established townships were also opened for settlement (Pendleton,
1994).
As a result of the liberalized laws, urbanization increased dramatically.
The population of Windhoek grew from approximately 100,000 in 1985 to
approximately 145,000 at independence in 1990 (Windhoek City Council,
1996). Squatter settlements sprang up on the outskirts of Katutura and the
other township areas.
Windhoek’s population has continued growing after independence and
is now believed to be roughly 250,000. Most of the urban expansion has
taken place in the township areas to the north and north-west of
Windhoek. Before the apartheid regulations were relaxed, the entire black
settlement was limited to an area of about 400 hectares located between 4.5
and 7 km from the city centre; currently, the furthest township settlements
are over 12 km from the city centre, and the entire township area covers
well over 2,300 hectares.
The average monthly household income in the townships is estimated
(Windhoek City Council, 1996) to be about N$850 (approximately 75US$).
These very low incomes seriously limit the level of services that can be provided at cost to the inhabitants of these areas. Many city governments in
Africa faced similar problems with low-income settlers in the 1960s and
1970s. These local governments often attempted to provide subsidized
full-service housing for city residents who could not afford cost-recovery
tariffs, and limit migration to the city in order to keep down the costs of
subsidies to low-income areas. However, the result has frequently been
that these cities have ended up both with costly subsidies to the formally
recognized low-income areas, and also (since the city governments cannot
afford to provide subsidized full-service housing for all settlers) with large
informal settlements where residents have no access to any municipal services and no legal rights whatsoever. The uncertainty of tenure and
constant threat of eviction in the informal settlements have led to low
levels of community involvement; these settlements are often characterized by high crime rates and other problems.
394 Michael Nokokure Humavindu
The local government in Windhoek, which faced the problem of rapid
urbanization later than local governments elsewhere in Africa, has
attempted to learn from these experiences. The municipal township policies aim both towards being financially viable and towards integrating
settlers in the formal economy.
Rather than provide subsidized full-service housing, the municipality
permits settlers to lease or purchase unused municipal land. Land purchases can be paid either in cash or through loans from the municipality at
market interest rates, with the purchased property as collateral. Once the
municipality has sold a property there are no restrictions on the resale
price, provided that any remaining debt to the municipality is paid in full.
Municipal services, such as water and electricity, are optional, but are
available at cost-recovery prices for those who choose to make use of them.
Refuse collection is the only municipal service which is compulsory for
all erven; each erf has its own refuse bin which is emptied once a week
either by municipal trucks (in the older townships) or by private contractors. Illegal garbage dumping in open areas was becoming a major
problem throughout the city in the early 1990s, but after municipal authorities improved refuse collection, converting some of the illegal dumps
into officially recognized dumping sites in the process, nearly all garbage
is now collected and dumped in the officially recognized locations. Unlike
many township areas in developing countries, Windhoek’s townships
therefore do not have any widespread sanitation problems related to
uncollected refuse at present.
Several public sector agents have been involved in the provision of
affordable housing in the township areas after independence. The government’s Build Together Programme was designed shortly after
independence to provide credit for building and building improvements
to ultra low-income households. The programme also provided technical
assistance to the program beneficiaries. Poor repayment levels, and high
subsidies from government, characterized the programme. The stateowned National Housing Enterprise (NHE) was also set up in order to
provide low-cost housing. However, due to profitability problems, the
NHE moved away from catering for the ultra low-income groups and
began targeting slightly higher-income groups. The NHE is currently reorganizing and plans to start building houses for the lowest-income
categories once more, but at present most construction in the township
areas is thus done either by residents or by other private agents.
There are few employment opportunities in the township areas; only
about 10 per cent of the township population are estimated to work there.
Those township residents who are formally employed primarily work in
the central areas of the city in the central business district or in the nearby
(no longer very aptly named) Windhoek North area, while unemployed
gather in the open areas in the central parts of the city in the hope of being
picked up by households or small businesses which need to hire labour for
short-term jobs. Nearly all work-related travel in the township areas is
therefore to and from the central business district, either on foot or by car.
There is a considerable number of relatively cheap, privately operated
taxis which carry large numbers of passengers at a time; rates are fixed by
Environment and Development Economics 395
a central association so that travellers from the townships pay the same
rates for a specific destination regardless of where in the township areas
they are picked up. In principle, these taxis are only permitted to pick up
passengers at specially designated taxi ranks throughout the city, but this
rule is only enforced intermittently; however, while taxis are permitted to
drop off passengers anywhere, they charge less for delivering passengers
to a designated taxi rank than they do for delivering passengers elsewhere.
Since Windhoek is still a fairly small city, many people with steady jobs are
picked up by their employers, while people without steady jobs may walk
into town to search for employment (Windhoek City Council, 1996).
Although there are few employment opportunities in the township
areas, there is a great deal of other activity going on. Schools are available
throughout the area. The municipality has also established a number of
market places, where commodities are traded and where cultural activities
take place. A large area around the Goreangab dam, where water is stored
for the dry season, was set aside as a combined conservation and recreation area in the late 1960s. Although the townships have since
expanded and now almost surround it, the area has been preserved and is
one of the largest open areas in Windhoek. It has considerable scenic
appeal and is used for activities such as barbecues, hiking, boating, and
picnicking.
3. Hedonic pricing
Real estate characteristics such as the area of the plot or the distance to the
nearest school are not themselves traded in any markets; they are tied to
the individual property being sold and are only traded as parts of the
bundle of characteristics constituting that particular property. However,
by examining the prices paid for different bundles of characteristics, it is
possible to estimate the value attached to a specific characteristic. This is
the basis for the hedonic pricing model (Rosen, 1974; Sheppard, 1999).
A property is characterized by a vector of attributes, H h1, h2, …, hk,
and the hedonic pricing method attempts to establish the relationship
between housing expenditure P(H) and the levels of the various attributes,
P(H) f(h1, h2, …, hk). If the price relationship is correctly specified, and if
property markets are functioning efficiently, it becomes possible to determine households’ implicit marginal valuation of each attribute, Pi P(H)/hi. Attributes which have been studied in hedonic studies fall into
two major groups; structural attributes of the property, such as the plot
size, the size of the house, the number of rooms, and the building materials
used; and location-specific property attributes, such as the distance to the
city centre, access to transport, and environmental and socio-economic
characteristics of the neighbourhood.
The few empirical studies which have been made of housing markets in
African countries have all indicated that access variables are important in
determining property prices. Asabere (1981a, 1981b) found that nearness
to the city centre and quality of nearby roads had an impact on property
prices in two Ghanaian cities. Megbolugbe (1989), Arimah (1992), and
Akpom (1996), in their studies of different Nigerian cities, similarly found
that a number of variables, measuring access to labour markets and/or
396
Michael Nokokure Humavindu
access to services, were important. The coverage of structural attributes in
these analyses varied considerably, from studies which only looked at the
sizes of the traded plots to studies which had access to detailed information on building materials as well as on the number and type of rooms
of each house. Most studies, however, ignored the issue of environmental
quality. The two studies by Asabere did include variables measuring
environmental bads (and found these to have significant impacts on property prices in the two cities studied), but the later studies did not take such
factors into account.
There are generally two stages to a hedonic pricing study; the first stage
is the estimation of implicit prices for various attributes, while the second
stage is the estimation of the implicit demand functions determining these
prices. The implicit prices of different attributes in a property market,
which reflect the marginal valuation of these attributes, can be estimated
using sales prices of properties and data on the attributes involved.
However, these implicit prices are in turn determined by the equilibria of
implicit supply and demand functions, which are affected by a large
number of factors. The market clearing implicit prices will be set through
a bargaining process between the agents in the property market and will
be affected by factors specific to the households buying and selling properties – household sizes, income levels, income distribution, and so on.
It is only possible to estimate the underlying implicit demand functions
for different attributes by including data, not only on the traded properties, but also on the households involved in the property market at hand.
Of the African studies cited above, only Arimah (1992) had this type of
household information and was able to proceed beyond estimating
implicit prices to estimating the implicit demand functions. In Windhoek’s
township areas, there is no detailed household-level information available
on variables, such as income, employment, or household size. At the
moment, there are in fact not even reliable figures available on the total
number of inhabitants in the different township areas, let alone inhabitants
in individual households, and the only data on income levels are estimated
average figures which are not sufficient for any detailed analysis. This
analysis is therefore limited to estimating the implicit prices, rather than
the implicit demand functions, for different property attributes.
Although formal segregation has been abolished for over a decade, it is
still unthinkable for white or coloured households in Windhoek to move
into the township areas in the northern and north-western parts of town,
regardless of household income or house price. Likewise, although the
former white neighbourhoods have seen an influx of black families in the
past years, there is still considerable reluctance on the part of white homeowners to sell their houses to blacks. This means that there are, effectively,
two separate housing markets in Windhoek, making it problematic to
apply one single hedonic model for the entire area.
Frequently, hedonic studies have attempted to capture market segmentation between different areas by using switching regressions; the area
being studied is divided into discrete segments, the model is estimated for
each segment separately, and the results for different segments are then
compared to see whether the differences are significant. However, there
Environment and Development Economics 397
are problems with using this approach on spatial data such as property
prices, since the delineation of the areas becomes crucial for the results.
Variation within the studied areas will produce misleading results, and
where there are significant differences between different market segments
the model will predict unrealistically large price differences between
neighbouring plots at the borders between those segments (Can, 1992). The
same problems occur when dummy variables are used for different market
segments, an approach used in many hedonic pricing studies (including
several of the African applications discussed earlier). In this study, rather
than attempting to model the precise relationships between pricing of
attributes in different sections of the city, we have chosen to limit the
analysis to the township area. Moreover, since the township areas are all
located in close proximity to each other rather than spread around the city
– the latter frequently being the case with townships in other cities – we
have also chosen not to subdivide the area by introducing neighbourhood
dummies, in order to avoid the delineation problems noted above.
For the individual properties being traded, information on the sizes of
the relevant erven is readily available. Unfortunately, detailed information
on the structural attributes of individual houses, which could also be
expected to affect prices, is not. However, when a house has been built, the
municipality makes a valuation of the replacement cost of that house. Any
changes made to the house have to be reported to the municipality, which
then makes a new valuation. This means that the municipal replacement
cost valuation can be used as a measure of overall house quality. Still,
experiences from other hedonic pricing studies indicate that factors such
as house size, number of rooms and building materials are very important
in determining property prices, and, although a measure of plot size and a
proxy measure of overall house quality are considerably better than
nothing, it would definitely have been preferable to have more specific
information on the buildings.
A number of access variables might be expected to be of importance in
determining property prices in townships. Namibian roads are of high
quality compared to those in other African countries and, unlike several of
the studies cited earlier, we have therefore not included any measure of
road quality. However, other access variables which are more likely to
play a role in determining property prices are the distances to the central
business district and to the nearest major market. Factors such as the access
to taxi ranks, and the walking distance to the nearest school, might also
play an important role in determining real estate prices in townships,
where few households own their own car (Blaauw et al., 1998).
While economic valuation of public goods has not been a major part of
urban planning in Windhoek or elsewhere in Namibia, a recent survey
(Humavindu and Masirembu, 2001) indicated that the Goreangab dam
recreation area was perceived as important by township inhabitants and
that they wished to have it preserved. It is of interest to see whether this
stated preference for the site is also reflected in actual market behaviour, in
which case properties with easy access to the area should be regarded as
attractive and might be expected to sell at premium prices. Alternatively,
the municipal garbage dumps which are located throughout the city are
398 Michael Nokokure Humavindu
probably not appreciated by their neighbours. If this lack of appreciation
for the dumps is reflected in property prices, one would expect a downward pressure on property prices in the vicinity of a dump.
4. Econometric specification and data
Economic theory provides no a priori reason to prefer one functional form
for the hedonic price function over others, and hedonic pricing studies
have frequently used Box–Cox transformations to find the functional form
that fits the data best. However, several authors (Cassel and Mendelsohn,
1985; Cropper, Deck, and McConnell, 1988; Sheppard, 1999) have argued
that it is problematic to use Box–Cox transformations in hedonic pricing
studies, both because the resulting parameter estimates tend to be highly
sensitive to small variations in the data and also because those parameter
estimates are frequently difficult to interpret. These authors have
suggested using simpler functional forms which produce more stable parameter estimates.
The use of a simple functional form is especially recommended in situations such as the one studied here, where some potentially important
attributes (such as house size or number of rooms) are not included due to
limitations in the data set. Rather than using polynomial expressions or
Box–Cox transformations, we have therefore chosen to test the following,
quite simple, model for the price of property i
Pi 1Sizei 2RCHi 3dCBDi 4dMarketi 5dSchoooli 6dRanki 7Garbagei 8Goreani 9dGori i
Size is erf size in square meters and RCH is the official municipal valuation
in N$ of the replacement cost of the house. dCBD is the distance to the
central business district where most of the township inhabitants find
employment (if any), dMarket is the distance to the nearest major marketplace where they are likely to make most of their purchases, dSchool is the
distance to the nearest school, and dRank is the distance to the nearest taxi
rank; all these distances are measured in meters. It is assumed here that the
Euclidean distance is a reasonable approximation of the actual travel distance, which is usually the case for dense road networks (Puu, 1997) such
as those in Windhoek’s township areas.
In order to study whether environmental quality has an impact on property prices, two dummy variables and one continuous variable are used.
Garbage is a dummy variable for proximity to garbage dumps, which takes
the value 1 for plots which are less than 250 m from a garbage dump and
0 for plots which are not. The reason for using a dummy rather than the
continuous Euclidean distance is that the perceived aesthetic difference
between a plot adjacent to a garbage dump and one 500 m away is likely to
be considerably greater than the perceived aesthetic difference between a
plot 1 km away from a dump and one 1.5 km away. While this type of consideration might alternatively have been captured by using both linear and
quadratic forms of the distances, this would have increased the risk discussed earlier of making the estimates highly sensitive to small variations
in the data and estimating parameters incorrectly, because of the missing
variable problem caused by the lack of structural information on houses.
Environment and Development Economics 399
The reason for choosing 250 m as a cutoff distance is that this captures
properties in housing blocks adjacent to a garbage dump, while excluding
properties located in housing blocks further off. Similarly, Gorean is a
dummy variable which takes the value 1 for plots which are less than
250 m from the Goreangab dam recreational area and 0 for plots which are
not; this variable is intended to capture the value of living directly adjacent
to the recreational area, with an attractive view and extreme ease of access
to the area. dGor, finally, is the distance to the recreational area, and is
intended to capture the ease of access to the area for those plots which are
located further off.2
Windhoek municipality registers sales prices, official property valuation, and erf area when individual erven are traded, and also if the
property is being sold by the municipality or by a close relative of the
buyer. The full data set (Windhoek City Council, 2001) consisted of 551
recorded sales of residential erven in the northern and north-western
suburbs (Goreangab, Hakahana, Katutura, Okuryangava, and Wanaheda)
during 1999. Of these, 72 were sales either by the municipality or by a close
relative of the buyer and were excluded from the sample, leaving a total of
479 sales in the reduced data set. Combining detailed maps of Windhoek
(Windhoek City Council, 2001) with GIS software, it has been possible to
calculate the distances from the centre point of each traded erf to the centre
points of, respectively, the central business district, the nearest major marketplace, the nearest school, and the nearest taxi rank, as well as to the
nearest garbage dump and to the Goreangab dam recreation area.
It deserves to be noted (table 1) that the average valuation of building
investments in the traded properties is close to 47,000 N$, so many households clearly spend large amounts of time and/or money improving their
dwellings once they have bought them. One of the goals of the settlement
formalization programme has, of course, been to encourage people in the
township areas to take greater responsibility for their surroundings, so this
effect was to be expected, but similar behaviour has been observed in
township areas in other developing countries where households only have
semi-permanent squatter rights and do not actually own their properties
(Jimenez, 1982). However, for some of the traded properties (28 of the
properties which remained in the reduced data set) the official valuation of
the replacement cost is zero, i.e. any existing structures are of such poor
quality that the municipality believes that, if destroyed, they could be
rebuilt at negligible cost. This means that the sample includes houses
ranging from the extreme lower end of the market to fairly high-quality
dwellings.
One may also note the considerable variation in the size of plots. Until
1997, the smallest erf size permitted by the Ministry of Regional and Local
Government and Housing was, in principle, 300 m2, a minimum which has
since been lowered to 200 m2, and well over half of the traded plots have
sizes between 200 m2 and 400 m2. However, some plots have sizes which
2
An earlier version of this paper used only the dummy variable; we thank a referee
for pointing out that this would only capture the ‘view’ aspect of the Goreangab
area.
400 Michael Nokokure Humavindu
Table 1. The data set
Price
Size
RCH
dCBD
dMarket
dSchool
dRank
Garbage
Gorean
dGor
N 479
Unit
Average
Minimum
Maximum
N$
m2
N$
m
m
m
m
Dummy
Dummy
m
60,192
337
46,737
6,862
870
435
344
0.06
0.01
2,661
2,060
131
0
4,587
103
57
41
0
0
118
220,000
1,191
266,300
9,323
2,763
1,468
2,223
1
1
4,955
Notes: Price is the sales price of each property; Size is the plot size; RCH is the
municipal valuation of the replacement cost of the house on the property;
dCBD, dMarket, dSchool, and dRank are the Euclidean distances to the central
business district, the nearest market, the nearest school, and the nearest taxi
rank, respectively; Garbage is a dummy variable which is 1 for plots which are
less than 250 m from a garbage dump and 0 for plots which are not, while
Gorean is a dummy variable which is 1 for plots which are less than 250 m
from the Goreangab dam recreation area and 0 for plots which are not;
finally, dGor measures the Euclidean distance to the Goreangab dam
recreation area.
are less than the official minimum, while a few plots are far greater than
the official minimum size.
The distance to the city centre varies by almost 5 km for the traded properties. As noted earlier the townships currently extend to a distance of
approximately 12 km from the city centre, but many of the outermost settlements have been established relatively recently and the properties there
have not yet been resold. Most properties are located relatively close to a
school and a taxi rank, but the average distance to the nearest major
market is considerably greater. The distance to the Goreangab reserve,
finally, varies from dwellings located in blocks directly adjacent to it to
dwellings located almost 5 km away.
5. Results
The model presented above was estimated using an ordinary least squares
regression. As a White test indicated the presence of heteroscedasticity, the
standard errors in the regression were corrected for heteroscedasticity
using a White estimator (White, 1980). The results are presented in table 2.
Results using semi-log and log formulations3 are shown for comparison in
tables 3 and 4; the results in terms of significant variables are largely
similar, but the linear form has greater explanatory power.
Some attributes which could potentially be important, such as individual attributes of houses, were not included in the available data,
3
In the log formulation, we used zero rather than ln(RCH) for those properties
where the replacement cost was valued at 0.
Environment and Development Economics 401
Table 2. Estimation results for the linear form
Size
RCH
dCBD
dMarket
dSchool
dRank
Garbage
Gorean
dGor
Constant
R2 0.4520
F(9, 469) 42.28
Coefficient
Robust SE
t-value
5.13
0.72
6.64
9.88
6.42
16.18
34706.26
21801.38
3.32
92367.02
19.94
0.07
2.10
3.57
8.42
6.92
4086.45
8032.49
1.82
19553.41
0.26
10.47
3.16
2.77
0.76
2.34
8.49
2.71
1.82
4.72
Notes: See table 1.
Table 3. Estimation results for the semi-log form
Size
RCH
dCBD
dMarket
dSchool
dRank
Garbage
Gorean
dGor
Constant
R2 0.3720
F(9, 469) 34.77
Coefficient
Robust SE
t-value
0.000339
0.000011
0.000211
0.000344
0.0000875
0.000582
0.769
0.740
0.000051
12.060
0.000379
1.59E-06
0.0000532
0.0000963
0.000209
0.000186
0.149
0.286
0.000046
0.451
0.895
6.936
3.971
3.575
0.42
3.125
5.149
2.588
1.109
26.723
Notes: See table 1. The regression used the logarithm of the dependent price
variable.
Table 4. Estimation results for the log form
Size
RCH
dCBD
dMarket
dSchool
dRank
Garbage
Gorean
dGor
Constant
R2 0.3477
F(9, 469) 67.27
Coefficient
Robust SE
t-value
0.728
0.125
0.904
0.226
0.011
0.172
0.886
0.794
0.033
15.961
0.120
0.019
0.385
0.077
0.087
0.081
0.144
0.213
0.102
4.150
6.044
6.72
2.351
2.951
0.123
2.117
6.153
3.736
0.321
3.846
Notes: See table 1. The regression used the logarithms of all variables except
the two dummy variables; for the variable RCH, the value 0 rather than
ln(RCH) was used in the 28 cases when RCH took zero values.)
402 Michael Nokokure Humavindu
leading to a relatively low R2 of 0.45 (the R2s for the other specifications
were even lower). Even so, at a 5 per cent significance level the results
support the hypothesis that several other attributes of the traded properties have an effect on property prices in the township areas. The R2 is not
much lower than those in several of the other hedonic pricing studies cited
earlier, and the F statistic for the entire regression is 42.28, which is also
significant at the 5 per cent level.
Erf size does not appear to have a significant impact on property prices,
while housing quality and nearness to the city centre do have significant
impacts. This suggests that the recent decrease in the statutory minimum
erf size can potentially be welfare enhancing because it means that the
municipality can open up for further densification of the older township
areas, making it possible for people in recently established townships
further out to move closer to the city centre. (Incidentally, the marginal
valuation of an additional N$’s worth of building investments is lower
than 1, which means that there are no arbitrage gains to be made by sellers
through making additional investments before selling.)
Proximity to a school has no significant effect on property prices – the
point estimate of the parameter even has the ‘wrong’ sign. This is presumably due to the fact that the distance to a school is fairly short for most
properties in any case, so that an additional meter is not perceived as particularly important. Despite the limited enforcement of the rank system
(which should mean that many commuters are able to catch a taxi
wherever they want to anyway) the distance to the nearest taxi rank has a
significant impact on property prices. Many people use commuter taxis to
travel into town and back, and it appears that they attach considerable
importance to having easy access to taxi transport. This indicates that the
municipal policy of establishing taxi ranks and taxi services quickly in
newly settled areas is likely to be appreciated by inhabitants and may play
an important role for the municipality’s success in integrating new settlers
into the local economy.
The distance to the nearest major market has a significant effect on property prices (the marginal valuation of an additional meter is actually higher
for the distance to the nearest market than it is for the distance to the city
centre), indicating that households attach considerable importance to
having access to major marketplaces. Although one cannot say anything
with confidence without having estimates of the underlying implicit
demand functions as well, it is possible that the cost of establishing
additional market places might be more than offset by the resulting
increase in social welfare. This is, at any rate, something that deserves to
be studied more closely.
The two dummy variables for environmental quality both had significant impacts on property prices. Proximity to a garbage dump is clearly
viewed as unattractive; the mean effect is to reduce the value of a property
by almost 35,000 N$. Close proximity to the Goreangab dam recreation
area, on the other hand, raises the value of a property by almost 22,000 N$.
The distance to the Goreangab area, on the other hand, does not appear to
have a significant impact on property prices. Thus, having a pleasant view
is valued highly while the ease of access does not have an impact on prices
Environment and Development Economics 403
for plots further away.4 This finding is in line with an earlier study
(Humavindu and Masirembu, 2001) which indicated that travel distance to
Goreangab did not affect how frequently people living in the township
areas visited the site. One possible explanation for this might be that many
people go to Goreangab by taxi and thus pay a fixed rate regardless of
where in the township area they start.
6. Conclusions
This paper has shown that property prices in Windhoek’s townships
reflect attractive and unattractive location-specific characteristics,
including proximity to environmental goods and bads. Public policy determines many of these location-specific attributes, and the results indicate
that the hedonic pricing method can be useful for evaluating public policy
not only in affluent neighbourhoods but also in townships. Keeping track
of property sales in townships, in the way that municipal authorities have
done in Windhoek, can thus provide urban authorities responsible for
administering townships with a powerful additional tool for policy
analysis.
Lack of reliable detailed information on household characteristics is
likely to be a problem for township studies in other cities as well. The usefulness of the method could nonetheless be increased further by at least
recording household characteristics at the time of sale/purchase of a property, even if it is likely be difficult for many municipal authorities to
update this household-level information on a regular basis. Property
market segmentation, and the welfare effects of this, also warrants further
exploration, as this is likely to be a complicating factor in the analysis of
property markets in many developing country cities.
An important finding in this study is the high value that inhabitants in
the township areas clearly attach to environmental quality. Proximity to a
conservation area or to garbage dumps have remarkably large impacts on
property prices. Townships in other developing country cities have often
been allowed to expand under less organized circumstances than in
Windhoek, and the issues of maintaining refuse disposal and secluding
garbage dumps from residential areas, as well as maintaining open spaces
in township areas have frequently been neglected by urban authorities.
Our results indicate that where this neglect has occurred it may have been
a very serious omission.
References
Akpom, U.N. (1996), ‘Housing attributes and the cost of private rental buildings in
Lagos, Nigeria’, Review of Regional Studies 26: 351–365.
Arimah, B.C. (1992), ‘Hedonic prices and the demand for housing attributes in a
third world city: the case of Ibadan, Nigeria’, Urban Studies 29: 639–651.
Asabere, P.K. (1981a), ‘The determinants of land values in an African city: the case
of Accra, Ghana’, Land Economics 57: 385–397.
4
Dropping one of the two ‘Goreangab’ variables does not affect the result for the
other one, so it would appear that the two different aspects of the Goreangab area
studied here are not closely related.
404
Michael Nokokure Humavindu
Asabere, P.K. (1981b), ‘The price of urban land in a chiefdom: empirical evidence on
a traditional African city, Kumasi’, Journal of Regional Science 21: 529–539.
Blaauw, L., J. Motinga, J.N. Oranje, E. Kaakunga, M. Humavindu, S. Masirembu, I.
Ngishoongele, and S. Shuuya (1998), ‘Needs assessment on residential erven
demand of Windhoek residents’, mimeo, Namibia Economic Policy Research
Unit (Nepru), Windhoek.
Can, A. (1992), ‘Specification and estimation of hedonic housing price models’,
Regional Science and Urban Economics 22: 453–474.
Cassel, E. and R. Mendelsohn (1985), ‘The choice of functional forms for hedonic
price equations: comment’, Journal of Urban Economics 18: 135–142.
Cropper, M.L., L.B. Deck, and K.E. McConnell (1988), ‘On the choice of functional
form for hedonic price functions’, Review of Economics and Statistics 70: 668–675.
Haines, R. and C. Tapscott (1991), ‘Health, housing and education’, in E. Leistner
and P. Esterhuysen (eds), Namibia 1990: An Africa Institute Country Survey,
Johannesburg: Colorpress, pp. 173–181.
Humavindu, M.N. and S. Masirembu (2001), ‘The economic value of urban recreation in Namibia: results of a pilot study at Avis and Goreangab dams’,
Directorate of Environmental Affairs Working Paper No. 7, Directorate of
Environmental Affairs, Ministry of Environment and Tourism, Windhoek.
Jimenez, E. (1982), ‘The value of squatter dwellings in developing countries’,
Economic Development and Cultural Change 30: 739–752.
Malpezzi, S. (1999), ‘Economic analysis of housing markets in developing and transition economies’, in E.S. Mills and P. Cheshire (eds), Handbook of Regional and
Urban Economics, Amsterdam: Elsevier, pp. 1791–1864.
Megbolugbe, I.F. (1989), ‘A hedonic index model: the housing market of Jos,
Nigeria’, Urban Studies 26: 486–494.
Pendleton, W.C. (1974), Katutura: A Place Where We Do Not Stay, San Diego: San
Diego State University Press.
Pendleton, W.C. (1994), Katutura: A Place Where We Stay, Windhoek: Gamsberg
Macmillan.
Puu, T. (1997), Mathematical Location and Land Use Theory, Berlin: Springer Verlag.
Rosen, S. (1974), ‘Hedonic prices and implicit markets: product differentiation in
pure competition’, Journal of Political Economy 82: 34–55.
Sheppard, S. (1999), ‘Hedonic Analysis of Housing Markets’, in E. S. Mills and P.
Cheshire (eds), Handbook of Regional and Urban Economics, Amsterdam: Elsevier,
pp. 1595–1635.
White, H. (1980), ‘A heteroscedasticity-consistent covariance matrix estimator and
a direct test for heteroscedasticity’, Econometrica 48: 817–838.
Windhoek City Council (1996), 1995 Residents Survey Report: Volume One,
Windhoek: Windhoek City Council Corporate Communications Division.
Windhoek City Council (2001), unpublished data.
Avhandlingar framlagda vid Institutionen för nationalekonomi,
Umeå universitet
List of dissertations at the Department of Economics, Umeå
University
Holmström, Leif (1972) Teorin för företagens lokaliseringsval. UES 1. PhLic
thesis
Löfgren, Karl-Gustaf (1972) Studier i teorin för prisdiskriminering. UES 2. PhLic
thesis
Dahlberg, Åke (1972) Arbetsmarknadsutbildning - verkningar för den enskilde
och samhället. UES 3. PhD thesis
Stage, Jørn (1973) Verklighetsuppfattning och ekonomisk teori. UES 4. PhLic
thesis
Holmlund, Bertil (1976) Arbetslöshet och lönebildning - kvantitativa studier av
svensk arbetsmarknad. UES 25. PhD thesis
Löfgren, Karl-Gustaf (1977) En studie i neokeynesiansk arbetslöshets- och
inflationsteori. UES 34. PhD thesis
Lundberg, Lars (1976) Handelshinder och handelspolitik - Studier av verkningar
på svensk ekonomi. Industriens Utredningsinstitut, Stockholm. PhD thesis
Johansson, Per-Olof (1978) Sysselsättning och samhällsekonomi - En studie av
Algots etablering i Västerbotten. UES 53. PhD thesis
Wibe, Sören (1980) Teknik och aggregering i produktionsteorin. Svensk
järnhantering 1850-1975; en branschanalys. UES 63. PhD thesis
Ivarson, Lars (1980) Bankers portföljvalsbeteende. En teoretisk studie. UES 64.
PhD thesis
Batten, David (1981) Entropy, Information Theory and Spatial Input-output
Analysis. UES 92. PhD thesis
Hårsman, Björn (1982) Housing Demand Models and Housing Market Models for
Regional and Local Planning. Swedish Council for Building Research,
D13:1981. PhD thesis
Holm, Magnus (1983) Regionalekonomiska modeller för planering och
samordning i en decentraliserad ekonomi. Byggforskningsrådet, R118:1981
and R5:1983. PhD thesis
Ohlsson, Henry (1986) Cost-Benefit Analysis of Labor Market Programs Applied to a Temporary Program in Northern Sweden. UES 167. PhLic
thesis
Sarafoglou, Nikias (1987) A Contribution to Population Dynamics in Space. UES
179. PhD thesis
Ohlsson, Henry (1988) Cost-Benefit Analysis of Labor Market Programs Applied to a Temporary Program in Northern Sweden. UES 182. PhD thesis
Anderstig, Christer (1988) Applied Methods for Analysis of Economic Structure
and Change. CERUM 1988:2, Umeå University. PhD thesis
Karlsson, Charlie (1988) Innovation Adoption and a Product Life Cycle. UES 185.
PhD thesis
Löfström, Åsa (1989) Diskriminering på svensk arbetsmarknad - En analys av
löneskillnader mellan kvinnor och män. UES 196. PhD thesis
Axelsson, Roger (1989) Svensk arbetsmarknadsutbildning - En kvantitativ analys
av dess effekter. UES 197. PhD thesis
Zhang, Wei-Bin (1989) Theory of Economic Development - Nonlinearity,
Instability and Non-equilibrium. UES 198. PhD thesis
Hansson, Pär (1989) Intra-Industry Trade: Measurements, Determinants and
Growth - A study of Swedish Foreign Trade. UES 205. PhD thesis
Kriström, Bengt (1990) Valuing Environmental Benefits Using the Contingent
Valuation Method: An Econometric Analysis. UES 219. PhD thesis
Aronsson, Thomas (1990) The Short-Run Supply of Roundwood under Nonlinear
Income Taxation - Theory, Estimation Methods and Empirical Results
Based on Swedish Data. UES 220. PhD thesis
Westin, Lars (1990) Vintage Models of Spatial Structural Change. UES 227. PhD
thesis
Wikström, Magnus (1992) Four Papers on Wage Formation in a Unionized
Economy. UES 287. PhD thesis
Westerlund, Olle (1993) Internal Migration in Sweden - The Role of Fiscal
Variables and Labor Market Conditions. UES 293. PhLic thesis
Bergman, Mats A. (1993) Market Structure and Market Power. The Case of the
Swedish Forest Sector. UES 296. PhD thesis
Johansson, Per (1993) Count Data Models - Estimator Performance and
Applications. UES 315. PhD thesis
Roson, Roberto (1994) Transport Networks and the Spatial Economy - A General
Equilibrium Analysis. UES 340. PhD thesis
Li, Chuan-Zhong (1994) Welfare Evaluations in Contingent Valuation - An
Econometric Analysis. UES 341. PhD thesis
Østbye, Stein (1994) Regional Labour and Capital Subsidies - Theory and
Evidence of the Impact on Employment under Wage Bargaining. UES 344.
PhLic thesis
Westerlund, Olle (1995) Economic Influences on Migration in Sweden. UES 379.
PhD thesis
Mortazavi, Reza (1995) Three Papers on the Economics of Recreation, Tourism
and Property Rights. UES 396. PhLic thesis
Østbye, Stein (1995) Regional Labour and Capital Subsidies. UES 397. PhD
thesis
Hussain-Shahid, Imdad (1996) Benefits of Transport Infrastructure Investments:
A Spatial Computable General Equilibrium Approach. UES 409. PhD thesis
Eriksson, Maria (1996) Selektion till arbetsmarknadsutbildning. UES 410. PhLic
thesis
Karlsson, Niklas (1996) Testing and Estimation in Labour Supply and Duration
Models. UES 413. PhD thesis
Olsson, Christina (1996) Chernobyl Effects and Dental Insurance. UES 428.
PhLic thesis
Vredin, Maria (1997) The African Elephant - Existence Value and Determinants
of Willingness to Pay. UES 441. PhLic thesis
Eriksson, Maria (1997) To Choose or not to Choose: Choice and Choice Set
Models. UES 443. PhD thesis
Widerstedt, Barbro (1997) Employer Change and Migration. Two Papers on
Labour Mobility in Sweden. UES 444. PhLic thesis
Lundberg, Sofia (1997) The Economics of Child Auctions in 19th Century
Sweden. UES 445. PhLic thesis
Lundberg, Johan (1997) Two Papers on Revenue and Expenditure Decisions in
the Swedish Local Public Sector. UES 454. PhLic thesis
Widerstedt, Barbro (1998) Moving or Staying? Job Mobility as a Sorting Process.
UES 464. PhD thesis
Bask, Mikael (1998) Essays on Exchange Rates: Deterministic Chaos and
Technical Analysis. UES 465. PhD thesis
Löfgren, Curt (1998) Time to Study Students: Two Essays on Student
Achievement and Study Effort. UES 466. PhLic thesis
Sjögren, Tomas (1998) Union Wage Setting in a Dynamic Economy. UES 480.
PhD thesis
Mortazavi, Reza (1999) Essays on Economic Problems in Recreation, Tourism
and Transportation. UES 483. PhD thesis
Rudholm, Niklas (1999) Two Essays on Generic Competition in the Swedish
Pharmaceuticals Market. UES 485. PhLic thesis
Olsson, Christina (1999) Essays in the Economics of Dental Insurance and Dental
Health. UES 494. PhD thesis
Marklund, Per-Olov (1999) Environmental Regulation and Firm Efficiency. UES
504. PhLic thesis
Berglund, Elisabet (1999) Regional Entry and Exit of Firms. UES 506. PhD thesis
Hellström, Jörgen (1999) Count Data Autoregression Modelling. UES 507. PhLic
thesis
Nordström, Jonas (1999) Tourism and Travel: Accounts, Demand and Forecasts.
UES 509. PhD thesis
Johansson Vredin, Maria (1999) Economics Without Markets. Four papers on the
Contingent Valuation and Stated Preference Methods. UES 517. PhD thesis
Schei, Torbjørn (2000) Natural recreation resources: production and a diversity of
interests related to the management of grouse as an outfield resource in
Finnmark, Norway, in the Euro-Arctic Barents region. UES 523. PhLic
thesis
Backlund, Kenneth (2000) Welfare Measurement, Externalities and Pigouvian
Taxation in Dynamic Economies. UES 527. PhD thesis
Andersson, Linda (2000) Job Turnover, Productivity and International Trade.
UES 530. PhLic thesis
Ylvinger, Svante (2000) Essays on Production Performance Assessment. UES
531. PhD thesis
Bergkvist, Erik (2001) Freight Transportation. Valuation of Time and Forecasting
of Flows. UES 549. PhD thesis
Rudholm, Niklas (2001) The Swedish Pharmaceuticals Market - Essays on Entry,
Competition and Antibiotic Resistance. UES 552. PhD thesis
Lundberg, Johan (2001) Local Government Expenditures and Regional Growth in
Sweden. UES 554. PhD thesis
Lundberg, Sofia (2001) Going Once, Going Twice, SOLD! The Economics of
Past and Present Public Procurement in Sweden. UES 557. PhD thesis
Eliasson, Kent (2001) University Enrollment and Geographical Mobility: The
Case of Sweden. UES 558. PhLic thesis
Samakovlis, Eva (2001) Economics of Paper Recycling. Efficiency, policies, and
substitution possibilities. UES 563. PhD thesis
Daunfeldt, Sven-Olov (2001) Essays on Intra-Household Allocation and Policy
Regime Shifts. UES 570. PhD thesis
Hellström, Jörgen (2002) Count Data Modelling and Tourism Demand. UES 584.
PhD thesis
Andersson, Linda (2002) Essays on Job Turnover, Productivity and State-Local
Finance. UES 586. PhD thesis
Rashid, Saman (2002) Invandrarinkomster, förvärvsdeltagande och familj. UES
588. PhLic thesis
Hanes, Niklas (2003) Empirical Studies in Local Public Finance: Spillovers,
Amalgamations, and Tactical Redistributions. UES 604. PhD thesis
Stenberg, Anders (2003) An Evaluation of the Adult Education Initiative Relative
Labor Market Training. UES 609. PhD thesis
Stage, Jesper (2003) Mixing Oil and Water. Studies of the Namibian Economy.
UES 611. PhD thesis
Marklund, Per-Olov (2004) Essays on Productive Efficiency, Shadow Prices, and
Human Capital. UES 621. PhD thesis
Rashid, Saman (2004) Immigrants' Income and Family Migration. UES 625. PhD
thesis
Sandberg, Krister (2004) Hedonic Prices, Economic Growth, and Spatial
Dependence. UES 631. PhD thesis
Sjöström, Magnus (2004) Factor Demand and Market Power. UES 633. PhD
thesis
Nilsson, William (2005) Equality of Opportunity, Heterogeneity and Poverty.
UES 652. PhD thesis
Quoreshi, Shahiduzzaman (2005) Modelling High Frequency Financial Count
Data. UES 656. PhLic thesis
Ankarhem, Mattias (2005) Bioenergy, Pollution, and Economic Growth. UES
661. PhD thesis
Quoreshi, Shahiduzzaman (2006) Time Series Modelling of High Frequency
Stock Transaction Data. UES 675. PhD thesis
Ghalwash, Tarek (2006) Income, Energy Taxation, and the Environment. An
Econometric Analysis. UES 678. PhD thesis
Westerberg, Thomas (2006) Two Papers on Fertility – The Case of Sweden. UES
683. PhLic thesis
Simonsen, Ola (2006) Stock Data, Trade Durations, and Limit Order Book
Information. UES 689. PhD thesis
Eliasson, Kent (2006) College Choice and Earnings among University Graduates
in Sweden. UES 693. PhD thesis
Selander, Carina (2006) Chartist Trading in Exchange Rate Theory. UES 698.
PhD thesis
Humavindu, Michael N (2007) Essays on Public Finance and Environmental
Economics in Namibia. UES 705. PhLic thesis
Norberg-Schönfeldt, Magdalena (2007) The Phase-Out of the Nuclear Family?
Empirical Studies on the Economics and Structure of Modern Swedish
Families. UES 708. PhD thesis
Granlund, David (2007) Economic Policy in Health Care: Sickness Absence and
Pharmaceutical Costs. UES 710. PhD thesis
Jonsson, Thomas (2007) Essays on Agricultural and Environmental Policy. UES
719. PhD thesis
Broberg, Thomas (2007) The Value of Preserving Nature – Preference
Uncertainty and Distributional Effects. UES 720. PhD thesis
Witterblad, Mikael (2008) Essays on Redistribution and Local Public
Expenditures. UES 731. PhD thesis
Thunström, Linda (2008) Food Consumption, Paternalism and Economic Policy.
UES 739. PhD thesis
Humavindu, Michael N (2008) Essays on the Namibian Economy. UES 745. PhD
thesis
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement