Development Accounting for the African Economy

(1)

Development Accounting

for the African Economy

(2)

ABSTRACT

With an emphasis on the whole African economy, the present thesis asks three questions that are quintessential for modern economic development: What are the sources of development accounting, what are the productivity dynamics that feature Africa’s economies and what are the fundamentals behind its structural transformation? The study is motivated by the fact that Africa has joined for the first time ever since independence the small club of growth economies that lifted the global economic growth around the mid-1990s. The answer to these questions is important to inquire whether Africa’s growth resurgence is sustainable or indicative of yet another false start that Africa is accustomed to.

Key words: Development Accounting, Labour Productivity, Structural Change, Africa.

Student: Maria Teresa Bou González m.t.bou.gonzalez@student.rug.nl Student Number: S2637197

Supervisor: Tarek M. Harchaoui t.m.harchaoui@rug.nl

Co-assessor: Robbert Maseland r.k.j.maseland@rug.nl

MSc International Economics and Business

University of Groningen, Faculty of Economics and Business

Note that the source reference for tables and figures in this paper is available in Table 20, Appendix C.

(3)

I. Motivation

In the recent years, the African continent has rapidly become the central topic of many studies (see McMillan and Rodrik 2014 and the references therein). The rapid economic growth that the continent has been experiencing, which markedly contrast with its poor track record has led to a large body of literature meant to understand the factors that held back Africa and the prospects that this continent can now offer. This paper contributes to the debate by shedding light on three related issues: 1) We perform a development accounting exercise in an effort to gain some insight about the factors underlying the pattern of Africa’s development; 2) Africa’s development is not uniform. It is full of contrasts, which motivates the need to examine the productivity distribution to assess whether poor performers are still falling behind and whether good performers are getting closer to the world frontier, and 3) We assess econometrically the primary determinants of structural transformation, regarded as the key element to economic development.

This paper is related to a well-established literature on economic development on a cross country basis but rarely place the focus on Africa’s economy. Our work is related to Caselli (2014) who performed a development accounting exercise for the Latin

America. His study is a cross-section for a sample of Latin America’s economies whereas ours tracks a panel of 48 countries over the 1980-2010 period. Our work is also related to Caselli (2005) and Hall and Jones (1999) in using a similar conceptual framework to perform a development accounting exercise.

Our work is also related to Duarte and Restuccia (2006) who investigated the pattern of the world productivity distribution. Our work though has the merit to offer a higher degree of resolution on Africa. Our focus on Africa makes our work related to the recent work by McMillan and Rodrik (2014) and de Vries et al. (2014) who examined the pattern of structural transformation in some Sub-Saharan African economies. We complement this set of contribution by estimating an integrated econometric model on structural transformation and highlight its primary determinants.

The reminder of the paper is organized as follows: We begin with an outline of

(6)

the dynamic of labour productivity and the determinants of structural transformation. The results suggest that, while there are some good practices in Africa’s economic development best represented by Mauritius and Botswana, the bulk of Africa’s economies still struggles and offer a bleak prospect for the future.

II. A

NALYTICAL

F

RAMEWORK

A. MACROECONOMIC ANALYSIS:DEVELOPMENT ACCOUNTING

According to Caselli (2005), the best way to study the income structure across countries in detail, and account for differences in factors of production and productivity, is to develop an analytical model in which income is based on factors and efficiency. To further develop the concept it is necessary to start by measuring the Aggregate Production Function and its sources. To continue with the Development Accounting exercise, the labour productivity equation- a function of Total Factor Productivity, Capital Intensity and Human Capital-, will also be discussed.

 GDP per Capita

Following the Development Accounting example of the aggregate production function (Caselli, 2005), the GDP per Capita is mainly composed out of the GDP per Employee (or

Output per Worker, it will be used indistinctively), the employed share of population and

the working age population. The following function will facilitate the study of the

population and employment structures and will allow checking the effects of population on economic development. Equation 1 shows the components of the aggregate

production function.

𝑞

𝑡,𝐴𝑓𝑟𝑖𝑐𝑎

_{= 𝑦}

_{𝑥 𝑒}

_{𝑥 𝑛}

𝑡,𝐴𝑓𝑟𝑖𝑐𝑎 ₍₁₎

GDP per Capita (q): Y

N

, where Y stands for GDP and N for Population. Output per Worker

(y): 𝑌

E, where E is Employment. Employment structure (e):

E

𝑁15−64, where N15-64 is

population aged between 15 and 64 years old. Finally, 𝑁15−64

𝑁 represents the

(7)

The variable “n”, representing the demographic dividend, is referred to as the ‘working age population’ variable. This phenomenon is captured by the age distribution of the population across age ranges (age structure). The demographic dividend reflects the effect of the Structural Change process on the age structure of a country or region. It is characterized by an increase in fertility and mortality rates (due to enhanced life expectancy), resulting in a rise of the working age population. Thus, when the working age population increases, the production capacity does so as well and leads to economic growth (Bloom et. al, 2003). This concept will be further discussed in the analysis of the results

 Labour Productivity

The Development Accounting exercise continues by the analysis of the GDP per Employee and its sources. Equation 2 represents the Cobb-Douglas equation and it shows the earlier mentioned dissemination.

𝑦 = 𝐴

1−α1

𝑘

1

1−α

ℎ

(2)

Where A represents total factor productivity (TFP), k (K/E) accounts for capital and h

represents the average human capital. 𝛼is a constant for the labour-share parameter,

weighting the importance of Labour. Knowing the shape of the production function, we can then trace the determinants of labour productivity. With (1) and (2), we can express real income per capita and labour productivity in relative terms, with the U.S. as the benchmark economy: 𝑞𝐴 qUS= 𝑦𝐴 yUS 𝑥 eA eUS𝑥 nA nUS

(3)

𝑦𝐴 yUS= (𝐴𝐴)1−αA1 (𝐴𝑈𝑆₎1−αUS1 𝑥 (𝑘𝐴) 𝛼 1−αA (𝑘𝑈𝑆₎1−αUS𝛼 𝑥 ℎ𝐴 ℎ𝑢𝑠

(4)

B. SECTORIAL ANALYSIS:THE AFRICAN STRUCTURAL TRANSFORMATION

(8)

Structural Change as employment experiences a structural shift. According to Duarte and Restuccia (2010), Structural Allocation involves systematic Labour Productivity gains as employment shifts away from agriculture to more efficient sectors. However, the results obtained regarding the African Labour Productivity projects an interesting counterview on the Structural Transformation theory.

Taken from Nickell et. al (2008), the model specified below presents the precedent for the set of equations in which the value added per sector is the dependent variable, while cross-country differences in prices and labour productivity function as independent variables, and natural resources and country fixed effect act as control variables.

𝑆

_𝑗

= 𝛼

_0𝑗

+ ∑

𝑁

𝛼

_𝑗𝑘

𝑘=1

ln𝑝

𝑘

+ ∑

𝑁𝑘=1𝛼𝑗𝑘

ln𝜃

𝑘

+ ∑

𝑀𝑖=1𝛾_𝑗𝑖

ln𝑣

𝑖

(5)

where 𝑆_𝑗represents the share of value added in sector j in GDP; 𝛼0𝑗 the constant

parameter;

ln𝑝

is the prices logarithm;

ln𝜃

technology logarithm;

ln𝑣

factor

endowments logarithm; and finally, index per sector is represented by

j,k

(1…N) and

the index per factor i (1…M).

In this regard, and following the insights of Nickell, et. al.(2008) the model presents a long run equilibrium between the income shares per sector, prices, technology and factor endowments. This equilibrium is ultimately based on technology and prices, which are mainly influenced the overall consumer preferences and productivity. Therefore, the equilibrium in one sector is also dependent on the other sectors. That is why the model also includes a set of constraints in order to assure this long run

equilibrium.

For this study purposes, equation 4 have been adapted to the following system of equations (representing each sector), in which the variable Share of Value Added per sector is dependent on the explanatory variables Prices and Labour Productivity. Natural Resources and the country fixed-effect have been used as control variables.

𝑉𝐴𝐶_𝑐,𝑡𝑎𝑔𝑟𝑖𝑐𝑢𝑙𝑡𝑢𝑟𝑒 = 𝛽0+ 𝛽1𝑙𝑛(𝑃𝑅𝐼𝐶𝐸)𝑖,𝑡𝑠1+. . . + 𝛽5𝑙𝑛(𝑃𝑅𝐼𝐶𝐸)𝑖,𝑡𝑠𝑘 + 𝛽6𝑙𝑛(𝐿𝐴𝐵𝑃𝑅𝑂𝐷)𝑖,𝑡𝑆1+ ⋯ + 𝛽₁₀𝑙𝑛(𝐿𝐴𝐵𝑃𝑅𝑂𝐷)_𝑖,𝑡𝑆𝑘_{+ 𝛽}

(9)

𝑉𝐴𝐶_𝑐,𝑡𝑖𝑛𝑑𝑢𝑠𝑡𝑟𝑦= 𝛽₀+ 𝛽₁𝑙𝑛(𝑃𝑅𝐼𝐶𝐸)_𝑖,𝑡𝑠1_{+. . . + 𝛽} 5𝑙𝑛(𝑃𝑅𝐼𝐶𝐸)𝑖,𝑡𝑠𝑘 + 𝛽6𝑙𝑛(𝐿𝐴𝐵𝑃𝑅𝑂𝐷)𝑖,𝑡𝑆1+ ⋯ + 𝛽₁₀𝑙𝑛(𝐿𝐴𝐵𝑃𝑅𝑂𝐷)_𝑖,𝑡𝑆𝑘 + 𝛽_𝑛(𝑁𝑎𝑡𝑢𝑟𝑎𝑙 𝑅𝑒𝑠𝑜𝑢𝑟𝑐𝑒𝑠)_𝑖,𝑡 + 𝐷𝑢𝑚𝑚𝑦(𝐶𝑜𝑢𝑛𝑡𝑟𝑦)_𝑖,𝑡 𝑉𝐴𝐶_𝑐,𝑡𝑚𝑎𝑛𝑢𝑓𝑎𝑐𝑡 = 𝛽₀+ 𝛽₁𝑙𝑛(𝑃𝑅𝐼𝐶𝐸)_𝑖,𝑡𝑠1_{+. . . + 𝛽} 5𝑙𝑛(𝑃𝑅𝐼𝐶𝐸)𝑖,𝑡𝑠𝑘 + 𝛽6𝑙𝑛(𝐿𝐴𝐵𝑃𝑅𝑂𝐷)𝑖,𝑡𝑆1+ ⋯ + 𝛽₁₀𝑙𝑛(𝐿𝐴𝐵𝑃𝑅𝑂𝐷)_𝑖,𝑡𝑆𝑘_{+ 𝛽} 𝑛(𝑁𝑎𝑡𝑢𝑟𝑎𝑙 𝑅𝑒𝑠𝑜𝑢𝑟𝑐𝑒𝑠)𝑖,𝑡 + 𝐷𝑢𝑚𝑚𝑦(𝐶𝑜𝑢𝑛𝑡𝑟𝑦)𝑖,𝑡 𝑉𝐴𝐶_𝑐,𝑡𝑠𝑒𝑟𝑣𝑖𝑐𝑒𝑠 _{= 𝛽} 0+ 𝛽1𝑙𝑛(𝑃𝑅𝐼𝐶𝐸)𝑖,𝑡𝑠1+. . . + 𝛽5𝑙𝑛(𝑃𝑅𝐼𝐶𝐸)𝑖,𝑡𝑠𝑘 + 𝛽6𝑙𝑛(𝐿𝐴𝐵𝑃𝑅𝑂𝐷)𝑖,𝑡𝑆1+ ⋯ + 𝛽10𝑙𝑛(𝐿𝐴𝐵𝑃𝑅𝑂𝐷)𝑖,𝑡𝑆𝑘 + 𝛽𝑛(𝑁𝑎𝑡𝑢𝑟𝑎𝑙 𝑅𝑒𝑠𝑜𝑢𝑟𝑐𝑒𝑠)𝑖,𝑡 + 𝐷𝑢𝑚𝑚𝑦(𝐶𝑜𝑢𝑛𝑡𝑟𝑦)𝑖,𝑡

(6)

Regarding the constraints mentioned below it is important to specify them as follows: the first and implicit constraint is what can be considered as a ‘natural’ restriction, or in other words, the fact that by definition the summation of all the shares will sum 1. The implication extracted from this natural constraint is that in a system offering five sectors –equations-, the last one’s result will be redundant. Therefore, the last equation does not add any relevance.

(10)

III. Q

UANTITATIVE

A

NALYSIS

A. SOURCE DATA

The data used in this paper is retrieved from three main sources, in accordance with the research questions. The first and main database is the Penn World Tables (PWT8), the second one is mainly complementing the gaps from PWT8 in the development of the first and second research questions and it is the World Development Indicators dataset (WDI). The third and last data source is the 10-Sector Database from the Groningen Growth and Development Center (GGDC).

The main data source used in the study is provided by the Penn World Tables 8.0 (PWT8.0), which is the latest version available for these tables. The African economy consists of 53 countries with a rich variety of growth paths and diverse data

accessibility. This paper encompasses a comprehensive study consisting of a sample of 48 African economies dating from 1980 to 2010. It is important to highlight that Somalia, Libya, South Sudan, Eritrea and Algeria are not part of the study, as data for those economies was not provided in PWT8.0. However, this version of the PWT offers many improvements in the conceptualization and data collection, (e.g. Expenditure-side

real GDP at chained PPPs –RGDPe_{-, Expenditure-side real GDP at current PPPs –CGDP}e_-

Average Hours Worked, Human Capital, Employment, etc.). A good example is the

variable ` CGDPe_{´, the expenditure-side real GDP in current PPPs, that is, the GDP}

including the net exports at current prices. Most of the necessary variables from both equations are provided by the PWT8.0.

Regarding the variable `Working Age Population (N15-64)´ (population aged between 15-64 years old), it was incorporated in order to capture the demographic dividend effect -which entails a rise of fertility and mortality rates, boosting the working age population and thus economic growth-. However, this variable is not included in the PWT8.0. That is the main reason why some additional data has been gathered from the World Development Indicators (WDI), taking the ratio from the population (named: Population ages 15-64 (% of total)) and then obtaining the absolute value by using the

variable `Population´ from the PWT8.0. As stated in the conceptual framework, CGDPe_,

Employment, Population and Working Age Population, are the main variables for the

(11)

(CGDPe_{/Emp) the Share of Employment (Emp/N15-64) and the Demographic Dividend}

(N15-64/Pop).

Moving towards the second equation, the variables for the Labour Productivity equation have all been taken from PWT8.0. This equation is based on the dissemination of the GDP per Employee in three variables: Capital Stock, Human Capital and the Total Productivity Factor, which is the residual of the first two variables. As mentioned in the Conceptual Framework, the variable `Labour Share´ measures the weight of Human Capital and indirectly that of Capital Stock.

The variable Capital Stock presents a complete dataset for all the 48 countries. However, due to the conceptualization complexity of Human Capital (i.e. this variable requires information about education- years of schooling- which many African

economies do not hold) and Labour Share, many countries had no data available for those variables. In order to fix this issue, the ´Nearest Neighbor Method´ has been used.

In the case of the first variable, the average of CGDPe_{per Capita for each year has been}

calculated and then the missing data for Human Capital has been filled with the data

from those countries with a similar CGDPe_{per Capita average. Table 1 shows this}

substitution, the list on the left stands for the countries with missing data for Human Capital, and the list on the right stands for the countries which data has been used

according to the most comparable CGDPe_{per Capita average to fill the missing}

information of countries on the left. Labour Share shows a similar case; the average from the variable Employment for each country has been used in order to fill the lack of data with the data from the most comparable average Employment country (Table 2).

(12)

average) dataset from the WDI has been used. Nevertheless, the use of USD instead of LCU does not have effects on the results, as the dependent variable is expressed in shares, prices are indexes and Labour Productivity is expressed in growth (1=1970).

A. ANALYSIS OF THE SOURCE DATA

As mentioned above, the new version of the PWT8.0 used in this study offers an

improved set of variables, mainly featured by a reinterpretation of former variables and integration of theoretical concepts in new variables. Although the improvements

provided by PWT8.0 offer a new and enhanced dataset, there still remain some

drawbacks. Exemplary, four specific variables are described as follows; first, the variable GDP is now differentiated between ‘expenditure-side GDP’ which includes net exports, while ‘output-side GDP’ does not. This differentiation, as well as the new

reinterpretation of GDP, which is based on an inclusion of different PPPs benchmarks, allows for a closer approximation to reality.

Secondly, in their paper, Inklaar and Timmer (2013) discuss the theoretical bases of ‘Human Capital’ from PWT8.0. In the new contribution of “Human Capital” in the dataset, they rely on Caselli (2005) and Hall and Jones (1999) to create this new

variable, measured through years of schooling, which has made it possible to model and integrate this variable. This measure addresses new information considering education inputs; however, it does not account for differences in the quality of the education provided, since obtained cognitive skills are not considered. Therefore, although the contribution provides a useful framework to check on human capital quantitatively, it does not provide a clear picture of human capital quality.

Thirdly, the “Capital Stock” variable offers an insightful approach as it includes the capital investment as a measure of capital. Investments in R&D produce a systematic Total Factor Productivity (TFP) variance, which implies a systematic efficiency

enhancement from more developed nations. Nevertheless, it does not include data on land, nor natural resources, and knowledge-based assets (Inklaar and Timmer, 2013). This variation is highly relevant within the African context, as many countries are dependent on various natural resources (oil and mineral exporters) and land

(13)

African countries will always experience lower TFP shares, and although it may reflect a real trend, it still may not present a clear picture of reality or a possible evolution.

Finally, the variable “Employment” is also a new contribution of PWT8.0, and it is genuinely useful for the purpose of this paper, as this variable is needed in the Output per Capita function. The availability of the data; however, urges to reduce number of years studied in the sample, making it necessary to limit the sample from 1980 and onwards. Moreover, the data used in PWT8.0 is retrieved from international surveys, conducted by the International Labour Organization. Notwithstanding, this information may be biased in the African context as it is based on governmental statistics (informal economy, etc. will be further discussed below).

In regard to the analysis of the variables for the study of the African Structural Transformation, the variable “Prices” has been obtained by dividing “Value Added in current Prices” by the “Value Added in constant prices”. Prices in a particular sector will always be statistically significant on its own sector. The variable ‘Labour Productivity’ is based, firstly on the calculation of Value Added (constant prices) divided by

Employment, and then indexed, using 1970 as a benchmark. The values have been indexed in order to check on the changes and evolution of sector-country Labour Productivity over time and also to reduce volatility.

The control variables used are: ‘Natural Resources’ and the ‘Country Dummy’ variable (which controls for fixed effects). The variable ‘Natural Resources’ is the percentage from the GDP that is obtained from the sale of oil, natural gas, coal, mineral and forest rents, the data has been obtained from the WDI (‘Total natural resources rents (% of GDP)’). The choice for this variable in the study was because of its

meaningful insights in, particularly, the African context as natural Resources availability produces numerous effects on the development process of economies. In this paper, this variable is meant to capture the specific effects of natural resources in the African

economy.

(14)

fixed effects’ in the econometric specification. This variable also allows controlling for time-invariant errors of measurement for specific countries. This variable controls specific events or features in a particular country. Specifically, in this context it results of great use on controlling for the institutional divergences among countries. Easterly and Levine (1997) argue that ethnic cross-country differences in the continent are an

important driver for countries divergent policies and institutional framework evolution. By controlling for this country fixed-effect, the really fragmented and diverse ranges of institutional frameworks characterizing the African continent are accounted.

Regarding the variable ‘Value Added Share per sector’, it varies depending on the independent variables “Prices” and “Labour Productivity” growth or decline. The Share of Value Added per sector is constructed from the portion per sector divided by the total economy value added. As previously mentioned, the development theory assumes that reallocation among sectors will entail a Structural Transformation leading to economic development and growth. The model specified in section two aims to state the income trend by studying Labour Productivity and price movements. Considering the case of developing economies, it would be expected to see a declining importance of agricultural share followed by an increase in productivity and relative prices in industry, which would lead to a further labour reallocation towards services. However, the results offered by the African case may suggest something different. This fact will be further discussed in section five.

B. DATA LIMITATIONS

As a matter of fact, data availability for the African continent has been enhanced throughout the years. Data accessibility has been considerable improved over the past years, showing the growing importance and internationally participation of the African economy. Nevertheless, the quality of the data, as well as complete availability is still an issue for the African continent. Many reasons justify this statement; although data is more and more demanded internationally, data collection is not yet a main issue for African countries. Therefore, official statistics remain poor due to the lack of incentives to improve their data collection, as those countries are still facing many other

(15)

might be biased, as it won’t show the ‘informal activities’ which are not listed in the official registers. Most of those economies are based on weak institutional systems, often featured by poor regulation and corruption which raises the existence of ‘Black Market’ activities in their economies. This provokes limited data quality and accessibility which may bias the conclusions extracted from the empirical study. For instance, alternative datasets will offer different information depending on their source, highlighting the lack of consistency in the available data (Alkire, 2007). This issue is, however, beyond the scope of this paper.

C. TAXONOMY

The presented sample of this study aims to overcome limitations from previous literature on the African continent. As stated in the introduction of this paper, the sample embodies the entire range of African countries, rather than just the Sub Saharan Africa. As a result, the Development Accounting study will serve as a more useful tool in the task of explaining the different economic performance patterns. In order to offer a broader picture of these growth patterns the sample has been divided into 4 types based on the World Bank Classification, which uses the Gross National Income (GNI) data for the year 2014. However, as this study includes data from 1980 to 2010 it would not be consistent to take a classification based on 2014 data. That is why following the insight of the World Bank Classification, the average GDP per Capita from 1980-2010 has been used in order to structure the country distribution (Table 3): Low income ($1,035 or less), Lower Middle Income ($1,036 to $4,085), Upper Middle Income ($4,086 to

(16)

Table 3: World Bank Country Classification with GDP per Capita Average (1980-2010) Low Income

(<$1,035) Middle Income ($1,036-$4,085) Upper Income ($4,086-$12,615)

Burkina Faso Angola Botswana

Burundi Benin Gabon

Central African Republic Cameroon Mauritius

Chad Cape Verde South Africa

Congo, Dem. Rep. Comoros Tunisia

Ethiopia Congo, Republic of

Guinea-Bissau Cote d`Ivoire

Lesotho Djibouti

Liberia Egypt

Madagascar Equatorial Guinea

Malawi Gambia, The

Mali Ghana Mozambique Guinea Niger Kenya Rwanda Mauritania Tanzania Morocco Togo Namibia Uganda Nigeria

Sao Tome and Principe Senegal Sierra Leone Sudan Swaziland Zambia Zimbabwe

(17)

a country is highly affected by capital and labour investments (human capital and physical capital-infrastructures). In this regard, the Lucas Paradox (Lucas, 1990) states that the earlier mentioned weak institutional environment (being, scarce human and physical capital) causes low levels of Labour Productivity and results in disincentives for foreign investment. This issue will be further discussed in the analysis of the results.

The overall continent is characterized by a large range of Low and Middle Income countries. Often, these countries’ typologies are related to ample availability of natural resources, usually non-fuel or mineral resources, and this common feature carries some implications. Congo and Sierra Leone are good examples of countries which are rich in natural resources (non-fuel resources –e.g. diamonds), but at the same time lack economic development and strong social infrastructures, namely, policies and institutions framing their economic reality. This relationship between the access to mineral resources and poor economic and institutional development is based in an endless loop. Those poorly developed countries often become dependent on their natural resources and this dependency limits their institutional and economic development. As a result of their economic dependence, their economies are not diversified and highly vulnerable to external shocks. (Haglund, 2011). Also, as their institutions are poorly developed, they don’t allow for efficient investments in labour and capital.

(18)

IV. A

NALYSIS OF THE

R

ESULTS

A. DEVELOPMENTACCOUNTING:GDP PER CAPITA AND ITS SOURCES

Figure 1 in Appendix A shows the evolution of the GDP per Capita over the last three decades, as well as the trend of its sources. The exposition of the sources of GDP per Capita in the chart is of great interest in order to explain the trend. In 1980 it presented its highest peak relative to the United States (considered as the benchmark economy) with a 6.6% of GDP per Capita for the whole African aggregate. Meaning that, when the US GDP per Capita was 100$, the African was 6.6$. However, the African economy was not able to catch up with the growing US GDP per Capita over time, presenting a final ratio of 5.2%, which ultimately meant that the African economy got weaker over time. Figure 1 indicates how the indicator for Output per Worker, following the same trend as GDP per Capita, shows a decreasing trend over time. Regarding the indicator Output per Worker, the chart shows how an average worker relative to US in 2010 produced 2% less than the one in 1980. Nevertheless, the variables ‘working age population’

(`demographic dividend´ in the chart) and ‘employment structure’ (`employment rate´), move steadily over the period of time in the sample.

These constant indicators from figure 1, the ‘working age population’ and the ‘employment structure’, are explanatory for the low performance of Output per Worker and ultimately, GDP per Capita. One of the conditions needed to obtain sustained

(19)

In the Appendix A there are three charts disclosed representing the GDP per Capita for the different groups according to the World Bank Classification. Low Income countries present a similar trend as the whole African aggregate (figure 2). However, the values are lower and the indicators for GDP per Capita and Output per Worker have a more sustained decline, from a 3% in 1980 to 2% in 2010, relative to the United States. Although the employment rate for Low Income countries is higher than in the whole African aggregate, the demographic dividend remains steady over time. Thus, Low Income countries have substantially deteriorated their productivity and income levels per capita over time, following the continent’s trend, with a big decline from the eighties until the nineties and raising levels for the new millennium.

Middle income countries present a different pattern and are the drivers for a higher aggregate GDP per Capita and Output per Worker (figure 3) because this group includes most of the countries from the sample (25 out of 48). Figure 3 presents a meaningful chart; its GDP per Capita and Output per Worker from 1980 to 2010 barely vary. Although they also suffered during the eighties’ and nineties’ decline, the GDP per Capita and the Output per Worker are the ones decreasing the least, compared with the other groups. This differentiated trend of Low Income countries can be explained by the behavior of the demographic dividend, which experiences higher rates than illustrated in the previous chart. The case of Upper Income economies (figure 4) deserves special attention as the results are lower than expected. Looking at charts 3 and 4, in both cases the demographic dividend indicator grows over time and as mentioned before, it is one of the GDP per Capita growth drivers. Since such economies are more developed than the Low Income economies, they have been able to benefit to a greater extent from its demographic transition. At the same time, the expected improved institutional

environment has provided a better framework for the Output per Worker to rise. The combination of those improved indicators has resulted in a higher GDP per Capita. However, this fact is not properly reflected on Upper Income Output per Worker and consequently also not on GDP per Capita.

Table 10 shows the growth rates per decades for the GDP per Capita and GDP per Employee for the United States, Africa and the subgroups, and relative to US. In

(20)

B. DEVELOPMENT ACCOUNTING:TOTAL FACTOR PRODUCTIVITY

Bloom et. al (2007) suggest that there is a significantly positive relation between the working age population and economic growth, which increases when institutional quality is present. A great amount of literature highlights the important role of institutions in the development process; Hall and Jones (1999) took this approach in order to explore the negative effects of weak institutions on Output per Worker. They suggested that the allocation of physical and human capital, which is fundamental in measuring the efficiency gap, is driven by the quality of the institutional environment. As differences in the allocation of production factors produce deviations in the composition of the Efficiency measure (TPF), it is of great importance to specify what conditions affect their allocation. The authors suggest that Output per Worker tendency is highly affected by the institutional framework.

Therefore, the second part of the Development Accounting exercise focuses on the decomposition of the Output per Worker. By disseminating the GDP per Employee by its components (physical and human capital), as stated in the conceptual framework, the different efficiency gap patterns can be further explored, in other words the TFP. The charts 5, 6, 7 and 8 in Appendix A represent the Total Factor Productivity of the African economy and its subgroups relative to the US.

Figure 5 shows how in 1980 an average African worker produced approximately 11% of the output that a worker produced in the US. The African Labour Productivity grew rapidly until the mid-eighties, reaching a higher TFP relative to the US in 1984 with 13%. The trend followed after the peak was characterized by a pretty much steady decline with a minimum peak of 5% in 2000. In other words, in 2000 the Labour Productivity was half of the one in 1980. Finally, in 2010, the TFP level was 7%.

Therefore, in 1984 an average African worker was producing almost twice as much as in 2010.

The efficiency gap between the benchmark economy and the African economy is significantly large. The other three charts can serve as an explanation driving the results of figure 5, as it encloses the aggregation of the whole set of countries. It seems

(21)

income, the subgroup with more countries, shows some volatility over the sample but remarkably, the Total Productivity Factor remains at 9% both in 1980 and 2010. However, in the same way as Low Income, Upper Income countries have also worsen their Labour Productivity, with a more distinctive decline for the Upper league (figure 8), with 40% in 1980, decreasing to 15% in 2010. It appears clear how all follow a similar trend, with a characteristic slump in 2008, reflecting the economic world crisis and a general raising tendency from the New Millennium on, with still little but steady growth.

The set of graphs previously explained have a great deal to say on the role of production factors efficiency, more specifically on the Labour Productivity as an

important driver for growth theories. The choice for the data and methods used in order to obtain the TFP measure play a fundamental role in the application of development and growth theories. According to Klenow and Rodriguez (1997), endogenous growth models offer the opportunity to further check on productivity differences. They offer a wide set of growth theories with endogenous variables which allow for a better

understanding of the dispersion across countries and efficiency divergences.

Table 11 in Appendix A shows the values driving the TFP, which is a product of human capital endowments, capital endowments and GDP per Employee. As the figures reflect, the African TFP has declined steadily over the three decades and it has been mainly driven by the dismissing capital-labour ratios, most pronounced in the Upper Income league. In regard to the Labour Productivity and the efficiency gap, as TFP comprises Human Capital and Capital stock endowments, the allocation of those will be highly affected by the institutional context. Likewise, institutions’ quality will also be influenced by the economic structure of those economies.

(22)

dependent on institutions still remains poorly developed, with low TFP levels. Moreover, Bello, et. Al (2011) discussed the effects of inadequate policies on the misallocation across production factors, as a main driver for declining levels of TFP and capital accumulation.

C. AFRICAN PRODUCTIVITY:THE DYNAMICS OF THE AFRICAN LABOUR PRODUCTIVITY.

The following section encompasses a particularly comprehensive study on the African Output per Worker trends and facts over the past 30 years. This study is based on data on GDP and Employment relative to the five outperforming African countries. As previously mentioned, the World Bank Classification has been of great use in order to check on different macroeconomic trends and patterns of all countries within the sample by comparing the different types with the benchmark economy, the United States. This section focuses on the decomposition of Labour Productivity within the sample. Hence, it aims to observe the Output per Worker within the African economies individually in order to explore how much of the economic growth is due to Labour Productivity and employment.

Inspired by Duarte and Restuccia (2006) example, the study will be divided in four sections: Dispersion, disparity, mobility, miracles & disasters. Using an intra-continental approach, a comparison of the top five performing countries with the rest of the sample, helps to observe the particular drivers of the African Economy. This paper follows that example and it divides the five richest and poorest countries in the African continent based on their average GDP per Capita values from the 1980 to 2010. Therefore table 4 provides that classification, showing the big gap amongst the African economies. The highest GDP per Capita registered is that of Gabon with 8,520$ (constant), while the lowest is Liberia with 581$ (constant).

Table 4: 5 richest and 5 poorest African countries.

5 Richest GDP per Capita 5 Poorest GDP per Capita

Gabon $8,520 Congo, Dem. Rep. $414

Mauritius $8,425 Mozambique $439

Botswana $6,681 Burundi $493

South Africa $6,138 Ethiopia $560

(23)

0 2 4 6 8 10 12 14 16 %  Disparity

The first section, devoted to the topic of disparity, focuses on the Output per Worker divergence across countries in the sample. Figure 9 represents the average worker productivity. It suggests that an average worker from the five richest countries produces between 6 and 15 times more Output per Worker than those in the poorest countries. It can be observed how from the mid-eighties until the mid-nineties those Labour

Productivity differences increased rapidly to a maximum of 15 points recorded in 1994 which was a result of the so called lost decades. In only a few years after the

mid-nineties, this tendency dropped to the level of 12 points in 2000, and kept declining to a 9% in 2010, with some ups and downs.

Figure 9: Disparity on GDP per Employee across countries

 Dispersion

Regarding the second part, the study of the dispersion of the sample focuses on the evolution of GDP per Worker over time; the chart shows the differences in

(24)

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 19 81 19 83 19 85 19 87 19 89 19 91 19 93 19 95 19 97 19 99 20 01 20 03 20 05 20 07 20 09 1980=1 5 Poorest 5 Richest enhance their situation too much over the past three decades as the average worker in 2010 is producing just 4% more than back in 1980. Regarding the five richest countries, they have gradually bettered their situation since 1980, positioning themselves 0.36 points above the benchmarking year, which means that the average worker in 2010 produced 36% more than they did in 1980. Obviously, the 2008 global crisis had affected the Labour Productivity growth of both the richest as the poorest countries.

Figure 10: Dispersion on GDP per Employee over times

Figures 9 and 10 indicate that the differences between rich and poor countries steadily increased over time. It seems that over the past three decades the dispersion and diversity of Labour Productivity increased, making the richest countries richer and the poorest countries poorer. Table 5 is presented in order to study this affirmation in detail. It is based on the data of the entire sample; thus, 43 countries relative to the aggregate five richest African economies.

(25)

Table 5: Relative Output per Worker by Decile

From the 80s to the 90s the dispersion increases. Thus, the relatively more productive countries got even more productive while the poor countries got poorer or less productive. In other words, in the 1980’s the 10% poorest countries (in terms of GDP per Employee) had an average Labour Productivity of 6% relative to the 5 richest economies. This value decreased to 5% in the 1990’s. However, from the 90’s to the New Millenium this trend changed and the dispersion decreased, coming back to 6% Labour Productivity for the poorest countries. In this regard, the 10% most productive countries during the 80’s produced 65% of what the 5 richest countries produced and in the 90’s that ratio increased to 90%. Yet, following the trend, in the 90’s it decreased just to 89%. In conclusion, dispersion across countries increased until the beginning of 2000. Afterwards there was a turning point and it moved backwards, but to a smaller extent.

In the period between the 90s and the 00’s, both of the sample’s extremes deciles (D1, D2, D3, D9, and D10) experienced a decrease in its dispersion. It can be observed, however, how the intermediate deciles’ dispersion increased the most (D4, D5, D6, D7, D8). In this regard it can be concluded, considering the whole dataset, that the

differences across countries were getting larger but that after the turning point in the 1990’s the differences became smaller.

(26)

Table 6: Relative Output per Worker by Quintile

Following the same insight, table 6 shows the Output per Worker of the African countries divided by the quintiles; each one represents 20% of the sample in an

ascendant order. The first 20% poorest countries remained somehow stable over the

three decades, while quintiles Q2, Q3, Q4 and Q5;however, follow a rising trend. By

2010, the 20% richest countries were producing 96% of the Output per Worker relative to the 5 richest economies. Consequently the ratio Q5/Q1 grew over time. The take away of this table is that richer countries have been able to absorb much more Output per Worker, therefore becoming more efficient than poorer countries over the past decades.

 Mobility

The third section is devoted to the mobility of the 48 African economies from their initial position, 1980, to the final year of the sample. Figure 11 is relative to the 5 richest

African countries. The countries close to the 45º trend line (with an average growth of 0%) are the ones which had similar GDP levels per Employee compared to the 5 richest countries over those 30 years. Therefore, the ones moving towards the left side –above the trend line- are the ones which have enhanced their GDP per Employee over time. Good examples are Equatorial Guinea, Gabon and Egypt. On the right side - below the trend line-, are countries which have worsened their Output per Worker (for instance Liberia and Congo Dem. Rep.).

(27)

Figure 11: Mobility on GDP per Employee over times

Following the same insight from Figure 11, Table 7 presents a mobility matrix. Both Output per Worker in 1980 and 2010 have been divided over quartiles (Q1, Q2, Q3, Q4), and the information is in percentages. The first quartile in year 1980 shows how in year 2010 45% of the countries from Q1 did not move to another quartile, or in other words, of the 11 countries constituting Q1, 5 remained there. 18% (two out of eleven) moved to Q2, 27% (three out of eleven) moved to Q3 and 9% (one country) moved to Q4. Countries with the lowest Relative Output per Worker (from quartiles Q1 andQ2) are the ones with the highest mobility throughout quartiles, while countries in quartiles Q3 and Q4 have not experienced such a high mobility. In other words, half of the poorest (relative to Output per Worker) countries have enhanced their situation, while the other half have remained the same or worsened their situation. A good example are the

countries from Q2, 50% of them have moved to Q1 while the other 20% stayed in Q2 and the remaining 30% have moved towards Q3 and Q4. Concerning the richest ones, most of them have kept their position, as can be seen in Q4, where 73% (eight out of eleven countries) have remained in Q5 and only three of them, 27%, have moved to Q3.

Angola Benin Burkina Faso Botswana Burundi Cameroon Cape Verde Central African Republic Chad Comoros

Congo, Dem. Rep.

Congo, Republic of Cote d`Ivoire Djibouti Egypt Equatorial Guinea Ethiopia Gabon Gambia, The Ghana Guinea Guinea-Bissau Kenya Lesotho Liberia Madagascar Malawi Mali Mauritania Mauritius Morocco Mozambique Namibia Niger Nigeria Rwanda

Sao Tome and Principe Senegal Sierra Leone South Africa Sudan Swaziland Tanzania Togo Tunisia Uganda Zambia Zimbabwe Rel at ive O u tp u t x Wor ker (201 0)

Relative Output xWorker (1980)

(28)

Table 7: Mobility Matrix- Relative Output per Worker 2010 ≤ $1,987-Q1 ≤ $2,846-Q2 ≤ $5,073-Q3 ≤ $11,446-Q4 Total 1 9 8 0 ≤ $1,987 (Q1) 45 18 27 9 100 ≤ $2,846 (Q2) 50 20 20 10 100 ≤ $5,073 (Q3) 0 27 36 36 100 ≤ $11,446 (Q4) 0 0 27 73 100

Figure 12 shows the big differences within the African economies regarding growth rates in Output per Worker. Annualized growth in Output per Worker is an indicator of the countries’ performance from 1980 to 2010. In this regard, the

annualized growth for the aggregate of the five richest African countries was -0.014% which suggests that many other African countries may also have rather low annualized growth rates. The figure shows how some countries like Equatorial Guinea, Egypt, Botswana and Cape Verde offer high annualized growth rates, reflecting an improved situation from 1980 to 2010. Most of the countries showing positive annualized growth are the ones which have shifted from the Low Income group to the Middle Income league in the recent years. Nevertheless, there are also numerous countries with

(29)

Figure 12: Relative Growth in Output per Worker

 Miracles and disasters.

The five richest and most productive countries in the African economy have an average Output per Worker growth of -0.014%. Therefore, a country has been considered to be experiencing a miracle (disaster) when it experiences an Output per Worker growth (decline) higher (lower) than 0%, therefore in positive rates, for a period of ten consecutive years. As to be expected, some of the five economies representing the benchmark of this sample show reasonably high growth rates in their output per

employee. Table 8 presents the 4 most remarkable countries on the basis of their labour productivity growth rates, the year it started, the number of years it lasted and the Output per Worker values at the beginning and at the end of the period of growth. Botswana is the country presenting the most exceptional performance over the other African economies.

Table 8: Miracles- Relative to five riches African Economies. Angola Benin Botswana Burkina Faso Burundi Cameroon Cape Verde Central African Republic Chad Comoros Congo, Dem. Rep.

Congo, Republic of Cote d`Ivoire

Djibouti Egypt

Equatorial Guinea

Ethiopia _{Gambia, The}Ghana Gabon

Guinea Guinea-Bissau _Kenya Lesotho Liberia Madagascar Malawi Mali

Mauritania Morocco Mauritius Mozambique

Namibia Niger

Nigeria Rwanda Sao Tome and

Principe Senegal

Sierra Leone South Africa

Sudan Swaziland Tanzania Togo Tunisia Uganda Zambia Zimbabwe -6% -4% -2% 0% 2% 4% 6% 8% 10% -1.6 -1.4 -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 Ann u al ized G row th in O u tp u t x Wor ker (%)

Relative Output x Worker 1980 (log) Growth in Output per Worker (1980-2010)

(30)

On the other hand, table 9 shows the 4 worst performing countries on the basis of their labour productivity growth rates. Same as table 7, the year when it started, for how many years did it last and the Output per Worker values at the beginning and at the end of the period of growth. Coming to similar results as Duarte and Restuccia (2006), the Democratic Republic of Congo is the country with the worst performance, with an annualized growth of -4% over the period of decline. Although South Africa’s decline takes more years, the decline of the Output per Worker growth rate is not as remarkable as in the other examples.

Table 9: Disasters- Relative to five riches African Economies.

It is important to mention that the general trend within the countries in the sample showed really volatile, unsteady growth rates. As aforementioned, the exceptional case of the 5 richest countries is what makes them the most productive. Although, they generally experienced positive trend in Output per Worker, their growth rates are not as steady as needed for further development. This assumption will be further discussed below by the descriptive analysis of some particular countries. As shown before in this paper, resource dependency plays a major role in driving countries’ performances. There is a wide range of literature addressing the issue of the so-called `resource curse´, the `Dutch disease´ and its drivers. The term `resource curse´ is commonly defined as the “tendency of natural resource abundant countries to suffer from low economic growth and disappointing development outcomes” (McSherry, 2006) and it involves market volatility and discouraging private investment, which along with the often unstable institutions and governments, produces counter development effects. On the other hand, the Dutch disease stands for the negative relationship between the growth of sectors with abundant natural resources and a consequent decline in other sectors. This is mostly due to the unidirectional shift of factors and efforts made in a resource abundant sector. Consequently, those sectors DISASTERS Annualized

Growth Start Year Number of Years Start Relative Y/L End

Congo Dem Rep -3.72% 1989 13 0.10 0.03

Guinea -2.61% 1996 11 0.25 0.11

Kenya -1.38% 1994 11 0.24 0.16

(31)

become more volatile and exposed to external shocks (Corden, 1984). Both, the `Resource Curse’ and the `Dutch Disease´ approaches claim that the lack of linkage between the resources abundant sectors and the other sectors of the economy was a main driver of underdevelopment and resource dependency.

Sala-i-Martin and Subramanian (2003) took the particular Nigerian case focusing more on the weak institutional environment of the country rather than the Dutch

disease. They conclude that resource-led countries follow a common pattern; the

inability to link the income through different economic sectors and therefore, to enhance their institutional context. Following Hecksher-Ohlin model based on trade theories, countries should focus on exporting the production of commodities based on their abundant factors rather than the export of the row factor. That way, countries would be able obtaining the linkage between economic sectors. The following section focuses this approach by studying the Structural Allocation of economic sectors within the African economy.

V. S

TRUCTURAL

C

HANGE

Many literatures focus on the development theories and the processes of Structural Change leading to catch up and stagnation cases. As stated in the introduction, such theories converge with the idea that labour reallocation entails Structural Change, a central step in the development race. The employment shift from agriculture to industry explains most of the development experiences, being the main stage of economic

development. Nevertheless, structural transformation continues by shifting towards the service sector. Duarte (2011)

Restuccia and Duarte (2009) studied the effect of sectorial differences in labor productivity on Structural Change; they analyzed the role of productivity in labour reallocation and concluded that the major productivity gaps are observed in agriculture and services. Consistent with many other studies, they conclude that the industry catch-up explains most of the productivity improvements and that this lack of sectorial

(32)

productivity by reallocating in the industry sector and slowing down their growth rates when moving towards services. In other words, labour reallocation towards the

industrial sector remains a main instrument in order to achieve the capability to absorb growth and further development in modern economies.

Figure 13 offers a first insight on the data offered by the 10-Sector Dataset, it shows how in the early days, from the seventies to the nineties, agriculture and services where the main drivers of the African economy. The second half of the graph shows the evolution of the sector allocation within two decades, where industry gained

importance, but the main driver kept being service market sector, which experienced the highest growth within the whole economy. The values to creat charts from figure 13 are shown in Appendix B.

The econometric analysis proposed below aims to study the African sectorial productivity patterns in order to set a stage of structural transformations in the continent, which, as broadly discussed, presents a different path. The study places productivity and prices as drivers of the Value Added (per sector) behavior in order to observe the particular African case in the structural transformation literature. In this study, Value Added is used as an income measure per country.

Figure 13: Value Added per Economic Sector 1970-1900 vs 1990-2010

Agricult ure Industry Manufac turing Market services Non-Market services

% GDP Value Added by Economic Sector (1970-1990) Agricult ure Industry Manufac turing Market services Non-Market services

(33)

A. ECONOMETRIC ANALYSIS

The use of the Seemingly Unrelated Regression offers a combination of equations as shown in the conceptual model. This analysis offers a meaningful insight in order to check the overall contribution of Prices and Labour Productivity from the sectorial reallocation by offering joint estimates from diverse regression models. What these combined regressions have in common (and therefore are added up together) is the possibility of their error terms being correlated within their dependent variables. In other words, the SUREG model allows the error terms of the different regressions to be correlated by regressing them in the same model.

At a first insight, the main independent variables in section three could involve some collinearity problems. That is because, by definition, changes in productivity are reflected in price changes. When a certain sector is influenced by an external shock (e.g. loss of arable land) or by internal causes (e.g. efficiency improvement), production is affected and therefore this change is reflected in the prices behavior. Because Labour Productivity can also be considered as a measure of human capital efficiency, once prices are already controlled in the regression, the inclusion of Labour Productivity could not offer much new information.

However, although by its intrinsic definition they should be somehow collinear (Prices = Value Added Current/Constant, Labour Productivity = Value Added Current/ Employment), the fact that Labour Productivity has been build, by indexing the data (1 = 1970) to reduce volatility in the variable, it does not show a collinear relationship

among the explanatory variables. After conducting the Breusch-Pagan test, it is

concluded that the hypothesis of correlation 0 cannot be rejected, the variables do not appear to be correlated, and therefore both have been included in the regression.

(34)

Table 15 in Appendix B presents the summary statistics of the sample and Table 16 shows the results obtained with the SUREG method (Appendix B). As expected, the variables in each sector behave accordingly; their own prices and Labour Productivity per sector appear to be statistically significant at a 1% confidence level. However, it is important to highlight as well that industry prices and labour productivity appear highly significant in each sector. It is also for the industry sector that R-square, with a value of 0.5, appears to be higher than in the other sectors. This means that the model explains the half of the variability of the values for the industry Value Added share. Moreover, the industry sector also presents a higher chi2, representing a stronger relationship among its variables. This evidence may be in accordance with the previous literature on the topic of the structural transformation, as the industrial sector appears to be a main driver for growth and development.

Nevertheless, it can be observed how the coefficient for labour productivity is much higher in agriculture and services, explaining a 13% of the income share in both sectors, while 3% and 4% in industry and manufacturing, respectively. This result is similar to the particular case of African Structural Change and its related literatures. It shows how both agriculture and services are driving income growth with higher rates. In any case, productivity per sector contributes positively to the dependent variable growth, while productivity of the other sectors appears to have a negative effect. The variable ‘prices’ follows just the opposite behavior than productivity. The contribution of ‘prices’ in the dependent variable per sector appears to be negative in its own sector; therefore, increases in relative prices per sector will result in the decline of that specific value added share. However, increases in relative prices for other sectors will contribute positively to the value added growth.

VI. C

ONCLUSIONS AND

I

MPLICATIONS

(35)

with an initial remarkable decline, from the eighties until mid-nineties and is followed by a lower, overall improvement a few years after the New Millennium.

Therefore, in accordance with the African historical context, data evidence shows how during the eighties and nineties the continent suffered from underdevelopment, low capital accumulation rates and low TFP, driving low labour productivity rates. However, as shown throughout this study, in the last decade since the New Millennium, their situation has been enhanced in contrast with the previous tendency. Yet, this tendency is not highly significant as the data is still showing little growth.

This paper also presents empirical evidence on the African structural allocation of

sectors. It shows how the economic drivers of the countries in the sample are agriculture and services, with little room for the industry and manufacturing sectors. Main part of the poor African development can be therefore related to the lack of industrialization. This fact has been preventing the continent from further development by impeding physical and human capital accumulation, which in sum have generated lower TFP ratios and therefore, declining economic development.

The conclusions reached by this study bring some implications. As discussed, the capital-labour ratios lower the TFP; however, there is little evidence that explains the cause of it. Many authors agree on the fact that the historical context of the continent plays a major role on its performance. Particularly, Acemoglu et. al (2001) agrees that the way the institutional framework of the continent has been settled, greatly affected the continent’s declining trend. However, there are other authors discussing the role of the capital accumulation in explaining institutional frameworks, rather than the other way around. (Glaeser et. Al, 2004). The authors claim that human capital drive

institutional improvement which ultimately stimulates economic growth. However, Kohler (2010), Schularick, and Steger (2008), Alfaro et. Al (2008) and many others pointed out the effects of institutional quality on the capital-labour flows across countries, concluding that the quality of institutions is a main driver for economic growth. There is an interesting discussion on the question of “What came first, the chicken or the egg?” Capital accumulation or an institutional favorable environment? A great deal of literature has extended Lucas’ Paradox (1990) theory in order to

(36)

trends of the African labour productivity which is very useful for identifying the drivers of its efficiency gaps, but further research should focus on the origins of these gaps.

The main limitation of this paper concerns the data availability of African

countries. It appears to be one of the main issues hampering further research regarding the countries’ trends and performances. In this particular study, the data availability just allowed for the study of thirteen African countries. In the same direction and still

regarding to the Structural Allocation, the lack of data on Education per sector has been a constraint on the study of the Structural Allocation as a measure of ‘factor

endowments’. The variable “human capital” can be considered a measure for education, it usually addresses new information considering education inputs; however, it often does not account for differences in the quality of the education provided, as it does not consider the cognitive skills obtained. Therefore, although it provides a useful

framework to check on human capital quantitatively, it does not provide a clear picture of human capital quality. According to Caselli (2014), differences in cognitive skills, rather than years of schooling, account for the larger productivity variation. When this topic is considered from the African context point of view, it needs to be further

discussed. PISA scores is the worldwide index for cognitive skills, thus they are a useful tool to check on human capital quality. Nevertheless, the availability of the data just covers a few African countries. There are alternative databases to check on students’ performances, but again, they are only available for a small sample of countries –not accidentally, mostly the ones with higher growth rates-, which does not provide a clear picture of the human capital value. (Harchaoui and Üngor, 2015). Hence, further

(37)

VII. R

EFERENCES

Acemoglu, D., Johnson, S., Robinson, J., 2001. The colonial origins of comparative development: an empirical investigation. American Economic Review, 91, 1369-1401. Alfaro, Laura, Sebnem Kalemli-Ozcan, and Vadym Volosovych. "Why doesn't capital flow from rich to poor countries? An empirical investigation." The Review of Economics and Statistics 90.2 (2008): 347-368.

Alkire, S. (2007). The missing dimensions of poverty data: introduction to the special issue. Oxford development studies, 35(4), 347-359.

Bello, Omar D., Juan S. Blyde, and Diego Restuccia. "Venezuela's growth experience." Latin american Journal of economics 48.2 (2011): 199-226

Bloom, D. E., Canning, D., Fink, G., & Finlay, J. (2007). Realizing the demographic

dividend: Is Africa any different. Harvard Program on the Global Demography of Aging, Working Paper, (23.2007).

Bloom, David, David Canning, and Jaypee Sevilla. The demographic dividend: A new perspective on the economic consequences of population change. Rand Corporation, 2003.

Caselli, Francesco, and James Feyrer. “The marginal product of capital” No. w11551. National Bureau of Economic Research, 2005.

Caselli, Francesco. "Accounting for cross-country income differences. “Handbook of economic growth 1 (2005): 679-741.

Caselli, Francesco. “The Latin American Efficiency Gap”. No. 1421. (2014.)

Corden, Warner Max. "Booming sector and Dutch disease economics: survey and consolidation." oxford economic Papers (1984): 359-380.

(38)

Duarte, Margarida, and Diego Restuccia. "The role of the Structural Transformation in aggregate productivity." The Quarterly Journal of Economics 125.1 (2010): 129-173 Easterly, W., & Levine, R. (1997). Africa's growth tragedy: policies and ethnic

divisions. The Quarterly Journal of Economics, 1203-1250.

Feenstra, Robert C., Robert Inklaar, and Marcel Timmer. The next generation of the Penn World Table. No. w19255. National Bureau of Economic Research, 2013.

Haglund, Dan “Blessing or curse? The rise of mineral dependence among low- and middle-income countries,” Oxford Policy Management, December, 2011

Hall, Robert E., and Charles I. Jones. Why do some countries produce so much more Output per Worker than others?. No. w6564. National bureau of economic research, 1999.

Harchaoui, T.M and Üngor M. “The lion on the move towards the world frontier: Catching up or remaining stuck?” (2015)

Hsieh, Chang-Tai, and Peter J. Klenow. "Development accounting." American Economic Journal: Macroeconomics 2.1 (2010): 207-223.

Hsieh, Chang-Tai, and Peter J. Klenow. Misallocation and manufacturing TFP in China and India. No. w13290. National Bureau of Economic Research, 2007

Inklaar, Robert, and Marcel Timmer. "Capital, Labor and TFP in PWT8. 0."University of Groningen (unpublished) (2013).

Kohler, Philippe, and Groupe Economie Mondiale SciencesPo. "Foreign direct investment in countries with weak institutions." (2010).

“Lions on the move: The progress and potential of African economies”. McKinsey Global Institute, 2010.

(39)

Nickell, S., Redding, S., & Swaffield, J. (2004). The uneven pace of deindustrialization in the OECD. London School of Economics mimeo.

Restuccia, Diego. "Recent developments in economic growth." Economic Quarterly 97.3 (2011): 329-57.

Rodrik, Dani. "AN AFRICAN GROWTH MIRACLE?."(2014).

Sala-i-Martin, Xavier, and Arvind Subramanian. Addressing the natural resource curse: An illustration from Nigeria. No. w9804. National Bureau of Economic Research, 2003. Schularick, Moritz, and Thomas M. Steger. The Lucas Paradox and the quality of

institutions: then and now. No. 2008/3. Diskussionsbeiträge des Fachbereichs Wirtschaftswissenschaft der Freien Universität Berlin, 2008.

VIII. A

PPENDIX

A:

D

EVELOPMENT

A

CCOUNTING AND

P

RODUCTIVITY

D

YNAMICS

(R.Q.

1 &

2)

 Table 1: Human Capital data compilation (GDP per Capita in $)

List of countries which substituted the missing information for Human Capital Country with missing hc GDP/POP Substituting countries GDP/POP

Angola 2,171 Cote d`Ivoire 2,008

Burkina Faso 671 Niger 670

Cape Verde 1,752 Congo, Republic of 1,805

Chad 993 Tanzania 1,013

Comoros 1,138 Benin 1,131

Djibouti 2,749 Morocco 3,019

Equatorial Guinea 2,938 Morocco 3.019

Ethiopia 556 Burundi 502

Guinea 1,588 Senegal 1,585

Guinea-Bissau 912 Rwanda 900

Madagascar 917 Rwanda 900

Nigeria 1,187 Gambia, The 1,166

(40)

 Table2: Labour Share data compilation (Employment in millions)

List of countries which substituted the missing information for Labour Share Country with missing labsh EMP Substituting countries EMP

Angola 5.11 Burundi 5.01

Cape Verde 0.15 Djibouti 0.18

Comoros 0.16 Djibouti 0.18

Congo, Dem. Rep. 13.98 South Africa 11.27

Congo, Republic of 1.10 Botswana 0.69

Equatorial Guinea 0.24 Swaziland 0.28

Ethiopia 2276 Egypt 17.35

Gambia, The 0.48 Mauritius 0.48

Ghana 6.31 Mozambique 7.42 Guinea 2.84 Senegal 2.89 Guinea-Bissau 0.46 Mauritius 0.48 Liberia 0.86 Botswana 0.69 Madagascar 5.67 Cameroon 5.01 Malawi 3.85 Zimbabwe 4.39 Mali 2.84 Senegal 2.89 Sudan 5.45 Cameroon 5.01 Uganda 8.01 Mozambique 7.42 Zambia 3.40 Rwanda 3.45

 Table 10: Average Annual Growth Rates - Per type & Relative to US (%)

Average Annual Growth Rates - Per type & Relative to US (%)

Development Accounting for the African Economy