One Island –Two Divergent Routes
A story of the different development paths of the Dominican Republic and the Republic of Haiti
Doctoraal Scriptie
Kuno W. R. Fischer Student No. 131 62 06
10th October 2005
Supervisor Dr. Marcel P. Timmer Co-assessor Dr. ir. Dirk J. Bezemer
University of Groningen Faculty of Economics
Department of International Economics and Business
Table of Contents
1. Introduction page 3
2. Literature Review 5
2.1. Theoretical Background 5
2.2. Trade and Economic Growth 7
2.3. Institutions and Economic Growth 8
3. Methodology 10
3.1. Out-of-Sample-Predictions 10
3.2. The Model 11
3.3. Possible Limitations of the Paper, Model and Research Method 15 3.3.1. Is it just an African Model? 15
3.3.2. The Problem with the constant 15
3.3.3. Is the model just valid for one period of time? 16
3.4. Data Collection 16
4. Discussion of Results 21
5. Sensitivity Analysis 25
6. Concluding Remarks 30
Bibliography 33
1. Introduction
Both Caribbean States, the Dominican Republic and the Republic of Haiti, are located on the island Hispaniola. Surprisingly, both countries differ in their economic achievement. The Dominican Republic has experienced strong economic growth in its history, while the Republic of Haiti has stayed one of the poorest countries in the world. Its performance had been stagnating over decades and has started to decline recently. The following graph 1 illustrates these divergent trends.
Graph 1 GDP per capita development of the Dominican Republic and Haiti
GDP per capita (constant 1995 US$)
0 500 1000 1500 2000 2500
1960 1966
1972 1978
1984 1990
1996 2002 Year
GDP per capita in constant 1995 US$
Dominican Republic Republic of Haiti
Source: World Development Indicators 2004 by the World Bank Group
Due to the fact that both states are on the same island, they have many similarities but they also differ in some important aspects that might explain their different development.
Economic growth theory suggests many possible factors that may have an impact on economic growth. The father of the neo-classical growth theory, Robert Solow (1956)1, built a model in which a country’s saving rate is a crucial determinant for economic growth. More recent studies investigate the relationship between different concepts and economic development by statistical regression models in order to go beyond the scope of the traditional proximates of growth. Such studies were done by Barro (1991) and Mankiw et al. (1992).
They found out that educational attainment, political stability and growth of population influence economic growth. Today more sophisticated and field-related regression studies are
1 See for a detailed description Debraj Ray (1998), “Development Economics”, chapters 3 and 4
present. Acemoglu et al. (2001) suggest that good institutions lead to economic growth.
Others like Dollar (1992) explain that liberal and open trade policies determine the success of an economy, while in the eyes of Sachs (1997) and Gallup et al. (1999) a country’s success depends on geographical factors like access to the sea and climate.
As it was shown so far, there are many econometric studies that support any and all of the different concepts that influence economic growth. A recent study conducted by Rodrik, Subramanian and Trebbi (2002) attempts to combine the three deeper determinants of growth:
openness, geography and institutions. It is looked for their interactions with each other and their direct and indirect effect on incomes. Furthermore, the authors come up with one of the deeper determinants that trumps the other two. According to Rodrik et al. the most important driver of economic growth seems to be institutional change.
In this paper an analysis method called out-of-sample-prediction will be used in order to find out the different growth paths of the Dominican Republic and the Republic of Haiti. This means that the estimated coefficients of an already existing regression model will be used in combination with data for the Dominican Republic and the Republic of Haiti to explain their per capita economic growth rates. The underlying model stems from Sachs and Warner (1997a) and will be applied to two time periods: firstly between 1965 and 1990 and secondly from 1990 to 2002. This analysis shows how the impacts of economic growth drivers have changed with time and which ones explain the divergent routes of the Dominican Republic and Haiti.
The paper is organised as follows. In the next section a general topic-related literature review will be presented. This is followed by the Methodology part of this research which describes the underlying model of this paper, the technique out-of-sample-prediction, the weaknesses of this theoretical framework and the data collection method. In the following section the results of the model will be discussed extensively, i.e. it will be pointed out what the major driving forces are. The next part deals with a sensitivity analysis, which shows whether the predicted outcomes are sensitive towards changes in variables. Additionally, the predictive power of the model will be tested. Finally, the paper ends with a conclusion and remarks for future research.
2. Literature Review
The introduction of this paper has shown that most literature on economic growth comes up with three deeper sources of growth. These sources are institutions, integration in world trade and geography. Therefore, the literature review will elaborate these concepts more deeply, whereas it should be noted that the concept geography will not be discussed. The concept geography does not give further insight for this study because both sates are located on the same island. Therefore both countries face identical geographic conditions. As a result, only integration and institutions will be discussed more extensively because it is very probable that they explain most of the diverging routes. But before that a basic theoretical background will be presented.
2.1. Theoretical Background
The basic underlying growth model of all cross-country regression models is from Solow (1956). This model makes use of a Cobb-Douglas Function, which determines output by means of capital, labour and technology and their respective elasticity towards labour and capital. The basic assumption of Solow’s model is the diminishing returns to capital. The crucial starting point is the saving rate in the model. Savings lead to investment, and hence to capital accumulation. This in turn increases labour productivity because there is more capital available for workers. But of course, as said before, the labour productivity will not increase at a constant level, the growth diminishes until it stops and becomes zero. Then the steady state or “frontier” is reached. At the frontier, output grows at the pace of the total population growth rate. The only element that can increase per capita economic growth in the long run is technology. Here a (positive) change shifts the production function upwards and creates per capita growth.
Further Solow predicted that due to the fact that poor countries have less capital, they must have a higher marginal productivity of capital than more developed countries. Therefore, growth in labour productivity will be much faster, i.e. there is a catch-up. This is supported by the assumption that technology is spread on a world-wide level without any losses. Thus, one
could conclude that in a long-term view rich and poor countries converge in their steady states. This phenomenon is called Unconditional Convergence.
Nevertheless, the past has shown that poor and rich countries do not converge on a world- wide level, so Unconditional Convergence does not hold. This finding may be due to differences in total population growth rates and saving rates. Thus, if these initial conditions are different, then countries converge in their economic growth rates but not in per capita incomes. When countries are grouped with comparable counterparts, one can recognise that poorer countries tend to grow faster and that the countries within the group converge, while there is a world-wide divergence. This concept is called Conditional Convergence.
Conditional Convergence is a weaker but more realistic form of Unconditional Convergence.
This also provides the basis for cross-country studies.
First steps of cross-country regressions were taken in Robert J. Barro’s paper “Economic Growth in A Cross Section Of Countries” (1991) and Mankiw et al.’s work “A Contribution to the Empirics of Economic Growth” (1992). Barro found out that poor countries can catch- up, if they have a high level of educational achievement. Furthermore his findings were positive relationships between per capita economic growth and physical investment, school enrolment and political stability. But Barro also found negative relationships. Thus, per capita economic growth was negatively correlated to the ratio of Government Consumption Expenditure to GDP and to price distortions. Mankiw et al.’s paper also proved that differences in saving, education and population growth explain differences in economic achievement across countries.
As already stated before, current literature focuses on the three deeper determinants of economic growth. In the next section two of them, namely institutions and integration in world trade, will be analysed more extensively.
2.2. Integration in World Trade and Economic Growth
It is a highly debated topic, whether open and liberal trade regimes foster economic growth.
Often it is argued that more open countries tend to grow faster. Classical explanations of unrestricted trade and growth are rooted in the principles of comparative advantage. The theory of Ricardo from the 19th century was further improved by Hecksher-Ohlin (1933)2. Both models predict that an economy will be successful – in terms of an increase in welfare – when it produces that product in which the country is factor-intensive and trades with another country which is highly endowed in another factor. Doing this raises welfare in both countries. More recent papers on this topic use cross-country regressions to show that low trade barriers, i.e. openness, will lead to more growth. A pioneer in this field is the author David Dollar with his paper “Outward-oriented Developing Economies Really Do Grow More Rapidly: Evidence from 95 LDCs, 1976-1985” (1992). Dollar created a regression model of 95 countries. He finds out that differences in countries’ GDP growth rates are partly due to exchange rate management and a country’s trade regime. The overall conclusion is that trade liberalisation, devaluation of the real exchange rate and continuous stable real exchange rates are factors which improve economic growth conditions in many LDCs. In the paper
“Trade Policy, Exchanges Rates and Growth” written by Sebastian Edwards (1993) one finds the similar relationship between trade policy and growth like in Dollar’s paper, meaning that more open countries experience a faster growth than countries with trade barriers. In “Trade and Growth in East Asian Countries: Cause and Effect?” by Frankel, Romer and Cyrus (1996) the authors investigate the rapid growth path of the East Asian economies. Their conclusion is that openness explains a large amount of economic achievement in Singapore and Hong Kong and smaller amounts in Korea, Taiwan and Malaysia.
There are also weak papers like David Dollar and Aart Kraay’s “Growth is good for the Poor”
(2000), in which both find a statistical insignificant openness parameter and then argue that openness is the crucial element for poverty reduction. This creates many doubts. Additional doubt is created by the paper “Trade Policy and Economic Growth: A Sceptic’s Guide to the Cross-National Evidence” (2001) by Francisco Rodriguez and Dani Rodrik. Here both criticize various earlier studies on openness-growth relationship. The basis for this is that they
2 See Husted & Melvin (2001) International Economics, 5th edition, Chapters 3 and 4
find statistical insignificance in earlier models, when they add outlier-countries or add other variables.
2.3. Institutions and Economic Growth
There are several studies that investigate the relationship between institutions and economic growth.
Also the concept institutions was discussed in the past. In Max Weber’s view (1922)3 bureaucratisation was the key to an efficient form of an organisation in a macro and micro environment, referring to the government level and business level. Weber was convinced that
“good institutions”, i.e. a stable environment in terms of efficient organisations, are a necessary condition for economic achievement
Further insight was achieved by North and Thomas (1973). Both authors stressed the finding that institutions are not a result of economic growth but a crucial determinant of economic growth. In their eyes efficient institutions lead to growth, for instance well-defined property rights create more innovation and risky investments that result in economic development.
Institutions can be defined as EFW which was done by Gwartney and Lawson (2001) in their article “The Concept and Measurement of Economic Freedom”. The authors define the highly debated Economic Freedom of the World (EFW) as an instrument for measuring the level of a country’s institutions. It is a mixture of personal choice, voluntary exchange and personal and property protection. After this definition of institutions (EFW), the authors conclude that weak forms of institutions are commonly present in less developed countries, i.e. in a broader sense also Gwartney and Lawson emphasize that good institutions lead to growth.
A very popular and recent study by Acemoglu, Johnson and Robinson (2001) confirms the same institutions-growth-relation as Gwartney and Lawson. Acemoglu et al. proxy institutional quality by colonial settler mortality. In the point-of-view of Acemoglu et al. these mortality rates had an important impact on the kind of institutions that were installed by
3 See Adam Szirmai (2004), “The Dynamics of Socio-economic Development – An Introduction “, Chapter 11, pp 407-9
European colonial powers in their former colonies. In case the colonial powers were only interested in extracting natural resources as quickly as possible, the development of good institutions was neglected. In case the colonies were of future importance, high quality institutions were necessarily. The authors’ assumption is that institutions change slowly over time, thus settler mortality can be seen as a good proximate for institutional quality. This is proofed by the authors’ finding that settler mortalities older than 100 years explain over 25 percent of the change in today’s institutions.4 In their regression analysis Acemoglu et al. find a positive relationship between institutions as defined and per capita incomes.
Using the institutional measurement of Acemoglu et al.; Rodrik, Subramanian and Trebbi (2002) stressed further the importance of institutions. In their analysis the authors investigate the three deeper determinants of growth and show that institutions are more important than either openness or geography.
Taking another definition of institutional quality Torsten Persson (2005) confirms the positive relationship of institutions and growth. He focuses on political institutions, especially on the form of democracy. His findings are that parliamentary, proportional and permanent democracy structures are most growth promoting.
Summarising this section, all different definitions of institutions have shown that good institutions enhance economic growth. Therefore, it can be assumed that differences in institutions shape the divergence between the Dominican Republic and Haiti.
4 See Acemoglu, et al (2001), “The Colonial Origins of Comparative Development: An empirical Investigation”, p. 1371
3. Methodology
Having provided the literature review, the next step to take is making the reader familiar with the methodology of this paper. This will be done in four steps. Firstly, the out-of-sample- prediction- method will be explained. Secondly, the underlying model of this paper will be described. Thirdly, possible objections and the weaknesses of the model, paper and research method will be discussed. Finally, it will be described how the data was collected.
3.1. Out-of-Sample-Predictions
Before going into more detail in the Methodology part, one must get familiar with this paper’s analysis method, which is called out-of-sample-prediction. For explaining purposes it is advisable to use a simple single-regression model with one predictor variable to get an understanding of this statistical tool. Such a model looks as follows:
y =
β
0 +β
1 * X + e (1)where
y = the predicted variable
β
0 = constantβ
1 = coefficientX = the variable that influences y e = error term
By means of a statistical procedure called least squares method, one can determine the coefficient
β
1 and the constantβ
0 out of a certain data set. In order to find out what the model predicts for a certain element of a sample, one can plug in values for the variable X. Now one can investigate what values for y will be predicted. So when one uses out-of-sample- predictions one makes use of already estimated coefficients of a previously estimated regression model. Just the values of the variables will be replaced by data of an element which is outside the data-set. Finally, one gets a predicted y for the element of the sample.More sophisticated models deal with more variables, not just with one as it was illustrated in equation (1). Such sophisticated models are widely used and known. They are most common in the field of finance. Here they are often used when making predictions for exchange rates or inflation as it was done in the papers of Soofi and Cao (1999) and Forni (2003). Out-of- sample-predictions are also applied in growth related studies. Acemoglu, Johnson and Robinson (2001) and Subramanian and Roy (2001) use out-of-sample-predictions in their papers “An African Success Story – Botswana“ and “Who can explain the Mauritian Miracle?- Meade, Romer, Sachs or Rodrik?”. In the first article the authors use the estimated coefficients of their earlier paper “The Colonial Origins of comparable Development: An Empirical Investigation” (2001) to proxy the quality of Botswana’s institutions. This proxy is then linked to the country’s economic performance. In the second article both authors make use of the estimated coefficients by Sachs and Warner in their paper “Sources of Slow Growth in African Economies” (1997a). Subramanian and Roy enter data for Mauritius.
Afterwards they compare Mauritian growth with baseline growth of Africa, fast-growing countries and other developing countries. From this the authors draw their conclusions concerning the peculiarities of the Mauritian growth path.
Like Subramanian and Roy, this study makes use of the estimated coefficients of the Sachs and Warner paper “Sources of Slow Growth in African Economies” (1997a). Both authors estimated their model on the basis of 79 countries. In this paper the variables will be replaced by data for the Dominican Republic and Haiti in the time periods of 1965 – 1990 and 1990 – 2002. This will give estimated growth rates. Then comparisons can be made and conclusions can be drawn. But before that one should have a closer look at the model.
3.2. The Model
The underlying model of this paper stems from the article “Sources of Slow Growth in African Economies” by Sachs and Warner (1997a). This model incorporates ten different concepts like geography, institutions, openness, primary exports and life expectancy that influence the level of total-factor productivity and therefore economic growth per capita. The model is a relative complete one due to the big variety of different concepts and especially the inclusion of the deeper determinants. A further advantage of the Sachs and Warner model is
the very high score of R², which stands for the predictive power of the model. The Sachs and Warner model scores 0.89, which is very close to the theoretical maximum of 1.00 and illustrates a high goodness-of-fit.
Various slightly different sub-models are provided in the Sachs and Warner paper.
Nevertheless, in this paper it will be made use of the regular baseline model. The baseline model looks as follows:
Ŷ = -1.63 LGGDPEA + 1.19 GEA-GPOP – 0.77 OPLGGDP + 8.48 OPEN – 0.58 LANDLOCK – 0.85
TROPICS + 45.48 LIFEE – 5.40 SQLIFEE + 0.12 CENGOVS +0.28 IQI – 3.26 SXP + C (2)
where
Ŷ = growth per capita of PPP adjusted GDP in the periods 1965-1990 and 1990-2002
LGGDPEA = natural logarithm of GDP over economic active population in 1970 and 1992
GEA-GPOP = mean annual growth of the labour force minus the mean annual growth of the total population in the periods 1965-1990 and 1990-2002 OPLGGD = openness to trade times LGGDPEA as an average in both time
periods
OPEN = openness to trade in both time periods
LANDLOCK = landlockedness of a country in both time periods (dummy)
TROPICS = share of land area of a country that is located in the tropical zone in both time periods
LIFEE = natural logarithm of life expectancy around 1970 and 1992 SQLIFEE = LIFEE squared
CENGOVS = Mean central government savings in the periods 1970-1990 and 1990- 2002
IQI = Institutional Quality Index in 1980 and 2005
SXP = Share of exports that originates from primary products in 1970 and 1990
C = constant
Now each variable will be discussed more intensively
Ŷ is the dependent variable in this model. It illustrates the real average GDP per capita growth rate in the respective time period, i.e. from 1965- 1990 and from 1990- 2002.
On the right hand side LGGDPEA means the log of real GDP per person that is economically active. The working-age of a population is defined as the age between 15 and 65. The necessary GDP is adjusted to Purchasing Power Parity. Another description of this variable would be labour productivity. But what it indeed wants to show is the catch-up potential, which was already described in the literature review on the Solow growth model (1956). For the investigation in this paper the years 1970 (chosen by Sachs and Warner) and 1992 (due to data availability) are taken.
The term GEA-GPOP stands for the mean annual growth of the working-age population minus the average population growth rate. This variable is necessary because the dependent variable is measured per capita. Therefore, it is essential for evaluating, whether a change in GDP per capita originates from a shift in the ratio of working-age population to total population. For instance a decrease in the working-age population should induce a decline in GDP per capita.
The next two variables deal with openness. The term OPEN stands for the fraction of years that are said to be open according to the definition of Sachs and Warner (1995). This definition says that an economy is open, when it fulfils the following criteria. Firstly, the average tariff rates are less than 40 per cent. Secondly, average quota of imports is also below the rate of 40 per cent. Thirdly, the black market premium is below 2 per cent. Fourthly, extreme controls on exports are absent and finally, the countries do not have a socialist regime. The variable OPLGGDP indicates openness multiplied by the natural logarithm of GDP per capita of the economic active population. This measures know-how-spillovers that are due to an open regime.
The following two variables deal with geography. LANDLOCK is a dummy variable that describes a countries’ landlockedness, i.e. a one indicates a completely landlocked economy, while a zero an economy that has access to the sea. The second variable is TROPICS. This variable was also created by the authors and shows the fraction of a country’s land area
belonging to tropical climate. Due to the fact that the geography is fixed, these values stay constant.
LIFEE and SQLIFEE deal with life expectancy of a country’s inhabitants. Life expectancy can be seen as an indicator for the quality of medical institutions and as an indicator for the people’s health status. Latter has impact on the quality of human capital which in turn spurs up development in terms of economic growth. SQLIFEE is the squared life expectancy of a country. It stands for the diminishing returns of life expectancy, i.e. at low levels one additional year of life expectancy has a larger positive impact on growth than at higher levels.
This relationship is found out by squaring life expectancy. LIFEE indicates the expected life expectancy in years (in the regression the natural logarithm will be taken).
CENGOVS stands for Central government saving. Central government saving is the result of the subtraction of current expenditures from current revenue and is expressed as a percentage of total GDP. The central government saving rate is used as an indicator for general savings in the Sachs and Warner paper. The variable is included in the model because it is assumed that a higher saving rate leads to higher economic growth. This assumption was already used in the Harrod-Domar Model (1939/1946)5, where the saving rate was seen as an engine for economic growth next to the capital-output-ratio. Further this assumption was the basis for the Solow Growth Model (1956), in which the saving rate led to capital accumulation which in turn determined the economic growth rate.
IQI means Institutional Quality Index. This index contains five elements, namely the rule of law index, bureaucratic quality index, corruption in government index, risk of expropriation index, and government repudiation of contracts index. These five subindexes are equally weighted (1/5) and summed to the “Institutional Quality Index”.
The last variable Sachs and Warner use is SXP, which means Natural Resource Exports to GDP. This measures how dependent a country is on its natural resource exports. In case that a country is highly economically depending on its natural resource exports means that it is highly vulnerable when price fluctuations occur, which are very common for primary products.
5 See Debraj Ray (1998), “Development Economics“, Chapter 3, pp. 51-64
The last term of equation (2) is the constant c. The constant has got a value of -81.94 and will be discussed in more detail in the next section “Possible Limitations of the Paper, Model and Research Method”.
3.3. Possible Limitations of the Paper, Model and Research Method
As one can imagine there are also possible objections and weaknesses concerning the “out-of- sample-prediction” method, the model and paper. These will be discussed in this section.
3.3.1. Is it just an African Model?
Concerning the title of the underlying paper “Sources of Slow Growth in African Economies”
one might think that this paper is applying just for African Economies. Already a closer look at the abstract of the article shows that this is not the case. In their abstract Sachs and Warner state:
“…The evidence suggests that the continent’s slow growth can be explained in an international cross-country framework, without the need to invoke a special explanation
unique to Sub-Saharan Africa…”
Rephrasing this sentence means that the model has also predictive power for other countries, i.e. in this case also for the Dominican Republic and Haiti.
3.3.2. The Problem with the constant
Unfortunately, the constant is not given in the article of Sachs and Warner. Both authors provide just the coefficients. Nevertheless, for the growth predictions, there is an unavoidable need for the constant. Due to that, a proximate will be provided which also originates from the article. Sachs and Warner provide 23 African countries with predicted and real average per capita growth rates. Firstly, the predicted rates are plugged in for Ŷ of equation (2) and secondly the rest of equation (2) except for c is calculated, i.e. the estimated coefficients are multiplied by the data of the respective country which is given in the text. After that the function is rearranged so that the constant c is left. This gives a number for every country.
Finally, all constants are added up and then the average is taken. This gives the number - 81.94. One may doubt this procedure because theoretically an average is not needed because the constant should always be the same. Nevertheless, this can be objected because it is obvious that there are rounding errors in the data of Sachs and Warner. Furthermore, the constant is moving in the boundaries of -81 and -82. Tests of the 23 African countries show quite good predictions when the outcomes are compared with the actual average GDP growth rates.
3.3.3. Is the model only valid for one period of time?
As stated before, the goal of this paper is to come up with the major driving forces that explain most of the divergent routes of the Dominican Republic and the Republic of Haiti.
These developments are investigated in two time periods: firstly, in the period of 1965-1990 and secondly from 1990 to 2002. At this place objections could be made because the estimated coefficients are only determined for the period between 1965 and 1990. Thus, one may argue that the impacts of the estimated coefficients have changed and are not valid any more for the second period. This might indeed be a problem, and our results to be presented later on suggest this might be the case. However, without a new re-estimation of the model with more recent data, this problem cannot be solved.
3.4. Data Collection
For the time period between 1965 till 1990 it is rather easy to collect the data needed because the data is completely provided in the Sachs and Warner paper. Therefore, decomposing economic growth for both countries in the period of 1965-1990 is straightforward.
Nevertheless, the data collection for the second time period demands more effort. But before explaining the collection process of this period, table 1 on the next page gives a broad summary, where the data comes from, from which year it is and what the values for the Dominican Republic and Haiti are.
17 Table 1 Data
Variable Year of data Source Value D.R. Value Haiti Year of data Source Value D.R. Value Haiti Ŷ 1970-1990 av. Sachs & Warner, 1997 2,12 -0,25 1990-2002 av. WDI 04 Database* 2,9 -3,02
LN Labour Productivity 1970 Sachs & Warner, 1997 7,85 7,4 1992 WDI 04 Database 9,1 8,28
Openness 1965-1990 av. Sachs & Warner, 1997 0 0 1990-2002 av. Penn World Table 6.1 0,66 0,43
Landlock Dummy Timeless Sachs & Warner, 1997 0 0 Timeless Sachs & Warner, 1997 0 0
Tropical Climate Dummy Timeless Sachs & Warner, 1997 1 1 Timeless Sachs & Warner, 1997 1 1
LN Life Expectancy around 1970 Sachs & Warner, 1997 3,98 3,78 1992 WDI 04 Database 4,2 3,98
LN Life Expectancy squared around 1970 Sachs & Warner, 1997 15,85 14,25 1992 WDI 04 Database 17,64 15,85
Primary Exports/ GDP 1970 Sachs & Warner, 1997 0,13 0,08 1990 WDI 05** 0,06 0,01
Central Gov. Savings/ GDP 1970-1990 av. Sachs & Warner, 1997 5,09 -4,72 DR 90-00/HT 90-98*** Penn World Table 6.1 5,3 -13,4
GEA-GPOP 1965-1990 Sachs & Warner, 1997 0,73 0,08 1990-2002 WDI 04 Database 0,91 -0,16
Institutional Quality Index 1980 Sachs & Warner, 1997 4,52 2,58 2004 PRS Online**** 2,2 2,1
Period 1965 until 1990 Period 1990 until 2002
* Source is the World Development Indicators 2004 database
** Source is the book World Development Indicators 2005
*** Penn World Table 6.1 provides just data from 1990 until 2000 for the Dom. Rep. and from 1990 until 1998 for Haiti
**** PRS online database for 2004. Samples are downloadable at http://www.prsgroup.com/icrg/sampletable.html
After this brief overview of the data a more detailed description of the collection method or derivation of each variable will follow.
The data for the dependent variable Ŷ stems from the World Development Indicators 2004, i.e. for each country the average per capita growth rate is taken for the period 1990 till 2002.
This is done by adding yearly growth rates up and dividing them by the number of years.
The World Development Indicator 2004 database provides also absolute numbers for the total labour force of both countries for the year 1992. Furthermore, the database delivers figures for GDP at PPP level for both countries in 1992. Dividing GDP at PPP level by the total labour force and taking logs gives GDP per economic active population or labour productivity (LGGDPEA).
A different openness measurement will be used for the second period. This is due to data problems because it is impossible to gather all the information needed to check the Sachs and Warner Openness-criteria that were used in the first period. The new measurement of openness is the sum of imports and exports over constant GDP. This method is also used by Frankel, Romer and Cyrus (1996), Dollar and Kraay (2000) and in the data-source for this variable (Penn World Table 6.1). Of course, this replacement should be treated with care because it is not the same measurement as before. Unfortunately, it is the only possibility to get quantitative data from the same data-source for both countries.
Due to the fact that an average openness value is needed for the second time period, all yearly openness values of the Penn World Table 6.1 are added up and divided by the amount of years (see av., table 2). Finally, these values are divided by hundred to make them suitable to the regression model (see av./100, table 2).
Table 2 Openness in constant prices in the Dominican Republic and Haiti in the period 1990-2000
Year 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 Av. Av./100 D.R. 62,56 61,6 68,6 70,79 70,07 65,44 66,27 64,64 67,17 63,8 66,1 66,10 0,66 HTI 57,76 75,81 33,02 47,35 33,68 39,62 35,39 31,41 30,74 n/a n/a 42,75 0,43
Source: Penn World Table 6.1
Remark: Data for the Dominican Republic was only in the period 1990-2000 available, while for Haiti only between 1990 and 1998
The data of table 2 indicates that the Dominican Republic is more open than Haiti, since higher values indicate a higher degree of openness. This finding goes in line with recent literature which also indicates that the Dominican Republic is more open than Haiti6. Therefore, the openness-values of 0.66 for the Dominican Republic and 0.43 for Haiti can be seen as good indicators for the current states of openness.
Both geographic variables are fixed and identical in both time periods. This seems to be logical because the island Hispaniola stays at the same place forever. Therefore, the variable TROPICS and the dummy access to the sea will not change.
Both figures dealing with life expectancy can be derived easily. The figures are taken from World Development Indicators 2004. Because the natural logarithm is needed, it is taken from both numbers for both countries. For the second variable the before-calculated-numbers are squared.
The data needed for primary exports as a share of GDP is also provided by the World Development Indicators 2005 of the World Bank. Because it is not possible to come up with exactly the categories Sachs and Warner had, primary exports are defined as food, agricultural, fuel and ore and metal exports in this paper. All the four categories are expressed as a percentage of total exports. Fortunately, the total exports for 1990 are also given, consequently the absolute number for every category can be calculated. These absolute numbers are scaled together and finally divided by GDP for 1990 in current prices (as Sachs and Warner did).
Unfortunately, central government savings as a share of GDP cannot be derived in the same way as it was done in the first time period by Sachs & Warner. The reason for this is lacking data, i.e. especially data on current expenditure and revenues is not available for Haiti in the second time period. Due to the fact that this paper focuses on a country comparison, one needs a similar derivation of the variables. Therefore, a proxy-method is needed and it is
6 Recent qualitative data stresses the goodness-of-fit of the calculated values from the Penn World Table 6.1.
Both countries have highly reduced their import tariffs and export subsidies, which is the first step of an increase in openness. Nevertheless, Haiti has not proceeded as good as the Dominican Republic because of its unstable political environment. Further reasons for the higher level of openness in the Dominican Republic are the huge amount of the Free Trade Zones that foster exports and attract foreign investors and the participation in several free trade treaties. Both is not so developed or even lacking in Haiti. Therefore, the grading of the Dominican Republic and Haiti can be seen as reliable as the qualitative has proofed.
assumed that this proxy is constant over time. The Penn World Table 6.1 provides current savings, which are derived differently than the central government saving rates used by Sachs
& Warner7. From the current savings rate an average value is calculated for the time period 1965-1990. These values are 6.00% and -2.04% for the Dominican Republic and Haiti respectively. Furthermore the values for central government savings as a share of GDP are taken from the Sachs and Warner article (1997a) which are 5.09% for the Dominican Republic and -4.72% for the Republic of Haiti in the period 1965 till 1990. Now central government savings are subtracted from current savings in order to determine the proxies. The Dominican proxy is valued as 0.91, while the Haitian as 2.68. As said before, these proxies are assumed to be constant over time. Thus, the proxies are also used for the determination of central government savings as a share of GDP in the second time period. Again one uses the Penn World Table 6.1 to get the values for current savings of both countries in the second period. From these values the average annual savings are calculated which are 6.24% for the Dominican Republic and -10.76% for Haiti. These values are corrected by means of the country-specific proxies. Finally, one gets a Dominican central government saving rate as a share of GDP of 5.3% and a Haitian of -13.4%. Latter means that the Haitian current revenues do not cover current expenditures, while in the Dominican Republic’s revenues exceed expenditures. These results go in line with Ray’s (1997) findings8 which show that low income countries cannot afford to save and middle income countries save a lot. These findings can be backed up since Haiti can be classified as a low income country and the Dominican Republic as a middle income country9.
The demography variable GEA- GPOP is calculated by means of the data of the World Development Indicators 2004. The Total Population Growth for both countries is given for each year in the period 1990- 2002. The growth rates of the labour force have to be calculated.
This is no obstacle because the absolute numbers for the labour forces are also provided.
Finally, for each year and each country the total population growth rate is subtracted from the labour force growth rate. Finally, for both countries the average GEA-GPOP – rate is calculated.
7 It is defined as 100-Consumption Share of Current GDP – Government Share of Current GDP. Sachs &
Warner used current revenues – current expenditure over GDP.
8 See Debraj Ray (1997) “Development Economics“, pp. 58-60
9 See World Bank, World Development Indicators 2002, CD-ROM and Adam Szirmai (2004), “The Dymamics of Socio-Economic Development – An Introduction”, pp. 14-17
As described earlier the Institutional Quality Index consists of five subindexes. The database of the PRS Group, which is the group Sachs and Warner used as well, offers its data in exchange for a payment (around 4,600 US Dollars). Fortunately, the group provides also a free sample with indexes for the year 2004. Nevertheless, only four of the five needed indexes are available – the subindex risk of expropriation is missing for all countries of the world.
Consequently one needs a replacement for the missing subindex. It is obvious that the best solution would be to replace the risk of expropriation index by an index that is highly correlated. But unfortunately also earlier data is lacking in which the risk of expropriation index is given. Thus, statistical correlations cannot be calculated. Therefore, the only possible solution is to replace the subindex by a simple number. Knowing that all subindexes vary in a boundary of zero and six, the risk of expropriation will also be in this boundary. Hence, it is assumed that the value is in the middle of both limits which is three for both countries. Of course, deviations from this value are possible and may cause changes in the IQI and predicted growth rates. Nevertheless already at this place it can be stated that these differences are so small that they do not matter. Therefore, one can be satisfied with the derivation of the IQI in the second period. For further details please see the “Sensitivity Analysis” of this paper.
4. Discussion of Results
In this section the sources of the Dominican Republic and Haiti’s economic divergence will be analysed. Therefore, the differences of each potential driver of divergence , e.g. central government savings, between the Dominican Republic and Haiti will be investigated.
Furthermore, it is analysed how these drivers result in an overall divergence in both time periods. Table 3 gives a broad overview of all the facts needed and makes the comparisons easy to understand.
Table 3* Potential drivers of growth and their impact on overall differences in predicted growth rates between DR and Haiti
Differences in concepts and predicted Differences in concepts and predicted growth rates in the period 1965-1990 growth rates in the period 1990-2002
Catch-up potential -0,7 -1,3
Openness ** 0 0,1
Landlockness 0 0
Tropical Climate 0 0
Life Expectancy *** 0,5 0,3
Central Government Savings 1,2 2,2
Institutional Quality Index 0,5 0
Primary Exports -0,2 -0,2
Demography Development**** 0,8 1,3
Constant 0 0
Estimated per capita growth 2,06 2,45
* Rounding errors are possible.
** Openness = The sum of OPEN and OPLGGD
*** Life Expectancy = The sum of LIFEE and SQLIFEE
**** Demography = GEA-GPOP
The first column of table 3 lists each variable in the regression model. Columns two and three show the contribution of each variable to the difference in GDP per capita growth rate between both countries. At the bottom row, the total difference in GDP per capita growth between DR and Haiti is given. Columns two and three can be derived by subtracting the Haitian out-of-sample-predictions from the Dominican ones in both time series. This gives the differences by variable and total between both countries.
Column two shows that there is an estimated per capita growth difference of around 2%
between both countries in the first time period. Dominican growth is mainly higher due to higher governmental savings (1.2%). Further drivers of divergence are the higher degrees of demographic development (0.8%), life expectancy (0.5%) and institutional quality (0.5%).
Only catch-up potential (-0.7%) and primary exports (-0.2%) contribute negatively to Dominican growth, which means that the growth difference becomes smaller. It means that Haiti has a higher catch-up potential than the Dominican Republic. This goes in line with the discussion on Unconditional Convergence in the literature review. Furthermore, Haiti is less dependent on primary exports. Both factors should lead to convergence. But the two items are
outweighed by the other items previously discussed. All in all, both countries diverge mainly due to differences in governmental savings and demographic development in the first period.
Column three shows an overall divergence of 2.45% in the second time period. The two most important drivers of divergence are the variables central government savings rate (2.2%) and demographic development (1.3%). Life expectancy makes up 0.34% of the difference, while openness (0.1%) and institutions (0.03%) contribute almost nothing. Again catch-up potential (-1.3%) and primary exports (-0.2%) contributed more to Haitian growth than to Dominican growth. Also in the second period the drivers of divergence are central government savings and demographic development.
When both time periods are compared with each other (see both columns 2 and 3) one can see that the estimated growth gap has increased in the second period compared to the first. This is conform the finding from graph 1. Nevertheless, the predicted divergence is smaller than in reality. Therefore, this finding will be discussed in more detail in the next section. A further comparison of both time series shows that the contribution to the overall per capita growth gap of life expectancy is still positive but smaller. The same holds for institutional quality that explains almost nothing of the difference in the second period. Thus, its importnace as a driver of divergence has become less compared to this first period. The concept openness is also of no importance. In the first period it was not present and explained nothing. In the second period it was in place but explained roughly 0.1% of divergence.
On the contrary, central government savings have even increased their initial importance in the second period; it explains now 2.2% of the difference up from 1.2% as in the first period.
The government of the Dominican Republic saves more, i.e. it has a better control over current expenditures and revenues than Haiti. As one knows from growth models like Solow (1956), a higher saving rate leads to growth. Furthermore, one can see the difference in central government savings as a measurement of institutional quality, meaning that Dominican Republic’s financial institutions are better than Haiti’s. A closer look at a few other indicators like the inflation rate, which is on average10 much lower in the Dominican Republic than in Haiti, support the assumption of better Dominican financial institutions.
10 See World Development Indicators 2004, the averages were calculated for the period 1995-2002. The Dominican Republic has an average inflation rate of 7.56% and Haiti one of 13.74%.
Demographic development is another variable that spurred the divergence up. Referring to demographic development, it seems to be the case that the Dominican Republic has reached the final stage of the demographic transition which is characterised by low birth and death rates, while the Republic of Haiti is in the second phase, where birth rates are (much) higher than death rates. The following graphs 2A and 2B support these findings.
Graph 2A Demographic Transition D.R. Graph 2B Demographic Transition Haiti
0 10 20 30 40 50
1960 1967 1975 1982 1990 1997Year
Rate in % Crude
Death RateCrude Birth
Source: Development Indicators 2004
With its values of 0.8% and 1.3% demographic development is of increasing importance and explains a lot of the difference between both states.
Summing all up leads to the following conclusion. Over the two time periods the difference between both countries has increased. The divergence of both countries is mainly due to big differences in the governmental saving rates and demographic development. It seems to be the case that the Dominican Republic has better control about its budget and that it is in the final stage of the demographic transition. On the other side Haiti has less control over its budget which is due to its unstable political environment and is still in the second phase of demographic transition.
0 10 20 30 40 50 60
1960 1967 1975 1982 1990 1997Year
Rate in % crude
Death Ratecrude Birth
5. Sensitivity Analysis
So far it was discussed what the major determinants of divergence are. Nevertheless, it was not discussed how good the predictions of the per capita growth rates actually are. A comparison of the real average annual growth rates and the estimated rates gives further insight.
Table 4 Predicted and Real Outcomes
Dominican Republic Republic of Haiti
Predicted Outcome 65-90
2.2% 0.1%
Real average Outcome 65-90
2.1% -0.3%
Difference 65-90
0.1% 0.4%
Predicted Outcome 90-02
1.3% -1.2%
Real average Outcome 90-02
2.9% -3.0%
Difference 90-02
-1.6% 1.8%
Table 4 demonstrates that the predictions of 1965-1990 are very close to the real average per capita growth rates. In the Dominican Republic there is just a gap of 0.1%. In Haiti the sign of the per capita growth is wrong but both values are so close to zero that one can be satisfied with the prediction. In the second period the differences between the predicted per capita growth rate and the real average per capita rate differ clearly. Nevertheless, the sign of the predicted growth rates is correct. Due to the estimation gap, which is in the Dominican case -1.6% and in the Haitian 1.8%, one can assume that there are some important drivers missing that explain differences in per capita growth in the second time period. For better predictions, one needs to find drivers that erase the level of underestimation in the Dominican case and decrease the overestimation in the Haitian case. These missing drivers could also solve the problem which arose in the previous section, when it was found out that the growth gap between the Dominican Republic and Haiti has increased but weaker than in reality. When missing drivers like FDI, educational achievement, tourism and inflation are included in the
model it is probable that predictions improve. The possible missing variables will be discussed in more detail in the “Concluding Remarks”.
Furthermore, a sensitivity analysis should be done in order to check how sensitive the predicted outcomes of the regression model are towards changes in its parameters. In case that a minor change in a parameter causes a larger change in the outcome, one can conclude that the outcome is very sensitive to the parameter.
Therefore, a sensitivity analysis is performed in which the initial values of the parameters central government savings, demographic development, openness and institutions will be changed. Central government savings and demographic development will be analysed because they seem to be the most important drivers of divergence according to the results predicted in the last section. Furthermore, central government savings are determined in another way than in the Sachs & Warner article. Openness and institutions are analysed for two reasons. Firstly, they are derived differently from the Sachs and Warner procedure, too.
Secondly, they explain a lot about the determination of economic growth – at least according to the literature review.
The sensitivity analysis of these four variables will be done in two stages: the values of central government savings, demographic development and openness will be changed by +10% and -10%, while all other parameters stay constant. In a second step institutions will be analysed. Because the weak-point in deriving IQI is one missing subindex, the values of this subindex will be changed. In more detail this means that the initial value of three will be replaced by the extreme points zero and six in the sensitivity analysis. This gives in turn a variation in the institutional quality index and results in changes in the predicted growth rates.
In general a large deviation of the predicted growth rates to their initial ones would imply that they are very sensitive towards changes in a certain parameter. As a consequence one should be very careful when drawing conclusion on the basis of this parameter.
Table 4 shows the initial situation, the change in the parameters central government savings, demographic development and openness and the resulting effect of the predicted growth rates in the Dominican Republic and Haiti in the second time period.
27 Table 5 Sensitivity Analysis of Central Government Savings, Demographic Development and Openness
Change in Predicted Difference from initial %-change to initial Predicted Difference from initial %-change to initial initial value Growth Rate predicted growth rate predicted growth rate Growth Rate predicted growth rate predicted growth rate Initial
Situation 0% 1,25 0 0 -1,2 0 0
Demographic
Development 10% 1,36 -0,11 -8,80 -1,22 0,02 -1,67
-10% 1,14 0,11 8,80 -1,18 -0,02 1,67
Gov. Savings
10% 1,31 -0,06 -4,80 -1,36 0,16 -13,33
-10% 1,18 0,07 5,60 -1,04 -0,16 13,33
Openness
10% 1,35 -0,1 -8,00 -1,11 -0,09 7,50
-10% 1,15 0,1 8,00 -1,29 0,09 -7,50
Dominican Republic Haiti
When one looks at columns three, four, six and seven of table 5, one can see that the differences of the new predicted growth rates from the initial predicted growth rates are small in the Dominican Republic and Haiti. When the variable demographic development is changed by +/- 10% the Dominican Republic experiences a maximum difference from its initial predicted growth rate of +/- 0.11. The second biggest difference in Dominican growth is caused by changes in openness which is followed by changes in central government savings. In Haiti a maximum difference of +/-0.16 from its initial growth can be found when the value of central government savings is changed. Openness causes the second biggest difference, while changes in demographic development cause almost no difference from initial Haitian growth.
When one investigates columns five and eight, it becomes clear that the findings of the previous section can be supported. In the Dominican Republic the relative values can be seen as robust since they vary only between five and nine percent, when the parameters are changed by +/- 10%. Haiti experiences more extremes. Percentage changes in demographic development cause only a change of +/- 1.7% in the predicted growth rate which is a very robust result. Changes in openness let the growth predictions vary between +/- 7.5% which is robust at moderate level. But changes in the central government saving rate cause a variation of around +/-13%. This means that the variation in the predicted growth rate exceeds the percentage change in the variable central government savings. One might conclude that the result is not robust. But this assumption can be rejected by mathematically reasons. The Haitian value for central government savings is relatively large. Therefore, percentage changes in savings result in larger deviations. Furthermore, as already shown in column seven, absolute differences from the initial predictions are so small that one can speak of a robust result.
These findings show three things: firstly, the deviations are so small that it can be concluded that the predicted growth rates are robust towards changes in all three parameters. Secondly, there is no clear order of the parameters, i.e. it is not clear which parameter causes the biggest variation in predicted growth rates. The results show that this differs from country to country.
Thirdly, the relative deviations are more constant and at a moderate level in the Dominican Republic, while the percentage deviations tend to be more extreme in Haiti.
Now the institutional measurement IQI will be analysed. An overview of changes in the missing subindex of IQI, the resulting changes in IQI and variations in predicted growth rates are presented in table 6.
Table 6 Sensitivity Analysis of Institutional Quality Index (IQI)
IQI Difference to %-change in Predicted Difference to predicted %-change in initial IQI IQI growth rate growth rate predicted growth rate Dominican Rep.
Initial Situation (3) 2,20 0,00 0,00 1,25 0,00 0,00
Lower Boundary (0) 1,60 0,60 27,27 1,08 -0,17 -13,60
Upper Boundary (6) 2,80 -0,60 -27,27 1,42 0,17 13,60
Haiti
Initial Situation(3) 2,10 0,00 0,00 -1,20 0,00 0,00
Lower Boundary (0) 1,50 0,60 28,57 -1,36 -0,16 13,33
Upper Boundary (6) 2,70 -0,60 -28,57 -1,03 0,17 -14,17
The second column shows the institutional quality index in the different situations. Its values differ from the initial values by +/- 0.60 in both countries (see column 3). This makes up a variance of 27.3% in the Dominican Republic and 28.6% in Haiti which can be found in column 4. These changes in institutional quality cause only minor differences in the predicted growth rates of around +/- 0.17 (columns 5 and 6). When the changes in predicted growth rates are analysed relatively to its initial values, two things are striking. Firstly, the results are robust in both countries since a 28%-change in IQI (column 4) leads only to a variation of 14% in the predicted growth rates (column 7). Secondly, there are just neglible differences between both countries. Taking all together, the differences in predicted growth rates are so small that one can conclude that the predicted growth rates for the Dominican Republic and Haiti are robust towards changes in institutional quality.
Concluding this section one can be satisfied with the robustness of the predicted growth rates towards the parameters central government savings, demographic development, openness and institutional quality. This outcome stresses the importance of this paper’s findings.
Unfortunately the growth predictions for the second period are not precise enough, when they are compared to the real average growth rates. This leads to the conclusion that some important drivers are missing that explain the gap between the estimated and real growth rates.
6. Concluding Remarks
As described in the literature review, many studies showed that openness and institutions shape a country’s economic growth path. In this paper it was tried to find out whether these determinants were also the sources of divergence between the Dominican Republic and Haiti.
The results showed that neither openness nor institutions explain most of the divergence between both countries. It was more surprisingly to see that the most important drivers of divergence are savings and demographic development. This finding brings the loop back to (neo-)classical growth models in which savings and population growth are the crucial determinants of economic growth [Harrod-Domar (1939/1946), Solow (1956)]. But still the question remains: what is behind these two factors? It is institutions - but different institutions than measured in the IQI of this paper.
Therefore, the first task for future research on this topic should be to look for the endogeneity of these concepts. Ray (1998) argues that a country’s level of savings is determined by income inequality. According to Ray rich and poor parts of the population do not or cannot afford to save. It is the middle class that saves most. At a first glance it seems to be the case that a strong middle class is the solution to saving and growth problems. Therefore, one might think about redistribution policies to strengthen this part of the population. But unfortunately, there is no general solution. In poor countries it might be the case that redistributive policies weaken that part of the population that normally saves. Consequently, no person saves at all.
Thus, the initial situation without redistribution would have been better. Therefore, redistribution policies should be treated with care. Sometimes inequality is even enhanced by the rich part of the population that wants to keep inequality in place in order to strengthen its position. Therefore, the rich do not invest in human capital for the majority. Furthermore, they might suppress democratic systems to their advantage [Sokoloff & Engermann (2000),
Bourguignon & Verdier (2000)]. These are only a few factors that should be kept in mind, when the endogeneity of savings is analysed.
Demographic development might have several determinants like education, (under-) nutrition and economic reasons. The educational level can lead to an acceleration of the demographic transition since education can impart methods of contraception which puts a brake on birth rates. Further education can impart facts about hygiene in order to decrease death rates. Additionally, a better nutritional level leads to lower death rates. But also economic reasons can explain why birth rates differ from society to society. Children do not only create benefits, they also cause costs. Firstly, they cause direct costs like feeding and schooling. Secondly, they cause opportunity costs. Opprtunity costs are the foregone incomes that occur when the child is brought up. In more detail this means that the family loses income because one part of the parents has to give up her work in order to take care of the child. In developing societies this opportunity cost is low because the general income level is low.
Therefeore, the birth rates are at a high level. In more sophisticated societies like the United States the opportunity cost is much higher. Consequently, children become a luxury good.
The findings of this paper rest on a number of assumptions and uncertainties which reduce the reliability of the outcomes. One might think about the assumptions that the coefficients of the regression model stay the same in the second time period, the derivation of openness and savings and the missing subindex in IQI. Therefore, the findings should be considered with care. Nevertheless, the sensitivity analysis has shown that the predicted growth rates are robust towards changes in savings, demographic development, institutions and openness. This outcome contributes positively to the findings.
Furthermore, this paper has also shown that the predicted growth rate was overestimated in the Haitian case and underestimated in the Dominican case in the second period. It seems to be the case that some important drivers are missing in the application of the Sachs and Warner regression model in the second time period. This leads to the second task of future research. It should think about several concepts that could improve the predictions. Other research has suggested additional variables to explain growth such as tourism, foreign direct investment, inflation and education. Authors like Durbarry (2004), Narayan (2004) and Dritsakis (2004) stress the positive impact of tourism on economic growth. Thus, an estimated coefficient for tourism would be positive, when it was included in the model. Due to the fact that the Dominican tourism sector has positively developed in the recent past, predicted economic per capita growth rates would be boosted. In Haiti the figures
show little and mostly declining tourism rates and therefore predicted rates would be brought down. Similarly, in the point-of-view of Akinlo (2003), Hunya (2004) and Hansen and Rand (2004) FDI inflows are an important growth-promoting factor. The Dominican Republic has a relatively high level of FDI inflows, while the level in Haiti is almost neglible. Studies of Dollar and Kraay (2000) and Easterly, Loayza and Montiel (1996) say that inflation growth decreases the possibility of positive economic growth. Due to the fact that the inflation rate is much higher in Haiti than in the Dominican Republic, this might provide another explanation.
Finally, one might think about educational attainment. Loening (2004), Musila and Belassi (2004) and Gutema and Mekonnen (2004) emphasize the positive education-growth- relationship. A logical consequence of this finding is that the estimated coefficient for education would be positive. The Dominican Republic has a high level of education, while Haiti has a small one. Therefore, the predicted Dominican economic growth would gain again. The concepts tourism, FDI, inflation and education are not included in the regression model yet. Nevertheless, the Dominican and Haitian data for these concepts seems to suggest improvements in the predicted growth rates, i.e. bringing down the Haitian overestimation and the Dominican underestimation. Future research has to find out the real impact of these parameters on economic growth. This would entail running the regression with a new data-set for many countries, including these new variables. But doing this would go beyond the scope of this thesis.
