Oil Boom and the Dutch Disease
A Study of Major Oil-exporting Countries
Bachelor Thesis Economics
Mengfan Cao
Student Number: 10397868
Supervisor:
Ms O. (Oana) Furtuna
Date: 15-7-2015
AbstractThis paper is devoted to discuss the presence of Dutch Disease in the major oil-exporting countries after a crude oil boom. The core model suggested by Corden and Neary (1982) is explained and as well as earlier empirical studies. The model constructed in this paper uses fixed effect model with panel data on 32 major
oil-exporting countries throughout ten years in the 21th century. The result we obtain is that the Dutch Disease is present in agriculture and service sector but not obvious in manufacturing sector.
Statement of Originality
This document is written by Student, Mengfan Cao, who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
1 Introduction
Over the decades, the natural resource curse has always been a controversial topic in international economics. Most famously, Sachs and Warner (1995) conclude in their study that economies with abundant natural resources tend to develop slower than economies without sustainable natural resources. Such a statement, though
astonishing, should be judged carefully because it has important implications for the government to make macroeconomic or development policies (Stijns, 2003).
The Dutch Disease is regarded as one of the explanations of natural resource curse but not exactly explains the idea that resource curse is trying to express(Davis, 1995). In a nutshell, the Dutch Disease refers to a situation that an increase in a booming export sector will hurt the rest of the tradable goods sector (Stijns, 2003). Corden and Neary (1982) suggested a core model for the Dutch Disease and the model was cited by many empirical studies for the following decades. A thorough study of the models for Dutch Disease and a conduction of an empirical study on this topic will bring us more insights in the international economics, and also theoretical preparation for the
suggestions of macroeconomics policies.
In this thesis paper, a research question will be tested that whether there is any Dutch Disease in the major oil exporting countries follows a crude oil boom. The paper is divided in five sections. In the second section, theoretical framework will be
explained, including the definition of Dutch Disease and the core model we use in our study. In the third section, Empirical studies will be discussed and follows by the hypotheses for our study. In the fourth section, there will be explanations for the model and dataset we use in the study, followed by a thorough analysis on the results we obtain and a discussion on the limitations. In the conclusion a statement will be made taking into account of all factors discussed throughout the thesis paper.
2 Literature Review
2.1 What is the Dutch Disease?
The Economist initially coined the term Dutch Disease on November 26, 1977. It was
used to describe the bad macroeconomic situation in the Netherlands following the discovery of the 1959 Slochteren gas field in Groningen (Corden and Neary, 1982). The discovery was followed by a sharp appreciation of the Dutch Gulden, and thus negatively influenced the Dutch export and manufacturing industry. According to Corden (1982) and Corden and Neary (1984), the Dutch Disease is summarized as the notion that an exogenous increase in resource prices or in resource output results in real exchange rate appreciation and a negative impact in manufacturing sector. Instead of limiting the effects to the manufacturing sector, in Stijns (2003) paper, he concludes that the Dutch Disease mainly refers to a situation in which a booming export sector drives up the relative price of non-tradable goods and services, thus hurting the rest of the tradable goods sector. This means that, the Dutch Disease may also negatively impact the agriculture sector and also other tradable goods sectors besides the manufacturing sector.
As for the reasons behind the resource boom, Corden (1984) concludes that a resource boom can occur in three different ways: first, an exogenous technological progress may take place in the booming resource sector, thus improving the efficiency of exporting the resource; second, a windfall discovery of natural resources may happen in a country; third, an exogenous rise in the world price of a natural resource of a country may happen.
the ‘natural resource curse’. He illustrated that the ‘natural resource curse’ hypothesis is based on the observation that on average, resource-rich economies grow slower than resource-poor economies. Most famously, Sachs and Warner (1995) proved this hypothesis by conducting research and showing that economies with a high ratio of natural resource exports as part of GDP in 1971 tended to have low growth rates during the following period 1971-1989. And thus they conclude, “One of the
surprising features of modern economic growth is that economies abundant in natural resources have tended to grow slower than economies without substantial natural resources.” This surprising result has been hold true after controlling for other growth determinants, such as initial per capita income, trade policy, government efficiency and investment rates (Stijns, 2003).
The other two explanations of the ‘natural resource curse’ in Davis (1995)’s paper are, firstly, the windfall discovery of resource may give rise to a fight over existing
resources. Therefore, governments and private agents have incentives to engage in rent-seeking behavior. This leads to poor institutional quality and lower growth. Secondly, the volatility of resource rents causes the lower growth and investment due to the existence of financial market imperfections.
However, the Dutch Disease and the ‘natural resource curse’ are not synonymous but two separate issues, as explained in Davis (1995). Davis illustrates that the natural resource curse regards the resource boom as “a net economic loss, where the present value of the positive effects of the boom are more than offset by the present value of negative effects”. This means that, the natural resource curse is more concerned with the worries about the struggles from the resource boom. In his paper, these worries includes five mechanisms: the growth-inhibiting deindustrialization; loss of
competitiveness of the firms; poor prediction of global price shocks; government’s overly optimistic expectation of the future revenue; and lastly, the inequitable factor income effects and worsening income distributions.
On the other hand, the Dutch Disease is a rather morbid term that only denotes the coexistence of booming and lagging sectors in an economy because of an increase (temporary or sustained) in export earnings. It describes the causes and structural effects of boom-induced growth. The essential problem explained by the Dutch Disease is resource allocation and the burden of adjustment originated from the losing factor and political pressure. Davis (1995) also mentions that if the resource boom is indefinite, the Dutch Disease may only describe the transformation of the economy from one long-run equilibrium to another.
After differentiating the two terms ‘Dutch Disease’ and ‘the natural resource curse’, we may be clearer about what the core model of Dutch Disease can explain developed by Corden and Neary (1982) in the next section.
2.2 Model Explaining the Dutch Disease
In order to provide a systematic analysis of the Dutch Disease in an open economy, Corden and Neary (1982) provide the core model of Dutch Disease. The core model has been cited by many empirical studies. Its framework is a small open economy in which produces three kinds of goods. The first two are traded at exogenously given world prices, which are labeled as ‘energy’, 𝑋!, and ‘manufactures’, 𝑋!. The other one good is traded at the price of which moves flexibly according to the equilibrium of domestic supply and demand, which is labeled as ‘service’, 𝑋!.
Corden and Neary (1982) make two simplifying assumptions. Firstly, the models are in real terms, so that monetary considerations are not here on the stage; only relative prices are determined. Therefore, national output always equals to expenditure, and so trade balanced entirely. Secondly, they assume that the markets for the factors of production (labor and capital in this case) are perfect. This assumption means that real
wages are perfectly flexible and that full employment is maintained at all times.
In Corden (1984), Corden presented the core model again and this time it consists of three sectors in a small economy: the booming sector (B) and the lagging sector (L), which produces the tradable goods (T). Here the booming sector includes oil, gas or mineral industry and the lagging sector includes manufacturing industry. The final sector non-tradable sector (N) produces services. Here the assumption is that capital and labor are both used in these three sectors.
Corden and Neary (1982) make an important distinction between the spending effect and the resource movement effect after a resource boom. The spending effect occurs when disposable income rises because of a resource boom in the booming sector, and there is an inflow of foreign exchange. Agents will consume the extra income directly or pay the government through the additional tax revenue. The inflow of foreign exchange is invested in imports or in the money supply within the country, which has no direct impact on the demand for domestically produced goods. If the foreign exchange inflow is converted into domestic currency, then it will be invested in the non-tradable sector. In Corden and Neary (1982), the income elasticity for
non-tradable goods is assumed to be positive, so that the demand for non-tradable and tradable goods will increase. Under the condition that the world demand will not be influenced by the increased domestic demand in a small open economy, the price of tradable goods will not increase more than that of the non-tradable goods.
To discuss it further, the increase of price will leads to two possible outcomes: first, the real exchange rate of a country will appreciate, which makes it less competitive with the tradable goods in the world markets; second, the rising wages within the non-tradable sector will attract a shift of labor from the tradable sector to the non-tradable sector, and this will continue until the factor prices within the three sectors become the same. Within the context of the model, the real exchange rate here is defined as the price of non-tradable products relative to that of tradable products
(Corden and Neary, 1982).
On the other hand, the resource movement effect occurs when there is an increase in the marginal product of labor in the booming sector (Corden, 1984). After the exogenous shock in the wealth, the booming sector will require more labor to
distribute the extra work. Under the assumption of international immobility of labor, the labor in the domestic country will be constant all the time. This means that domestically, labor will shift from the lagging sector and non-tradable sector to the booming sector.
Corden (1984) explains the resource movement effect in two separate processes: direct de-industrialization and indirect de-industrialization. Direct de-industrialization describes the labor movement from the lagging sector to the booming sector, resulting in the output decrease in the lagging sector. It’s called ‘direct’ because this kind of movement of the production factors does not involve the non-tradable sector, so that it doesn’t make any change in the real exchange rate.
Indirect de-industrialization describes the labor movement from the non-tradable
sector, such as service, to the booming sector. The movement is interacted with the spending effect in the way that due to the outflow of labor in the non-tradable sector, and then the non-tradable goods are in higher demand. This causes that the real exchange rate appreciates and so a subsequent flow from the lagging sector to the non-tradable sector to meet the high demand.
In a conclusion, the spending effect results in an output increase in the non-tradable sector, but the resource movement effect results in an output decrease in the
non-tradable sector. Thus it’s ambiguous about the movement in the non-tradable sector since it depends on the relative strength of the two effects. However, it is certain that in the lagging sector there will be an output decrease under these two effects.
2.3 Empirical Works on the Dutch Disease
In Stijns’s (2003) study ‘An Empirical Test of the Dutch Disease Hypothesis Using a Gravity Model of Trade’, the author tests the hypothesis of negative effect of a resource boom on a country’s manufacturing exports by using a gravity model of trade. In his paper, he uses trade data that covers most countries for most of the last thirty years in the 20th century.
According to Stijns (2003), four hypotheses may be identified after a resource boom: ‘first, the real exchange rate unambiguously appreciates; second, there is a likely though theoretically ambiguous increase in non-traded output; three, production in the manufacturing sector unambiguously falls; and fourth, there is a fall in manufacturing exports.’ Among these hypotheses, Stijns (2003) chooses to only test for the last hypothesis, because of the richness of the trade data. For others, some researchers have already tested the first hypothesis, while the second and third hypotheses are having not enough available sectorial production data for every country.
Stijns (2003) explains that by using the gravity model of international trade, it’s more feasible to implement the variables without worrying about other macroeconomic effects not explained enough. The origin of the gravity analogy is the fact that trade between two countries is related to their GDPs and also their distance. The gravity model, as assured by Stijns (2003) after referring to previous people’s work, has been proved to be simple and credible.
Three approaches are mentioned to define the Dutch Disease terms in the model of Stijns (2003): first, only using the world price of energy as the independent variable; second, only using the total net energy exports from each country on both sides of the manufacturing export flows as the independent variable; third, using both the world price of energy and the total net energy exports to combine the advantages from the
proceeding two approaches.
Stijns (2003) finds that the results from all the three approaches significantly supports the Dutch Disease theory, that is, the world price of energy and the total net energy exports both negatively impact the manufacturing exports. The exact results in his words are ‘a one percent increase in the price of energy will, ceteris paribus, decrease a net energy exporter’s real manufacturing exports by half a percent,’ and also ‘a one percent increase in the net export exports will, ceteris paribus, decrease a net energy exporter’s real manufacturing exports by one eight of a percent.’
In Lartey, Mandelman & Acosta’s (2008) study ‘Remittances, Exchange Rate Regimes, and the Dutch Disease: A Panel Data Analysis’, the authors test the relationship between remittances and also other macroeconomic factors to find the presence of Dutch Disease. In their research, they find data on 109 developing and transition countries for the period 1990 – 2003. They implement a dynamic panel data model and the model is estimated by using generalized method of moments estimator. The model is used for the reason to adapt endogeneity in the explanatory variables. The macroeconomics variables they are interested in are: GDP, money supply (M2), trade index (good and services), trade openness (sum of exports and imports of goods and services as percentage of GDP), TNT (ratio of the sum of agriculture and
manufacturing output as share of GDP over services’ output as share of GDP), Agriculture (as percentage of GDP), Manufacturing (as percentage of GDP) and Services (as percentage of GDP), etc.
Lartey, Mandelman & Acosta’s (2008) implement two different equations to find the presence of positive signs of Dutch Disease on remittances in the research. In the first equation, the dependent variable is the ratio of tradable to non-tradable output. In the second equation, the real exchange rate is used as the dependent variable. They also would like to test whether the spending effect and the resource movement effect under the Corden and Neary (1982) model can be different by following different exchange
rate regimes.
The results found by Lartey, Mandelman & Acosta’s (2008) shows that if remittances increase, then there will be a decrease in agriculture and manufacturing, but also an increase in services (all as percentages of GDP). This may because that ‘the spending effect culminates in an increase in the relative price of nontradables and real exchange rate appreciation’, while ‘the resource movement effect favors the nontradable sector at the expense of tradable goods follows an increase in remittances’. Moreover, the authors find that the country with a fixed exchange rate regime has a larger real appreciation if there is an increase in remittances. The Dutch Disease here in Lartey, Mandelman & Acosta’s (2008) is the evidence that ‘the share of services in total output rises while the share of manufacturing declines.’ All of these results are tested to be hold after controlling for the other variables and dealing with endogeneity issues.
In Fardmanesh (1991) study ‘Dutch Disease Economics and the Oil Syndrome: An Empirical Study’, he integrates the Dutch Disease model analysis into a reduced-form three-sector model. Fardmanesh (1991) does the research on five developing
oil-exporting countries, all having significant agricultural and manufacturing sectors. These five countries are: Algeria, Ecuador, Indonesia, Nigeria and Venezuela. Time-series annual data are used for the period of 1966-1986 (Fardmanesh, 1991). The hypotheses made by Fardmanesh (1991) are: first, if there is an increase in world relative price of manufacturing, the agricultural sector will contract and the
semi-traded/semi-nontraded manufacturing will expand, whereas the movement in nontraded sector will be ambiguous; second, if there is an increase in oil revenue, the agriculture sector will contract, while the semi-traded/semi-nontraded manufacturing and nontraded sector will expand.
The two explanatory variables used are O (ratio of oil revenues to GDP) and P (the index of world relative price of manufactured goods to agricultural products for less
developed countries). The results of the research are significant for presence of the Dutch Disease in the most countries except the agricultural sector of Venezuela: ‘In all other countries an oil boom contracts the agricultural sector via the increase both in their oil revenues and in their world price of manufacturing. The increase in their oil revenues expands both their manufacturing and nontraded goods sectors.’
2.4 Formulation of Hypotheses
After studying the core model from Corden and Neary (1982) and other empirical studies, we would like to find the presence of Dutch disease by discovering the impact on different sectors: agriculture, manufacture and service sectors, if there is an oil boom in a country. It seems that from different empirical studies mentioned above, the hypothesis for the presence of Dutch Disease are not the same, especially when it comes to the service sector (nontradable sector). Thus here we will base our
hypothesis mainly on the core model suggested by Corden and Nearly (1982).
From the core model we may expect that the impact on agriculture and manufacture sectors will be negative (since they are counted both as lagging sector), whereas the impact on service sector and also tradable to nontradble ratio will be ambiguous. Thus we may conclude the presence of Dutch Disease in this study as the output decrease in agriculture and manufacture sectors after an increase in the oil export. At the
meantime, the service sector is also tested but not making any hypothesis on it. Therefore, we can test the research question using the following hypotheses:
H0: An increase in crude oil exports in a major oil exporting country will not make negative impact on the agricultural and manufacturing sectors of that country.
H1: An increase in crude oil exports in a major oil exporting country will make negative impact on the agricultural and manufacturing sectors of that country.
3 Methodology and Data
3.1 The Regression Models
In order to test the hypotheses, the regression analysis here is using panel data that varies in two dimensions, in this case individual country and each year. According to Stock and Watson (2007), panel data consists of observations on the same n entities at two or more time periods T. By using panel data here, the result will be more accurate than by using single time series or cross-sectional data. Since my research is based on analyzing 32 countries within 10 years instead of only by looking into one of them. However, the panel data we use in this study is unbalanced because some data are not available for all the years we cover.
Fixed effect model may control for omitted variables in panel data when the omitted variables vary across entities but do not change over time. The reason to use fixed effect model instead of random effect model is well explained in Stock and Watson (2007). Fixed effect model is preferred when all the studies included in the analysis are functionally identical, and also when the research goal is to compute the common effect size for the identified population instead of generalizing to other populations. These two conditions are met in my research. On the other hand, random effect model is preferred when the researcher is accumulating data from a series of studies that had been performed by researchers operating independently. Thus in my research using fixed effect model rather than random effect model makes better sense.
The four regression models in our research are presented as follows: Model 1: 𝑎𝑔𝑟𝑖𝑐𝑢𝑙𝑡𝑢𝑟𝑒!" = 𝛽!𝑂𝑖𝑙!" + 𝛽!𝑔𝑑𝑝𝑔𝑟𝑜𝑤𝑡ℎ!"+ 𝛽!𝑡𝑟𝑎𝑑𝑒𝑜𝑝𝑒𝑛!"+ 𝛽!𝑚𝑎𝑛𝑢𝑓𝑎𝑐𝑡𝑢𝑟𝑒!" + 𝛽!𝑠𝑒𝑟𝑣𝑖𝑐𝑒!"+ 𝛼!+ 𝜇!" Model 2: 𝑚𝑎𝑛𝑢𝑓𝑎𝑐𝑡𝑢𝑟𝑒!" = 𝛽!𝑂𝑖𝑙!" + 𝛽!𝑔𝑑𝑝𝑔𝑟𝑜𝑤𝑡ℎ!"+ 𝛽!𝑡𝑟𝑎𝑑𝑒𝑜𝑝𝑒𝑛!"+ 𝛽!𝑎𝑔𝑟𝑖𝑐𝑢𝑙𝑡𝑢𝑟𝑒!" + 𝛽!𝑠𝑒𝑟𝑣𝑖𝑐𝑒!"+ 𝛼!+ 𝜇!" Model 3: 𝑠𝑒𝑟𝑣𝑖𝑐𝑒!" = 𝛽!𝑂𝑖𝑙!"+ 𝛽!𝑔𝑑𝑝𝑔𝑟𝑜𝑤𝑡ℎ!" + 𝛽!𝑡𝑟𝑎𝑑𝑒𝑜𝑝𝑒𝑛!"+ 𝛽!𝑎𝑔𝑟𝑖𝑐𝑢𝑙𝑡𝑢𝑟𝑒!" + 𝛽!𝑚𝑎𝑛𝑢𝑓𝑎𝑐𝑡𝑢𝑟𝑒!"+ 𝛼!+ 𝜇!" Model 4: 𝑇𝑁𝑇!" = 𝛽!𝑂𝑖𝑙!"+ 𝛽!𝑔𝑑𝑝𝑔𝑟𝑜𝑤𝑡ℎ!"+ 𝛽!𝑡𝑟𝑎𝑑𝑒𝑜𝑝𝑒𝑛!"+ +𝛼!+ 𝜇!"
Where i indicates country and t indicates time.
3.2 Variables and Data
Since we are studying the impact on the different sectors in the economy of a country after an oil boom, the result will be more significant if we use the major oil-exporting countries. We choose 32 major oil-exporting countries by referring to the World
Development Indicators dataset of the World Bank. The dataset calculates the oil rents by subtracting the total costs of production from the value of crude oil production at world prices. We also choose the time period from 2001 to 2010 because that most of the data are available at this period. The minimum average oil rent is 5 percent of GDP for all the chosen countries. In Appendix 1 there presents a table of the 32 chosen countries.
The datasets for dependent variables: agriculture, manufacturing and services, refer to the database of the World Bank by calculating the total value added as percentage of GDP in nominal term. Value added is the net output of a sector after adding up all outputs and subtracting intermediate inputs. In the World Bank dataset, the economy of a country is divided in four sections: agriculture, services, industry excluding manufacturing and manufacturing, according to the International Standard Industrial Classification (ISIC), revision 3. Here the oil industry is included in the industry excluding manufacturing sector. Agriculture comprises value added from forestry, hunting, and fishing as well as cultivation of crops and livestock production. Services comprise value added from wholesale and retail trade (including hotel and restaurants), transport, and government, financial, professional, personal services such as education, health care and real state services, according to the World Bank.
The using of TNT variable as dependent variable in the model is inspired and followed by Lartey, Mandelman & Acosta (2008) paper. Here TNT is calculated as tradable output divided by non-tradable output, where tradable output is the sum of manufacturing and agriculture and non-tradable output is the sum of services.
The explanatory variable of crude oil exports is calculated based on the database of the U.S. Energy Information Administration (EIA). Here the crude oil exports describes for all countries per year in thousand of barrels per day.
and Lartey, Mandelman & Acosta (2008). Both of these two empirical studies use GDP growth as a control variable. It’s estimated based on the dataset from the World Development Indicators in the World Bank. In their dataset, the annual percentage growth rate of constant GDP at market prices based 1995 USD. The variable trade openness is used also inspired by Lartey, Mandelman & Acosta (2008) paper. Here trade openness is calculated as the sum of exports and imports of goods and services as percentage of GDP. The exports and imports as percentages of GDP are based on the dataset from the World Development Indicators in the World Bank.
Besides the correlations, in Table1 we also put the unit root test result in the last column to see whether the datasets of the variables are not stationary in any degree. The unit root test for panel data we use is Fisher-type unit-root test based on augmented Dickey-Fuller tests, because the Fisher-type test is available for the unbalanced panel data. In the last column, the numbers are referring to the Inverse chi-squared P statistic since it’s the one to test with finite number of panels. The null hypothesis is that all panels contain unit roots and the alternative hypothesis is that at least one panel is stationary. The star on top of the number in the last column shows the significance of rejecting the null hypothesis. From the p-value we may conclude
Table 1 agriculture manufacture service tntratio oil gdpgrowth tradeopen DF test
agriculture 1.000 80.331* manufacture -0.187 1.000 125.867** service -0.152 0.681 1.000 77.400* tntratio 0.923 -0.108 -0.282 1.00 184.500** oil -0.238 0.002 0.097 -0.187 1.000 98.986** gdpgrowth 0.028 -0.151 -0.205 0.033 0.018 1.000 106.749** tradeopen -0.146 0.043 -0.214 -0.050 -0.173 0.082 1.000 117.336** **, * correlation is significant at 0.01 and 0.05 level respectively
that if we put significance level at 0.05, then all of the variables are not having unit-root test. However, if we set significance level at 0.01, then the agriculture and service variables are having a little degree of unit-root. Therefore, in the next regression session, we will also test the agriculture, manufacturing and service variables in difference as dependent variables to see elasticity. Besides that, we will also use the oil export variable in difference to see the robustness of the results.
4 Results and Interpretation
In this section, the results of the regression analysis are described. We conduct regressions on our four models and also make sensitivity analysis to test the robustness of the results we obtain from each model.
4.1 Agriculture Sector
Table 2 (1) (2) (3) (4) (5)* oil -0.0022 (0.006) -0.0024 (0.000) -0.0027 (0.000) -0.0026 (0.000) -0.0001 (0.023) oil2 -0.388 (0.590) -0.1259 (0.867) -0.1378 (0.857) 0.0200 (0.670) gdpgrowth -0.0920 (0.003) -0.0257 (0.235) tradeopen -0.0416 (0.023) -0.0296 (0.026) -0.0359 (0.011) -0.000 (0.872) manufacture -0.2712 (0.043) 0.3111 (0.003) 0.3872 (0.000) 0.3756 (0.001) -0.0018 (0.798) service -0.1074 (0.061) -0.1327 (0.001) -0.1678 (0.000) -0.1391 (0.001) 0.0042 (0.101) 𝑹𝟐 0.1287 0.1382 0.1598 0.1349 0.0401 Observations 258 254 259 259 257The results of the regression analysis on agriculture sector (Model 1) are shown in Table 2. The analysis is conducted in five steps. The first regression on the agriculture variable in level includes the explanatory variables of crude oil export, GDP growth, trade openness, manufacture and service. From the first column of the table, we can
read that the coefficient of crude oil export is negative as the theory supported before, and is significant at less than 1% level. This implies that an increase in crude oil export leads to a decrease in the total value of the agricultural sector. From the 𝑅!
we can see that the model explains 12.87% of the variation in the data.
Regression two in the second column includes an extra explanatory variable of crude oil export in difference, in order to test the robustness. We see that the 𝑅! is
increased, and the coefficient of crude oil export in level is more significant. Thus it makes more sense to include the crude oil export variable in difference in this model. In the third column, we exclude the GDP growth variable. Surprisingly, 𝑅! is
decreased while the coefficient of crude oil export is still having a negative sign. This means that by including GDP growth variable is not going to help us any further in proving the existence of the Dutch Disease. However, in the fourth column, we exclude further the trade openness variable. But this time 𝑅! is decreased while the
coefficient of crude oil export is still negative. So we should include trade openness as an explanatory variable.
The last fifth column is a regression on agriculture variable in difference by using all the explanatory variables mentioned except the GDP growth. This regression is conducted to see the elasticity of oil export on agriculture sector. We can read from the coefficient of oil export in difference that 1% increase in crude oil export will cause an increase of 2% in agriculture sector, which is surprising because the
economic theory predicts a negative relationship. But the significance level is 0.670, which means that the positive relationship not significant. From all the analysis above, we may conclude that the impact on agriculture sector after an oil boom is slightly negative.
4.2 Manufacturing Sector
Table 3 (1) (2) (3) (4) (5)* oil -0.0002 (0.669) 0.0003 (0.415) -0.0002 (0.633) -0.0002 (0.595) -0.0000 (0.392) oil2 0.2869 (0.531) -0.0117 (0.808) gdpgrowth 0.0236 (0.120) 0.0168 (0.223) 0.0248 (0.103) -0.0031 (0.032) tradeopen 0.0094 (0.302) 0.0090 (0.290) 0.0104 (0.259) -0.0009 (0.322) agriculture -0.0667 (0.043) 0.1263 (0.003) -0.0544 (0.083) -0.0721 (0.027) -0.0018 (0.685) service 0.1565 (0.000) 0.1344 (0.000) 0.1395 (0.000) 0.1497 (0.000) 0.0054 (0.038) 𝑹𝟐 0.1750 0.1422 0.1424 0.1710 0.0741 Observations 258 254 263 258 245As before, in Table 3, the regression analysis on manufacture sector (Model 2) is also conducted in five steps in the same logic. The first column shows that the coefficient of oil export is negative as the theory explained, but is not a significant level. While we are including the oil export variable in difference, the 𝑅! decreases and the sign
of the coefficient of oil export in level changes to positive. Thus we do not include oil2 in the model for the rest of regressions here. In the third and fourth column, we exclude the GDP growth and trade openness variable separately, and both decreases the 𝑅! without changing the sign of coefficient of oil. So it makes sense to put those
two variables in the model. In the last column we conduct regression on the
manufacture variable in difference. From the coefficient of oil2 we may read that an increase of 1% of oil export will bring a decrease of 1.17% in manufacturing sector,
though the conclusion is not at a significant level. Since for all other regressions the coefficients of oil export variable are not at significant level as well, there is not enough statistical evidence to conclude that a change in oil exports negatively influences the growth in the manufacturing sector.
4.3 Service Sector
Table 4 (1) (2) (3) (4) (5)* oil -0.0023 (0.015) -0.0028 (0.003) -0.0025 (0.011) -0.0021 (0.033) -0.0000 (0.487) oil2 0.3869 (0.745) 0.0198 (0.622) gdpgrowth -0.1178 (0.001) -0.0970 (0.006) -0.1352 (0.000) -0.0027 (0.022) tradeopen -0.0817 (0.000) -0.0911 (0.000) -0.0933 (0.000) -0.0013 (0.078) agriculture -0.0145 (0.061) -0.3627 (0.001) -0.1577 (0.036) -0.1076 (0.175) -0.0165 (0.000) manufacture 0.8662 (0.000) 0.9042 (0.000) 0.8056 (0.000) 0.8823 (0.000) 0.0148 (0.008) 𝑹𝟐 0.2562 0.2464 0.2121 0.2044 0.1280 Obervations 258 254 263 258 252The regression of oil export on the service sector (Model 3) is conducted in five steps and the results are in Table 4. In the first column, the coefficient of oil export variable is negative at a significant level and it’s not suggested by the theory. This means that if there is an oil boom, there will be contract in the service sector. In the second column, we include the oil export variable in difference. The 𝑅! slightly decreases
and the coefficient of oil export variable is not changing much. In the third and fourth column, we exclude the GDP growth and trade openness variables separately. The 𝑅!
decreases in both cases so it makes sense to include both of them in the model. In the last column, we make regression on the oil export variable in difference, and read that if there is a 1% increase in the crude oil export, the service sector will expand 1.98%. This result goes with the theory, also not at a significant level. To conclude from all the regressions, we may conclude that there is a negative relationship between the oil export and service sector.
4.4 TNT Ratio
Table 5 (1) (2) (3) (4) oil -0.0000 (0.246) -0.0001 (0.108) -0.0001 (0.071) -0.0001 (0.100) oil2 0.0075 (0.865) 0.0294 (0.514) 0.0071 (0.874) gdpgrowth -0.0008 (0.550) 0.0006 (0.625) 0.0008 (0.518) tradeopen 0.0003 (0.700) 0.0008 (0.302) 0.0008 (0.294) 𝑹𝟐 0.0085 0.0182 0.0203 0.0135 observation 258 254 259 254In Table 5 we make regression on the TNT ratio, which is calculated as the sum of agriculture and manufacture sector divided by service sector. From the last three tables we may conclude that there will be ambiguous movement of the impact on TNT ratio after the oil boom. And the conclusion is seen clearly from the results in the table. In the first regression, we can see little relations between oil export and
TNT ratio, not at a significant level though. From the second, third and fourth
regression, we may conclude that oil export in difference and trade openness variable makes more sense to be in the model. However, the explanatory level of the model is still weak, as seen from the 𝑅!. Therefore, we can see that there still isn’t enough
statistical evidence to conclude the movement of TNT ratio after an oil boom.
In conclusion, the results shown in these four tables can be concluded as that: an increase in the crude oil export will leads to contract in the agriculture sector and service sector, but not sure about the impact on the manufacturing sector. This result partially rejects the H0 hypothesis but not fully rejects.
4.5 Discussion and Limitation
Although in the study we tried to make the analysis as reliable and logical as possible, there are still many limitations we are not able to solve at this moment. Here is the section to devote to a discussion on some of these limitations and simplifications.
First, we omitted many of the variables mentioned in the Stijns(2003) and Lartey, Mandelman & Acosta (2008) because it’s not feasible to include all those variables. For example, in the Stijns(2003), he uses many dummy variables, such as Lang as a dummy which is equal to 1 if two countries share a common language, or Cont as a dummy which is queal to 1 if two countries share a land border, etc. But it’s
enormously difficult to check every country on this information within limited time. However, omitting all these control variables makes bias to the results of analysis. Thus it’s better to include as many control variables as we can for the next time regression analysis.
Second, it will better if I use the exact gravity model used in the Stijns(2003) since it was tested credible and simple. However, the model is not feasible for me since the
author is using complicated methods to construct the Dutch Disease variable. The method is out of my ability to construct. Also in the Lartey, Mandelman &Acosta (2008), the model they use is a dynamic panel data model with generalized method of moments estimator. Again this model is more difficult to implement and for me to understand in such a limited time.
Third, endogeneity is a potential problem. For example, in the first model we use agriculture value added as dependent variable but as control variable for the other two models. It is possible that the high correlation between agriculture, manufacture and service causes the problem of endogeneity. This endogeneity problem may potentially cause bias in the results. However, the growth in each three sector may be crucial driver of other sectors, so they cannot be excluded from the analysis.
Fourth, there may be lagging time before the sectors reacts to the oil boom. However, here is this study we omitted this lagging issue for simplification, and may cause bias in the results. Therefore, it is better to use the time lagging factor as in the Lartey, Mandelman & Acosta (2008) to improve the credibility of the model.
5 Conclusion
This thesis paper sets out to determine whether the Dutch Disease was present in major oil exporting countries after a crude oil boom. We made hypotheses based on the core model suggested by Corden and Neary (1982). In order to test the research question we tested 32 major oil-exporting countries based on their percentage of GDP dependent on the oil-exporting sector. We tested countries from 2001 until 2010 on agriculture, manufacturing and service sectors, and use crude oil export as the major explanatory variable. We control for the GDP growth, trade openness and oil export in difference variables. The hypotheses were transformed into four models. We constructed the four models by using fixed effect model with panel data. The models are tested by making regressions in STATA and also by doing sensitivity analysis to test the robustness of the results.
The final conclusion we obtained is that there will be negative impact on agriculture sector and service sector, but no significant proof on the manufacturing sector. Therefore, the hypotheses were partially rejected. This means that the Dutch Disease was not tested to be present in all of the sectors in an economy in our study. With many limitations existed, improvement should be made in the future studies about the models and controlling variables.
References
Corden, W. M., & Neary, J. P. (1982). Booming sector and de-industrialisation in a small open economy. The Economic Journal, 825-848.
Corden, W. M. (1984). Booming sector and Dutch disease economics: survey and consolidation. Oxford Economic Papers, 359-380.
Davis, Graham A. "Learning to love the Dutch disease: Evidence from the mineral economies." World Development 23.10 (1995): 1765-1779.
Fardmanesh, M. (1991). Dutch disease economics and oil syndrome: An empirical study. World Development, 19(6), 711-717.
Lartey, E. K., Mandelman, F., & Acosta, P. A. (2008). Remittances, exchange rate regimes, and the Dutch disease: A panel data analysis.
Sachs, J. D., & Warner, A. M. (2001). The curse of natural resources. European economic review, 45(4), 827-838.
Stijns, J. P. (2003). An empirical test of the Dutch disease hypothesis using a gravity model of trade. Available at SSRN 403041.
Stock, J. H., & Watson, M. W. (2003). Introduction to econometrics (Vol. 104). Boston: Addison Wesley.
Appendix
Appendix 1 CountriesAlgeria Congo, Rep. Iraq Norway Sudan Yemen, Rep.
Angola Colombia Kazakhstan Oman Syrian Arab Republic Azerbaijan Ecuador Kuwait Papua New
Guinea
Trinidad and Tobago Brunei
Darussalam
Egypt Libya Qatar United Arab
Emirates Cameroon Gabon Mexico Russian
Federation
Venezuela
