Resource abundance and
returns to education in
developing countries
- the resource curse revisited
Frans van Mastrigt
Universiteit van Amsterdam
Economie & Bedrijfskunde Bachelor Thesis
1. Introduction
A connotation of the word developing is that the thing that is developing is moving from one stage to the next, higher stage. Developing countries are implicitly supposed to become developed countries one day. The reach these stages of developed, governments have to secure their growth and lead their country forwards. One major advantage some governments have is their access to natural resources. Resource abundance would, logically, lead to higher economic growth because of its relatively ‘easy money’.
However, literature has shown that higher economic growth due to resource abundance is not so ‘logical’. There have been arguments supporting both the ‘resource richness leads to higher economic growth’-side and the other side which is called the resource curse.
Although the effects of the resource curse are not always clear and straightforward, an investigation on other factors that influence economic growth might provide better insight in resource abundance theory. One of the key factors for economic growth that has been highly stressed by development scientists is education. By raising levels of education countries can sustain their economic growth and improve living standards which is especially important for developing countries.
This paper will focus on the effects of resource abundance on returns to education in developing countries. The research question is: is there an influence of resource abundance on returns to education? It is, however, not the aim of this paper to prove whether or not there is significant evidence for any deviation between the returns to education among developing countries. Education is important. There is plenty of evidence supporting this and there is a strong necessity for governments to acknowledge the importance of education. Resource abundance does not change the weight of education. It is, however, helpful to investigate the effects of resource abundance on returns to education because this can explain and clarify certain issues that developing countries are facing. By bringing clarity to the subject
governments and helping institutions can improve their ability to improve education by tackling clearly specified problems.
i. Composition of paper
The paper consists four parts. In the first two parts literature regarding (2) resource abundance and (3) returns to education will be discussed. In part (4) indicators of resource abundance and education will be discussed. In the last part (5) the regression outcomes will be presented and discussed, followed by concluding remarks, a bibliography and the appendix.
In part (2) the resource curse will be elaborated. The effects of the resource curse have both been questioned (Stijns, 2006; Haber and Menaldo, 2010) and supported (Sachs and Warner, 1995; De Medeiros Costa and Moutinho dos Santos, 2013; Sinott & al., 2010). An overview of the effects will be provided and discussed, emphasizing on the Dutch disease, rent-seeking behaviour, policy-making and neglect of education. In the last section of this part resource abundance indicators will be discussed that will be partly used in the regression. The elaboration of resource curse theory will shed a light on the main factors of influence on economic growth. This, together with resource abundance indicators, will provide support for the decisions
made for the regression in part (4).
In part (3) returns to education will be discussed. Returns to education have been studied in human capital theory since the late 1950s (Becker, 1964; Mincer, 1970; Schultz, 1996). The section can be divided into personal returns to education, social returns to education, differences in educational returns and educational investment. Education is in a way more layered than resource curse theory. In a theoretical light education can be regarded as personal investment and social investment. There is an effect and trigger for individuals and for society which must be regarded both jointly and individually. The returns to the investment decision play a key role in both cases. Personal returns will determine the decision of the individual and his decision is of importance to economic growth. Social returns just as important and cannot be left out in the process of creating education policy. The trade-off between social and personal return must be taken into consideration in policy-making and educational investment.
In part (4) the regression analysis will be presented. The Mincer equation will be elaborated and the data that was used will be discussed. The decisions for taking certain measures will be elaborated using literature from parts (2) and (3). The regression of the following countries: Brazil, Guatemala, Nicaragua, Panama and Peru. The results of the regression will be compared to prior research and will be put into perspective. The paper will conclude with a discussion of the results and the most important outcomes. 2. The resource curse
Resource curse theory has been excessively investigated, resulting in ambiguous outcomes. Pioneers in this field of study are Sachs and Warner (1995). The idea of a resource curse came up in the 1950s and 1960s but data was limited. The idea has been supported and questioned in literature. The very simple intuition behind resource curse theory is the idea that greater natural resource wealth is associated with higher GDP per capita. Production of natural resources has been the initial source of nearly all development, it creates a source of foreign exchange reserves, it attracts foreign capital and skills and it creates raw materials for processing and a market for manufactured goods (Mikesell, 1998). The positive relationship between resource abundance and economic growth seems to hold for Latin American countries (Sinott et al., 2010). There are, however, plenty of negative effects. Because of economic and political effects resource abundance has on a country’s economic position and its institutions the negative effects might outgrow the positive ones. This is what is understood by the resource curse: a negative effect of resource richness on economic development.
There are four leading arguments supporting the resource curse: the Dutch disease (i), rent-seeking behaviour (ii), overconfidence and policy-making (iii) and neglect of education (iv). Although the existence of the resource curse is questioned and stressed, there are strong indicators for its existing in many individual cases. Taking the effects from the resource curse into consideration can lead to a more concise explanation for possible educational inefficiency or failing, thereby providing policy-makers with more accurate information to provide for solutions. In section (v) the indicators of resource abundance which will be taken into account for the regression.
i. The Dutch disease
When a country is rich of natural resources it can produce at a low marginal cost in comparison to the cost of production elsewhere. The advantage that the country has leads to an increase in exports and large profits which are generated from the production. The Dutch disease is the economic phenomenon that high exports of natural resources lead to a deterioration of a country’s competitive position, harming other domestic industries. It can be divided into two main effects: exchange rate appreciation (a) and a relative increase in returns to production of natural resources (b).
(a) Because of a rise in exports, the demand for domestic currency increases which leads to an appreciation of the currency and a rise in domestic income. The expensive domestic currency leads to a deteriorated competitive position for other domestic products (Stijns, 2006) and it raises the price of nontradeable goods relative to the price of tradeable goods (Sinott et al., 2010). The rise in both the nominal exchange rate and domestic price inflation results in a rise in the real exchange rate.
The real-exchange rate appreciation reduces relative prices for tradeable goods in relation to nontradeable goods and takes labour and capital away from the tradeable sector to let it flow into the nontradeable sector. Because the nontradeable sector is relatively labour intensive and the tradeable sector is capital intensive, there will be a movement of labour in the direction of the nontradeable sector, raising wages and lowering capital returns. Wages rise in all sectors due to the change in exchange rate (Corden, 1984). As a consequence exports of nonresource tradeables decline and the imports rise. To fight the rise in imports and the fall in exports, governments impose import restrictions and subsidize exports. This leads to even more imbalances by attracting investment to high-cost importing, therewith substituting manufacturing. The total effect of the higher prices for manufactured goods and the rise in wages is a reduction in competitiveness of all tradeables in the export markets, including the exports of natural resources (Mikesell, 1998). (b) The second major effect of the exchange rate appreciation is that it increase the returns of production of the recourse relative to other tradeable good (Sinott et al., 2010). It is
therefore more attractive to invest in the resource industry than in other industries, creating a further imbalance in investments. Moreover, the technological progress is faster in traded sectors than in non-traded sectors (Balassa, 1964). Thus, because technological progress is largely confined to the non-traded goods sector, a decline in the traded goods sector may permanently lower income per capita in comparison to the situation where the shift did not occur (Hahn and Matthews, 1965).
According to Dutch disease theory, a risen exchange rate due to the rise in natural resource exports leads to a decrease in competitiveness in all sectors and relative loss of economic growth because of missed out technological progress in the tradeable sector. Government intervention to protect the domestic economy offsets the attractiveness of the tradeables sector, leading to an imbalance between the tradable and non-tradable sector. The advantage of having resource abundance is offset by the negative effects of the higher domestic currency.
These effects do, however, not necessarily occur. It depends on the economic framework of the country. The shift from the tradeable to the nontradeable sector might be in line with the economic
benefits of more resource exports. A strong rise in exports may be necessary to obtain market share in the resource market and can the benefits from this position can therefore outgrow the negative effects of the appreciated currency in the long run. However, the imbalance that has been created between labour in the tradeables and the nontradeables sector has strong long-run effects too. This is related to the topic of rent-seeking behaviour discussed below.
ii. Rent-seeking behaviour
Natural-resource-rich economies tend to have a relatively high level of rent-seeking behaviour on the part of producers (Gylfason, 2001). Rent seeking refers to all largely unproductive, expropriative activities that bring positive returns to an individual but not to society (Krueger, 1974). It often takes shape as import licenses that are distributed or sold by the government. When quantitative import restrictions are imposed, the licenses must be regarded as a valuable commodity. This is especially the case when imports have become more and more important due to the deterioration of the domestic tradables sector, as discussed in (2.i.a.). Two forms of rent seeking will be discussed below: (a) licenses pro rata and (b) bribery. Although there are possible positive outcomes for rent seeking the negative ones outnumber the positive ones and will therefore be solely discussed.
The first type of rent seeking is found in developing countries and concerns (a) licensing mechanisms for imports of consumer goods. Here, licenses are allocated pro rata in proportion to firms’ capacities. It comes in different forms. If licenses are distributed pro rata, a plant-owner will invest in extra capacity if the gains from the increased import divided by the costs of the investment are equal to return on investment in other activities. If this is the case a plant-owner will invest in extra capacity and will thereby obtain extra licenses. Thus, the plant-owner uses his resources not for direct competition in his own market but for competition in the license market. This leads to market inefficiency due to the pro rata licensing. With another form of license distribution the loss of efficiency is caused by market entry competition. Entry generally is free into importing-wholesaling, which results in a larger-than-optimal number of firms. Although these firms still make profits that satisfy their needs, each importer-wholesaler receives fewer imports than would be the case in absence of licensing. The rate of efficiency would be higher if the number of firms would decrease, like in the case of licensing absence, and firm size were optimal. In the case of resource-rich developing countries rent seeking can be problematic because of the high incentives for obtaining licenses and the consequent loss of efficiency.
The second type concerns (b) bribery and the hiring and employing of relatives of officials. In this case efficiency is lowered because of an irrational selection process. Rent seeking may also breed
corruption in business and government, thereby distorting the allocation of resources and reducing both economic efficiency and social equity. Empirical evidence suggests that import protection and corruption both tend to impede economic growth (Bardhan, 1997). Bribery is especially problematic when there are high incentives within a sector to obtain rights, positions and thereby power. In resource-rich developing countries this is the case in the resources industry. The high profitability makes it attractive to attain high
positions in the industry which can be achieved by using the grey area of bribery.
In general, rent-seeking behaviour leads to production inefficiency and an imbalance of capital distribution. Government policy, as discussed in (i), has great impact on the incentives within the market. In countries where institutions and industries are not high-developed meddling is problematic in the case of conflicts of interest of policy-makers and industry executives.
iii. Overconfidence and policy-making
Natural resource abundance may give people a false sense of security and governments tend to
insufficiently focus on adequate economic management which includes free trade, bureaucratic efficiency and institutional quality (Sachs and Warner, 1999). This argument that can be called overconfidence is of great importance to educational development. The pseudo-confidence that seems to be provided by the economic prosperity caused by resource abundance does not necessarily lead to an improvement of institutional quality. As will be discussed in part (3) education does. By having a false sense of security the incentive to develop and invest in education seems to decrease. In judging the effects of economic growth in developing countries the sustainability of the growth has to be taken into account seriously, especially in the case of resource-abundant countries.
There have been several studies that state that institutional performance and quality are the most important transmission mechanism of natural resource abundance (Oskenbayev et al., 2013). Knack and Keefer (1995), Hall and Jones (1999) and Acemoglu et al. (2002) show that institutional quality is a crucial determinant of growth. Capital accumulation, productivity and therefore output per worker are
determined by government policy and the institutional framework, which is usually called social infrastructure (Hall and Jones, 1999). The institutional quality mechanism can be divided into three channels: rentier effects, delayed modernization and entrenched quality (Isham et al., 2005). Rentier effects are the most influential group on institutional quality and will therefore be solely discussed below.
Rentier effects are the persistence of elites and therewith economic institutions that have an impact on economic growth (Acemoglu and Robinson, 2007). Economic institutions tend to survive political power changes because of investments by the elite. The balance of power over economic and political institutions, between citizens and elites, determines economic growth. This balance is based on incentives (Oskenbayev et al., 2013), which result in two kinds of institutions: ‘grabber’ friendly and producer friendly (Mehlum et al., 2006). The ‘grabber’ friendly institutions have a negative effect on economic growth, whereas the producer friendly institutions provide positive economic outcomes. In addition, the relationship between natural resources and institutional performance suggests that there is a minimum level of natural resources that will lead to poor performance and therewith low economic activity (Couttenier, 2009). Evidently, Acemoglu concludes that by extending fiscal tools and instruments, resource misallocation could be avoided (2007). Furthermore, rent-seeking activities along with corruption lead to a failing of establishing positive macroeconomic policies (Auty, 2001). Incentives to create wealth tend to become too blunted by the ability to extract wealth from the soil or the sea (Gylfason, 2001).
The often fast and sudden growth that is caused by the discovery of natural resources does not lead to a change in economic powers. Hereby, policy-makers are subject to inefficient incentives that are not optimally beneficial for a country’s economic growth. The political situation in a country is of strongly influential to the development of industries and economic growth. It is also worth mentioning that geographic variables influence economic growth via institutional quality and performance (Easterly and Levine, 2003; Rodrick et al., 2004).
Institutional quality is of great importance to economic growth. The role of policy-makers is crucial in the improvement of institutional quality. Because natural resource abundance leads to easily acquired income, a focus to invest in the institutional framework diminishes in comparison to resource-poor countries. Ideally, conflicts of interest should be eradicated, leading a country to higher productivity and economic growth and, eventually, more sustainable growth.
iv. Neglect of education
Resource wealth tends to take away a focus on effective investment in institutions, human resources and education (Gylfason, 2001). Another, quite obvious, problem with resource abundance is that diamonds may last forever, but resources do not. At some point, once resource-rich economies have to make a transition into another type of economy. The investment in education is crucial for this purpose.
The level of school attainment at all levels is inversely related to natural resource abundance (Gylfason et al., 1999). The focus on the primary production sector causes little incentives to invest in education. Secondary school attainment in OPEC countries in 1997 is 57%, whereas the world average is 64%. The amount of money spent on education is 4% of GDP, whereas the world as a whole has an average of 5% (in 1997). The point being made here is that resource-rich countries have a decreased interest in developing their educational levels despite the need for it.
The relatively low levels of investment are not a consequence of impossibility to invest in education, but rather the failure of public authorities. Countries like Norway and Botswana have shown, with high levels of educational investment as a proportion of GDP, that it is not impossible to introduce effective educational policy despite their resource abundance. Norway’s total exports of goods and services are no larger in proportion to national income than they were before the oil fields were
discovered in the North Sea (Gylfason, 1999). This shows it is not a matter of possibility but a matter of policy to ensure educational development.
It is obvious that education is good for a country’s development and economic growth. Nonetheless there seems to be a great imbalance between investment and the need for education. This problem will be elaborated in part (3) that focuses on returns to education and therewith the incentives to invest in it.
3. Returns to education
There is little doubt about the general need for education. Sen (1999) states that education contributes to a more desirable civic society and is valuable in its own right. Taking this statement as an absolute truth, there seems no need to calculate the returns to education. Unfortunately, education’s right in itself is not always being respected, creating a need for showing what education brings, can bring and must bring to society. This is especially relevant for developing countries, where policy-makers need to be convinced of the importance of education (Hanushek, 2013).
Returns to education can be divided in returns to the individual and returns to society. Both are important indicators for economic growth. Educational policy determines the extent to which it is desirable to personally invest in education. The quality of education is subject to policy as well. In the first section (i) personal returns to education will be discussed. In the second section (ii) the social returns to education will be discussed. Section three provides comparative results The last section (iv) focuses on educational investment and will provide connections to policy-making.
i. Personal returns to education
Classical human capital theory is built around the ideas of Mincer (1970, 1974), Becker (1964) and Schultz (1961). Educational investment is seen as a personal decision that depends on increased earnings due to extra years of schooling and the cost of education. The estimated return is obtained as the coefficient on a years of education variable in a log wage equation that contains controls for work experience and other individual characteristics. In this semi-log earnings function the coefficient on years of schooling can be interpreted as the average private rate of return to one additional year of education, regardless of the educational level to which this year of schooling refers. Hanushek (2008, 2010, 2013) stresses the importance of a more accurate measure of education, advocating a qualitative measure instead of the quantitative years of schooling measure. He states that cognitive skills of the population are
powerfully related to individual earnings, to the distribution of income and most importantly to economic growth, where cognitive skills are not merely school attainment but the whole of skills and human capital of an individual. Cognitive skills measures, however, are hard to come by and years of schooling seem to be an appropriate measure (Psacharopoulos, 1994, 2004; Harmon et al., 2003). In section (a) the Mincer function will be discussed. Section (b) provides possible supplements to the function.
In Mincer’s function (a) years of schooling are considered the most important influence on personal income. In human capital theory education is considered an investment of current resources. The result of this investment will be the future returns minus the opportunity cost of the time involved and direct costs such as tuition fees, teaching materials and travelling expenses. The question people implicitly ask themselves is: what are the returns to taking one extra year of schooling, given the opportunity costs, e.g. working. The theory assumes that individuals seek to maximise their expected present value of the stream of future incomes minus the costs of education (Harmon et al., 2003). This results in the following formula:
�𝑤𝑤(1 + 𝑟𝑟𝑠𝑠− 𝑤𝑤𝑠𝑠−1
𝑠𝑠)2 𝑇𝑇−𝑠𝑠
𝑡𝑡=1
= 𝑤𝑤𝑠𝑠−1+ 𝑐𝑐𝑠𝑠
where 𝑠𝑠 is years of schooling, 𝑟𝑟𝑠𝑠 is the internal rate of return, 𝑤𝑤𝑠𝑠 the stream of future incomes up to
retirement at date T and 𝑐𝑐𝑠𝑠 are the net costs of education. By increasing the direct education costs, the net
benefits of undertaking an extra year of schooling will be lowered. 𝑠𝑠 is chosen in that way that, at optimum, the benefits of the 𝑠𝑠𝑡𝑡ℎ year of schooling are equal to the costs of undertaking the investment.
The optimal investment decision implies that one would invest in the 𝑠𝑠𝑡𝑡ℎ year of schooling if the return
to the extra year of schooling is bigger than the market interest rate.
If T is large the return to education can be approximated with the following equilibrium: 𝑟𝑟𝑠𝑠≈ 𝑤𝑤𝑠𝑠𝑤𝑤− 𝑤𝑤𝑠𝑠−1
𝑠𝑠−1 ≈ log 𝑤𝑤𝑠𝑠− log 𝑤𝑤𝑠𝑠−1
log 𝑤𝑤𝑖𝑖 = 𝑋𝑋𝑖𝑖𝛽𝛽 + 𝑟𝑟𝑠𝑠𝑖𝑖+ 𝛿𝛿𝑥𝑥𝑖𝑖+ 𝛾𝛾𝑥𝑥𝑖𝑖2+ 𝑢𝑢𝑖𝑖
where 𝑤𝑤𝑖𝑖 is a measure for individual i’s earnings, e.g. earnings per hour or week. 𝑠𝑠𝑖𝑖 represents a measure
of schooling, 𝑥𝑥𝑖𝑖 is an experience measure and 𝑋𝑋𝑖𝑖 is a set of variables that represent assumed influences on
earnings. 𝑢𝑢𝑖𝑖 is a disturbance term, representing measures that are not correlated to 𝑋𝑋𝑖𝑖 and 𝑠𝑠𝑖𝑖, but are not
explicitly measured. Consequently, at the optimal schooling level the marginal rate of return of an
additional year of schooling is equal to the marginal cost of this additional year of schooling (Card, 1999). For Mincer’s formula to be clear, a number of implications must be pointed out. Firstly,
individuals are considered rational beings that seem to make a decision that follows a classical risk-return function and that the set of opportunities they face functions as a market (Harmon et al., 2003). In other words, schooling is considered an investment decision. Secondly, the internal rate of return is used and discounted to derive the present value of the investment benefits to the present value of costs. This implies as well that a person who prefers current income to future income has a different internal rate of return than someone who is indifferent to this time value. The individual who prefers present income would have a higher discount rate, so future income will have a lower present value. As a consequence, this type of individual is less likely to undertake education. Thirdly, if the probability of being employed is higher when education is undertaken a decrease in this benefit would lower the reward for undertaking education (Harmon et al., 2003). Fourthly, more schooling may increase the likelihood of receiving further work related training. Therewith, cognitive skills would increase as a secondary effect of undertaking education (Blundell et al., 1996). Fifthly and finally, by taking individual income as the most important feature for the education decision, cultural benefits such as status, having a highly skilled job, are not taken into account although they might influence the investment decision and are not reflected in wages only (Chevalier and Lydon, 2001).
seen as inherent to an individual, being a complementary factor to education in producing human capital. More ability would thus imply more human capital produced in one year of extra schooling. At the same time, one might regard ability as having higher opportunity costs, because possible earnings in different investment might be higher too (Harmon et al., 2003). As mentioned, individual preferences to future and current earnings can differ due to for example a variation in taste for schooling or access to funds (Lang, 1993). However, it is very well possible that lower discount rates are correlated to ability. High-ability parents, who might very well be wealthy, will be more capable of providing their children sufficient funds and resources for education. This could lead to an endogeneity bias. People with relatively low marginal returns will choose higher levels of schooling (Harmon et al., 2003). Measuring ability is hard and there is little data availability on it. Leaving out ability could lead to less accurate outcomes. However, when testing returns to education in developing countries the overall returns to education are subject and need not necessarily be adjusted for ability.
ii. Social returns to education
Social returns to education are the returns that concern society and are not included in personal returns. For policy-makers, the additional return to education for society is of importance, rather than increased personal returns (Fulford, 2014). The magnitude of the social return to education is important for assessing the efficiency of public investment in education (Moretti, 2004). Another effect that concerns policy-makers is the fact that the variables that are used in the Mincer equation are subject to government policy. By decreasing tuition fees, governments open their universities to a less wealthy part of society, shifting their returns and increasing the attractiveness of high-skilled jobs.
The social returns to an educational investment indicate the desirability of this investment to society. Both literature and data on the social returns to education are not as extensive and of the same quality as those on personal returns (Temple, 2001; Cohen and Soto, 2007). As a consequence, there are no clear estimates for social returns. Yet, it is important not to trivialise social returns because of their relationship with policy-making and their influence on economic growth.
If personal educational investment is suboptimal the incentive for governments to intervene increases. This means that if investments are not at the desired level, given personal and governmental preferences, an incentive to increase investment arises. Capital externalities, government subsidies, taxes and labour market institutions are the main external force for narrowing the gap between private and social returns (Venniker, 2000). This part is divided into three sections that focus on externalities. Externalities are the factors that influence the desirability to invest in education and can be divided into three categories: (a) static externalities, the effect of human capital externalities on current productivity, (b) dynamic externalities, the effect of human capital externalities on learning and technological change and (c) externalities that are related to non-pecuniary effects of human capital.
The most important static human capital externality (a) is that human capital of an individual enhances the productivity of others, e.g. physical capital and human capital of others (Lucas, 1988). This
overtaking of productivity occurs through channels that are not internalised by individual families and firms, most importantly human interaction within cities. If productivity in a city rises due to risen educational levels citizens who have not enjoyed this education will improve their productivity as well. The effect of increased education rises in this way.
There is no strong evidence for these positive externalities, but no evidence against their existing either. Benhabib and Spiegel (1994) find evidence against the existence of externalities, whereas Heckman and Klenow (1997) and Acemoglu and Angrist (1999) find evidence in favour of the externalities. Rauch (1993) finds no evidence of an existing causal relationship. There are no obvious indicators for
externalities which makes any consensus about the subject very hard to achieve, leaving the empirical evidence for significant positive human capital externalities in production inconclusive (Venniker, 2000).
Dynamic externalities (b) can be divided in three categories. Firstly, the creating and adopting of new technologies: the higher the level of human capital the more effective the creation and adoption of new technologies is. Secondly, learning-by-doing is more effective with higher average human capital. Thirdly, human capital accumulation is more effective within groups of the same level of prior human capital. This argument is not directly related to the increasing of human capital. In fact, it says something about the composition of human capital, not about the average level.
Research on dynamic externalities provides more univocal evidence than evidence on static externalities. Nelson and Phelbs (1966) Benhabib and Spiegel (1994) and Romer (1990) all emphasise the role of human capital stock in creating new technologies. Krueger and Lindahl (2000) point out the importance of two underlying assumptions: firstly, all countries have the same relationship between initial education and growth. Secondly, this relationship is to be linear. They argue that the positive coefficient is not necessarily representative for the existence of externalities when more plausible assumptions
employed. It can carefully be concluded that dynamic externalities cause higher efficiency of capital accumulation due to spill-over effects from the environment. If the average level of human capital rises, not only does the level of those directly concerned rise, it rises the levels of those not-directly concerned as well.
Non-pecuniary effects are effects that are not directly measurable in terms of money or human capital measure. There is positive evidence that education evokes non-marketed effects of education (c) such as intra-family productivity, childcare, family healthcare, teenage pregnancy, crime reduction, social cohesion and charitable giving (Haveman and Wolfe, 1984; Lochner, 1999), but existing estimates on their economic value lack in the work of Haveman and Wolfe. Lochner finds social returns of 20 percent of private return. Although evidence is ambiguous the non-pecuniary effects must be taken into
consideration. Economic growth and the improvement of living standards is not only caught in numbers of income. The overall quality of life is strongly represented in the non-pecuniary effects mentioned above.
The evidence on social externalities is ambiguous and data on externalities are hard to come by. Nonetheless, externalities do account for returns to society. It is up to local governments to take the
effects into consideration, attach the right conclusions and use them for policy creation and implementation.
iii. Educational investment
Educational investment must depend on both the personal and the social returns to the investment. There are two ways of measuring the value of education, using either the input method or the output method. The input method takes the resources that are committed to education by families, students and state and measures therewith the input to education. The output method measures what is produced by education, such as comparatively high living standards. By comparing the two methods, measures of efficiency in the use of resources can be found. Intervention by the state as a tax collector or financier of education influences the distribution of education resources (Le et al., 2003).
By comparing the costs and benefits of education on a personal level, the private rate of return to investment in education can be calculated. On the social level, however, states invest in education by public subsidisation and wish for returns in a higher productive workforce. Social benefits can be measured as pre-tax earnings differentials to reflect the full productivity potential of the workers by level of education. By comparing costs and benefits one can calculate the social rate of return to investment in education. GDP growth could be set off against the costs of education as a measure. However, calculating social returns is subject to two factors that make it hard to rightly adjust the influence of education. Firstly, earnings are not a very concise proxy for personal productivity. Secondly, education may be associated with non-market or external benefits that are not easily determined, such as reduced criminality and a reduction in the spread of infectious diseases (Psacharopoulos, 2006).
iv. Differences in educational returns
Knowing the returns to education, both at the personal and social level, provides a field for policy-making to consider the best implementation for economic growth. Psacharopoulos (1994) shows that Latin-American countries have a social return of 17.9% to primary education, 12.8% to secondary education and 12.3% to tertiary education. The returns on the personal level are 26.2% to primary education, 16.8% to secondary education and 19.7% to tertiary education. OECD countries have a social return of 8.7% to tertiary education, which is almost equal to long-term opportunity cost of capital, indicating that the profitability for human and physical capital is in virtual equilibrium. Psacharopoulos finds an average for years of schooling for Latin-American countries of 7.9 years with an average return to education of 12.4%. He makes several remarks about his findings: (a) countries show diminishing returns as income per capita rises; (b) countries have declining returns over time; (c) females show higher returns than males; (d) faculties have different returns amongst each other and (e) returns are dependent of the sector of employment.
For all countries, returns to primary education are highest. Returns diminish as the level of education rises due to subsidisation of education, resulting in relatively high private returns. As income per capita rises (a) social and private returns at all levels diminish. It is another reflection of the law of
diminishing returns to the formation of human capital at the margin (see: table 3.1.). As time passes (b) returns to education decline by between 2-8 percentage points in average on a 15 year period. Notably, higher education returns diminish too, by approximately 2 percentage points, indicating that college graduates have increased their accessibility to public funds. The returns to female education (c) are about 1 percentage point higher than those for males on the whole picture. Although there is a concern literature of female returns to education relative to male returns (Heckman, 1979), no influence is found by Psacharopoulos and Tzannato (1992a, 1992b) for the selection bias which, meaning that women make a decision whether to participate in the labour market whereas men do not. The faculty of tertiary education influences (d) the rate of return. The lowest social returns can be found in physics, sciences and
agronomy, while engineering, law and economics show the highest private returns. Returns to education are higher (e) in the private and competitive sector than in the public and non-competitive sector. Returns in the self-employment sector of the economy are lower than returns in the dependent employment sector. The labour market duality is difficult to identify due to differences in low-pay and high-pay sectors and a scarcity in data.
(table 3.1. Coefficient of schooling: Mincerian mean rate of return, Psacharopoulos 1994)
4. Indicators of resource abundance and education
In this part possible indicators for resource abundance (i) and education (ii) will be discussed. Resource abundance indicators will be discussed at a national level whereas educational indicators are discussed both at the national as at the personal level.
i. Resource abundance indicators
Resource abundance (a) can be measured in different ways. There are different sources of resource income that can be divided to obtain a good estimator. The World Bank provides information about natural capital wealth which can be used to create ratios for resource abundance. Stijns (2006) mentions ten estimators that are commonly used in literature. The first one to define resource abundance is the share of natural capital used by Gylfason (2001). To calculate the share of natural capital in national wealth, natural capital is divided by the sum of natural capital, physical capital and human resources which, combined, is the total capital ratio. Physical capital is a perpetual inventory value of produced assets. The World Bank provides information on natural resource wealth, a part of which is used to estimate resource abundance. The share of natural capital as a part of total capital gives a percentage of resource abundance that is comparable among countries. However, because the human capital in this ratio hard to define, the
Low income ($610 or less) 301 6.4 11.2
Lower middle income (to $2,449) 1383 8.4 11.7
Upper middle income (to $7,619) 4522 9.9 7.8
High income ($7,620 or more) 13699 10.9 6.6
World
estimator is difficult to use, especially in developing countries. As a solution, Stijns proposes several alternative estimators that serve as an indicator for resource richness: three natural capital ratios, three indicators that measure resources per capita, three export indicators and an indicator concerning resource rent intensity.
The first three indicators are natural capital as share of physical capital, subsoil wealth as share of physical capital and green capital as share of physical capital. By taking these ratios, the measurement problem of human capital is discarded and resource abundance is regarded in respect to different sorts of sources. The latter is meant to distinguish between minerals and fuels, subsoil wealth, and green capital. The next three indicators are subsoil wealth per capita, arable land per capita and resource rent per capita. Subsoil wealth per capita is a more precise measure than the physical capital ratios, because the effects of economic development in, for example, the mineral sector will not be wrongly ascribed to higher human capital accumulation (Stijns, 2006). Arable land per capita is used by Auty (2001) and Birdsall and al. (2001), but the indicator is plausibly an inexact measure of resource richness because it does not concern all types of resources. Sachs and Warner (1995) use primary export intensity as resource abundance indicator, i.e. all natural resources exports as a share of total GDP. Stijns (2005) criticizes this measure for its emphasis on intensity instead of mere resource abundance. Agricultural and mineral export intensity have the same emphasis, but distinguish between the two sectors, making it possible to indicate more rightly the specific parts of primary export intensity. The World Bank (1997), Stijns (2006) and Hamilton and Clemens (1999) support resource rent intensity as an estimator. The ratio is obtained by taking the world price of natural resources and subtracting country-specific costs of extracting the recourse. Dividing this by current price GDP gives resource rent intensity. Resource rent per capita can be supported by the same arguments as resource natural wealth per capita and resource rent intensity.
ii. Educational level indicators
Years of schooling can be considered a concise indicator for the level of education. However, there are more indicators that could give a more complete image of the level of education, because the level of education achieved with one year of schooling in Nicaragua can and will differ from one year of schooling in Ecuador or for example The Netherlands (Hanushek, 2008). The indicators vary on level and are either personal or societal. Firstly, beside years of schooling, secondary enrolment can be used
(Mankiw et al.l, 1992; Birdsall et al., 2001; Gylfason, 2001) to give a broad overview of the average level of education. Secondly, Davis (1995) and Birdsall et al. (2001) use the adult literacy rate. Literacy on a personal level could serve as a measure of skill to calculate income. In developing countries the
distribution of human capital is a concern (Stijns, 2006) and the adult literacy rate could shed some light on possible returns to education. Thirdly, life expectancy at birth is used by Davis (1995). Health being an important indicator of one’s working efficiency, the life expectancy at birth can function as an indicator of productivity. Fifthly, Gylfason (2001) uses public expenditure on education as a percentage of aggregate expenditure. This, however, does not cover the need for public education in society or private
expenditures on education (Stijns, 2006).
Psacharopoulos (1994) mentions the lack of usage of the different levels of education in existing research. He states that using educational level dummies could describe the behaviour of wage effects in relationship to education. Authors who do use these dummies often describe the effects as returns to investment, meant as the personal investment in education, although they really are wage effects and not rates. The returns notion necessitates taking into account the cost of education, whether private or social, and relating this cost to the wage effect. As a second notion, Psacharopoulos mentions the difference between primary education and the other levels. Primary school children do not forego earnings because of their studying. The assignment of foregone earnings is therefore unjust. Consequent estimates will grossly underestimate the returns to education due to the unjust discounting of primary school years. 5. Methodology and regression analysis
The purpose of this paper is to investigate the effect of resource abundance on returns to education in developing countries. Methodology decisions and issues will be presented in this part and the regressions outcomes will be analysed. In the first section (i) the research method will be presented. The decisions for variable selection and data quality will be presented in the second section (ii). In the third section (iii) the regression outcomes will be discussed and compared to existing literature. In the fourth sector (iv) conclusions will be drawn and put into perspective for policy-making.
i. Research method
In human capital theory the Mincer equation is most widely used to calculate returns to education. In the model, the natural logarithm of earnings is modelled as a function of years of schooling and potential experience. In the most commonly used version of Mincer’s function, log earnings are modelled as the sum of a linear function of years of schooling, potential experience and a quadratic function of potential experience:
ln 𝑦𝑦 = ln 𝑦𝑦0+ 𝑟𝑟𝑟𝑟 + 𝛽𝛽1𝑋𝑋 + 𝛽𝛽2𝑋𝑋² + 𝐷𝐷 + 𝑢𝑢
where 𝑦𝑦 is earnings, 𝑦𝑦0 is earnings of someone with no education at all, 𝑟𝑟 is years of schooling, 𝑋𝑋 is
potential experience, 𝐷𝐷 are dummies and 𝑢𝑢 is an error term. 𝑟𝑟 represents the rate or return to an additional year of schooling. The dummies are gender, age² and potential experience because of their proven effects on earnings. In the 1950s, earnings were seen as a function of age, although it was known that the higher educated had a steeper earnings curve than the less educated. By using potential
experience, both the age-earnings profile and differential slope of the age earnings profile across education groups are caught. Another advantage of the model is that there is a single labour market return, 𝑟𝑟, conditional on years of potential experience (Lemieux, 2003).
Returns on education will be measured in three different ways. The first set of regressions will provide information and educational returns with all available forms of education. Because the data
collection differs between the datasets and because of differences in their educational institutions available education variables differ in number from four to eleven. In all the regressions the dummy variable for no education has been left out and is implicitly captured by the constant. All other variables are therefore relative to one without education. The regression outcomes using this method were best at explaining the overall composition of earnings.
The results, however, are hardly comparable. The difference in educational systems and data collection between, for example, Brazil and Nicaragua leads to four variables for secondary education on Brazil, whereas Nicaragua has only one. The different dummy variables for the same level of education, secondary education in this case, have therefore been substituted by one new dummy variable that has the value of one if one of the sublevels of education has been finished. By doing this three new categories were created: secondary, tertiary and other. To carry out comparable regressions these dummies have been used instead of the original ones. This led to the second set of regressions that has been used in the final tests. The explanatory quality of these regressions is only slightly lower that the quality of the first set of regressions.
A third set of regressions has been run using only the summarised dummies from primary, secondary and tertiary education plus gender, age² and potential experience, thus, leaving out other education. The quality of this set of regressions was significantly lower, which led to the conclusion that neglect of the other education dummy does not raise the quality of the outcomes. This test will therefore not be used or discussed.
After the returns to education have been calculated for each of the countries involved the results will be compared. The outcomes of these tests will not provide reliable information on the effects of resource abundance on returns to education. The test will provide insight into the different returns between countries. If there is any difference in returns to education, which there is, it may be concluded that there is a difference in returns, but we cannot conclude that this difference is caused by resource abundance. This problem is caused by the size of the test. The test involves five countries which is relatively small and cannot provide reliable information. The influence of resource abundance will
consequently be calculated but here again the sample size is problematic. The outcome can only be used as a guideline and is not reliable on itself.
To compare the outcomes between the countries a t-test is used. The subsequent t-values are checked and if they are significant a test is followed controlling for resource abundance.
ii. Data quality and variable selection
In this section data selection and quality (a) will be discussed. Consequently, (b) the selection of variables is divided into resource abundance and education indicators.
The datasets (a) that are used are downloaded from the World Bank LSMS database and represent household surveys of Brazil, Guatemala, Nicaragua, Panama and Peru over years between 1990 and 2000.
provides sufficient overall quality of the surveys and judges it as high-quality studies. This statement concerns the way the surveys were carries out. Nonetheless, several tests had data missing. This is due to the fact that the surveys are assigned to measure all available household information and does not specifically specialise in the topics this paper is concerned with. For this reason regressions on Ecuador and Peru (1994) could not be carried out. In the Peru (1994) dataset income is included as a variable, but lacks information, e.g. there are no existing income values. The Ecuador datasets (1994; 1995; 1998) were accessible, but there were not usable for the software that has been used for the regressions.
Another problem was the juridical accessibility of the data. For the data of Jamaica and Guyana legal permission had to be granted after an application for the data. Due to the time limitation of this paper regressions on these countries could not be included.
Not all countries are examined with the same frequency. Some countries only have one dataset available for the past 25 years, whereas others have been measuring their living standards every year. Dependent of data-availability, the datasets that have been used are chosen so that they are centred on the same years as much as possible. The influence of economic trends worldwide is major. Although returns to education will not increase or decrease very rapidly, exports and imports are strongly dependent on worldwide trends. Economic trends will therefore have a great impact on the indicator that is used for resource abundance. Due to the fact that the datasets vary in a range of ten years (1991-2000) the comparative test on resource abundance is not likely to be very reliable.
The countries that are included are all Latin-American countries that are all regarded as developing countries according to IMF standards. The geographical equivalence should make the regressions more concise because of economic and cultural similarities. Nonetheless, the countries are in no way the same. They all have their own economic and educational structures and face their own political opportunities and difficulties. These characteristics are crucial for a precise explanation of the results. Unfortunately, that goal exceeds the purpose of this paper and will be left for further investigation. The outcomes of this paper will serve as a clear indication that education is needed in these countries, independent of their specific characteristics.
The dependent variable (b) in the regression is earnings. The measurement of earnings differed between countries. Although some datasets, i.e. Brazil and Guatemala, showed earnings per month, other countries showed earnings per week, i.e. Nicaragua. This did not lead to any problems because the earnings could be all adjusted to earnings per month. Furthermore, the log of earnings was used in the regressions, which measures the growth in earnings dependent of the given variables. If the data within one dataset is consistent, i.e. all earnings are earnings per month, the log earnings does not differ between monthly returns or weekly returns. Nonetheless have the data been adjusted to one way of measuring earnings. However, there are problems involved with taking a monthly returns and adjusting returns to monthly returns. The influence of seasonal work was not included in the data. It could therefore well be that earnings may have differed if the data were collected in a different time of year. Another remark that must be made is that earnings are the earnings of paid labour. Earnings of entrepreneurs have not be taken into account which may lead to an inconsistent representation of reality. This is especially the case in
developing countries with a high number of small farmers who work independently. There was no possibility of correcting the data for this.
The most important independent variable is highest level of education completed, represented by dummies, depending on the information provided by the different datasets. Most common variables were no education, primary, secondary, higher education and other. Other variables included the different forms of secondary, higher and other education, i.e. non-university higher education and adult education. To overcome the differences between countries substituting dummies have been created that summarise one level of education, using the five different levels mentioned. In the dummy other all education was put that was not primary, secondary or higher education. No education was used as base. This usually captured levels such as no education at all, nursery school or pre-elementary school. The results are, as mentioned, relative to this level of education.
Other variables include age squared, gender and potential experience. Literature shows that age squared has a significant impact on the estimated return to education and is therefore included. Gender is included because returns to women tend to be significantly higher, especially in developing countries, because non-participation in the labour market is more common among women than among men. Including gender could reduce a bias in the returns to female education. Ideally, a direct measure of experience would be included, but this is not provided by the datasets used. Literature shows ambiguous findings on the way of measurement of experience. Kjellstrom and Björklund (2001) state that the way potential experience enters the model does not have severe impact on the quality of the schooling coefficient. Heckman, Lochner and Todd (2001) show opposite findings and that the measurement error will bias the estimated return downwards. Because of an absence of direct experience, potential experience is measured as age minus 6, implicating that one has the potential of earning money while not studying from age 6 on, following the descriptive suggestions by Lemieux (2003). Another possible way of including potential experience is by taking current age minus age of education completion, but in the datasets provided this could not be calculated accurately. Including other changes in specification generally does not lead to significant changes in the estimated return to education (Harmon, 2003).
The choice of indicators for resource abundance is based on direct applicability of the indicator. Resource rent intensity, used by i.e. The World Bank, is used as indicator for resource abundance. It does not make any distinction between the nature of resource income which makes it easy in comparing countries. It does, however, distinguish between the costs and prices of the resources, making it more precise than the primary export intensity indicator used by Sachs and Warner (1995). For education highest level of education completed is used.
The choice of an indicator for education was less ideal, but is supported by the theories of Psacharopoulos (1994) and Hanushek (2008, 2010, 2013). The difficulty with this indicator is that
educational levels differ between countries, not only in a qualitative way, but in a quantitative way too. The age of finishing secondary school, for example, can differ between countries, what would lead to different years of schooling, the variable that is used in the Mincer equation. This problem is, however,
years of schooling are purely quantitative. One additional year of schooling on the university level is equal to an additional year of schooling on the secondary level when using years of schooling. When using the highest level of education completed a distinction is being made between the different levels of education and their returns. The qualitative difference between countries remains as would have been the case when using years of schooling as a explanatory variable. One last problem is the comparability of the results. As years of schooling are usually used in human capital theory, most articles provide information only on this type of measurement. Information on returns to level of education is hard to come by. This makes it difficult to put the results in line with existing literature. Recalculating years of schooling by assigning a fixed number of years for each of the educational levels would lead to major inconsistency because educational systems have changed over time and the time used for each level by an individual may differ strongly. The results will therefore be compared where possible, keeping in mind the research method differences.
iii. Regression analysis and comparison
Coefficients Brazil Guatemala Nicaragua Panama Peru
R² 0,3698 0,3994 0,1918 0,3936 0,1571 PRIM -1,026 0,399 0,764 -0,240 0,825 SEC -0,261 1,054 1,660 0,213 1,266 HIGH 0,865 1,677 2,738 1,209 2,061 OTHER -0,353 0,413 2,084 0,474 2,985 GEND -0,545 -0,443 -0,228 -0,385 -0,615 POTEXP -0,001 -0,001 0,022 0,123 -0,001 AGE2 0,118 0,085 0,000 -0,001 0,069 CONST 5,076 5,558 3,978 3,632 1,901
Brazil Guatemala Nicaragua Panama Peru
Returns NO EDUC 5,076478 5,55798 5,076478 3,631751 1,900748 PRIM 4,050201 5,9568086 4,050201 3,39181 2,725392 SEC 4,8156292 6,611742 4,8156292 3,844606 3,166928 HIGH 5,9415172 7,235083 5,9415172 4,840632 3,961581 OTHER 4,7237094 5,9709872 4,7237094 4,105884 4,885675
(table 5.1: parameters and total returns to education) To calculate returns to education the Mincer model was used as described above, including dependent variable log earnings, independent variables highest level of education completed, represented as dummy variables for each level, potential experience, age squared and gender as a dummy variable. The results of the first two sets are included in the appendix, i.e. the results of a regression including all education dummies and those of a regression including only combined dummies. The results of these two sets are
significantly similar. The outcomes of the second set were used to calculate differences in returns between countries. Where needed these outcomes will be used to further explain any abnormalities.
There are a couple of trends showing in the results. Firstly, the returns to education seem to be upward-sloping as the level of education increases. Secondly, some of the returns are negative, implying that having an education leads to lower earnings than having no education. Thirdly, returns to female education are relatively higher than returns to male education. Fourthly, potential experience has some negative values, meaning that having experience leads to lower earnings than having none. Fifthly, the overall returns are relatively low compared to literature. Having completed higher education in Guatemala leads to a return of 7,23%, which is well below the average returns in developing countries.
(figure 5.1: % returns to educational levels) The coefficients show the return to education to an individual with no education. This might seem confusing given the term return to education. However, it simply means that a person with no education has earnings too. This may be caused by market preferences, but it could be caused as well by the fact that the explanatory value of the tests is relatively low, ranging from 0,16 to 0,40. If more explanatory variables were included the constant might be upward or downward corrected.
An upward-sloping return to education is in line with literature and intuition. The more education you have, the more well-paid your job will be. However, the second finding, negative returns to primary education in Brazil and Panama, does not support this. This could be due to the composition of the workforce and education inequality. If the demand for unschooled labour is relatively high and the distribution of education is concentrated on only one part of society, a higher level of education does not necessarily guarantee a better paid job. Another explanation might be that it is unlikely to have only primary education in comparison to having either no education or a higher level of education than the primary level.
Returns to education are higher for women than for men. The parameter for gender shows 0 1 2 3 4 5 6 7 8
NO EDUC PRIM SEC HIGH OTHER
Brazil Guatemala Nicaragua Panama Peru
negative values. This means that if one is female earnings are lower. Consequently can be argued that this increases the relative return to education for women in comparison to men. If earnings are lower but the absolute return to, for example, higher education remains the same for a woman the impact of education determines a relatively greater part of her earnings.
Potential experience, obtained by taking age and subtracting 6 years, shows a mixture of small negative values in Brazil, Guatemala and Peru and positive value in Nicaragua and Panama. This seems as if more potential experience, which must theoretically lead to more experience, leads to lower returns. Potential experience, however, is dependent of age. The values can be explained by referring to labour market preferences or the composition of the labour force. If young people are more demanded than old people potential experience can show negative values. The same goes for labour force that is relatively young and therefore controls most of the jobs. The latter is in line with positive returns to the age²-parameter that shows positive returns.
(table 5.2.: returns to education by gender, Psacharopoulos 1994) The overall returns to education are low in comparison with literature. This is especially the case when closely interpreting the results. The highest parameters to be found are 2,99 to other education in Peru and 2,74 to higher education in Nicaragua. This means that if an individual completes all years of schooling successfully up until university her earnings only 2,74% higher are in comparison to somebody without education. This is very unlikely. When comparing to Psacharopoulos’ (1994) findings the values are extremely low. Psacharopoulos finds values ranging from 12,8% to 20,1%, distinguishing between men and women, which is high above the values provided in this paper. His coefficient on years of schooling gives 12,4%. Minding that this number is the return to one additional year of schooling the values found in this paper even seem to be lower. A return of 2,74% university level education, which would require at least 15 years of education, is put in contrast with 12,4% average return to one additional year of
education found by Psacharopoulos. Another remark that has to be made is that by using dummies the returns to level of education are measured, not correcting for levels of education already finished. This is important when comparing increasing or diminishing returns to the levels of education. As can be seen in table 5.2. the returns are diminishing. This does not mean that an individual with a higher education level has lower earnings than someone with only a primary education level. It simply means that finishing primary education leads to more increased earnings than finishing higher education. This can be due to the composition of the labour market, the cost of education and education inequality.
The comparison of returns to level education between the countries led to insignificant results. Men Women
Primary 20,1 12,8
Secondary 13,9 18,4
Higher 13,4 12,7
This means that there is no proven difference in returns to education between the countries. It can therefore not be concluded that the countries have different returns to education. Testing the influence of resource abundance on the returns to education can obviously not change this. The returns are
nonetheless shown in figure 5.2. in combination with the resource abundance ratios. It can, finally, not be concluded that resource abundance is of any influence on the returns to education in developing
countries. All in all, the results are in strong contrast with earlier findings. This can be due to data
selection, i.e. focusing on employment and leaving out other types of work, leaving out seasonal effects or the other arguments given in 5.ii.
(figure 5.2: resource abundance and returns to education) -2,000 -1,000 0,000 1,000 2,000 3,000 4,000 5,000 6,000 0 2 4 6 8 No Educatoin Primary Secondary Higher Education Other
Bibliography
Acemoglu, Daron, and Joshua D. Angrist. 2001. “How Large Are Human-Capital Externalities? Evidence from Compulsory Schooling Laws.” In NBER Macroeconomics Annual 2000, ed. Ben S. Bernanke and Kenneth Rogoff, 9–59. Cambridge and London: MIT Press.
Acemoglu, D., Robinson, J.A., 2007. Persistence of power, elites and institutions. NBER Working Papers 12108, Cambridge, MA.
Acemoglu, D., Johnson, S., Robinson, J.A., 2002. Reversal of fortune: geography and institutions in the making of the modern world income distribution. Quarterly Journal of Economics 117, pp. 1231–1294. Auty, R. (1994). Industrial Policy Reform in Six Large Newly Industrializing Countries: The Resource Curse Thesis. World Development, Vol. 22, No. 1, pp. 1 l-26.
Auty, R. M. (2001). Resource abundance and economic development: Introduction and overview. In R. M. Auty (Ed.), Resource abundance and economic growth (pp. 3–16). Oxford University Press.
Bardhan, P., 1997. Corruption and development: A review of the issues. Journal of Economic Literature 35, 1320-1346.
Becker, G. (1964). Human Capital: A Theoretical and Empirical Analysis, with Special Reference to Education. New York: Columbia University Press.
Benhabib, Jess, and Mark M. Spiegel. 1994. “The Role of Human Capital in Economic Development: Evidence from Aggregate Cross-Country Data.” Journal of Monetary Economics, 34(2): 143–73.
Birdsall, N., Pinckney, T., & Sabot, R. (2001). Natural resources, human capital, and growth. In R. M. Auty (Ed.), Resource abundance and economic growth (pp. 57–75). Oxford: Oxford University Press.
Bjorklund, A. and Kjellstrom, C. (2000) Estimating the Return to Investments in Education: How Useful is the Standard Mincer Equation? Economics of Education Review, forthcoming.
Breton, T. 2010. The quality vs. the quantity of schooling: what drives economic growth? Economics of Education Review 30 (2011), 765–773
Card, D. (1999) Education and Earnings. In O. Ashenfelter and D. Card (eds), Handbook of Labor Economics. Amsterdam and New York: North Holland.
Card, D., & Lemieux, T. (2001). Can falling supply explain the rising return to college for younger men? A cohort-based analysis. The Quarterly Journal of Economics, 116(2), 705–746.
Chevalier, A., Lydon, R. (2001). Estimates of the Effect of Education on Job Satisfaction. Mimeo, University of Warwick
Cohen, Daniel, and Marcelo Soto. 2007. “Growth and Human Capital: Good Data, Good Results.” Journal of Economic Growth, 12(1): 51–76.
Corden, 1984, Corden, W.M., Neary, J.P., 1982. Booming sector and de-industralisation in a small open economy. The Economic Journal 92, pp. 825–848
Couttenier, M., 2009. Relationship between natural resources and institutions. In: ESNIE TEAM Conference Paper, Cargèse, Corsica.
Davis, G. A. (1995). Learning to love the Dutch disease:Evidence from the mineral economies. World Development, 23(10), pp. 1765–1779.
De Medeiros Costa, H.L., dos Santos, E.M. 2013. Institutional analysis and the “resource curse” in developing countries. Energy Policy 63 (2013), 788–795.
Easterly, W., Levine, R., 2003. Tropics, germs, and crops: how endowments influence economic development. Journal of Monetary Economics 50, 3–39.
Fulford, S. 2014. Returns to education in India. World Development Vol. 59 (2014), 423-450.
Greiner, A., Semmler, W. (2002). Externalities of Investment, Education and Economic Growth. Economic Modelling 19 (2002), pp. 709-724.
Griliches, Z. (1979). Sibling Models and Data in Economics: Beginnings of a Survey. Journal of Political Economy, 87, pp.37–64.
Gylfason, T., Herbertsson, T.T., Zoega, G., 1999. A mixed blessing: Natural resources and economic growth. Macroeconomic Dynamics 3, 204-225.
Gylfason, T. (2001). Natural resources, education, and economic development. European Economic Review, 45(4–6), pp. 847–859.
Haber,S., Menaldo, V. 2010. Natural resources in Latin America: neither curse nor blessing. Mimeo, Stanford University.
Hall, Robert E., Jones, Charles I.. 1999. Why Do Some Countries Produce So Much More Output per Worker than Others? Quarterly Journal of Economics, 114(1): 83–116.
Hamilton, K., & Clemens, M. (1999). Genuine savings rates in developing countries. World Bank Economic Review, 13(2), pp. 333–356.
Hanushek, E., Woessmann, L. (2008). The Role of Cognitive Skills in Economic Development. Journal of Economic Literature, V 608 ol. XLVI (September), pp. 607-663.
Hanushek, E., Woessmann, L. (2010). Education and Economic Growth. International Encyclopedia of Education (2010), vol. 2, pp. 245-252.
Hanushek, E. 2013. Economic growth in developing countries: The role of human capital. Economics of Education Review 37 (2013), 204–212
Harmon, C., et al. (2003). The Return to Education: Microeconomics. Journal of Economic Surveys (2003), vol. 17 no.2, pp. 115-152.
Haveman, R., and B. Wolfe. 1984. Schooling and Economic Weil-Being: The Role of Non-Market Effects. Journal of Human Resources 19(3): 377-407.
