Master Thesis/ MSc International Economics and Business
Sources of Inequality in Latin America: a deep look at the inequality
legacy of the import substitution period.
Joaquin Hueso Gonzalez/s3232506 J.J.Hueso.Gonzalez@student.rug.nl
University of Groningen
1
Abstract
This work explores the sources of inequality in Latin America. It argues that these sources have varied throughout the different economic periods that Latin America has experienced. It focuses on the inequality legacy of the period of import substitution industrialization. It is claimed that the inability of the industrialization process in Latin America to promote industrial upgrading motivated rising inequality forces in the trade liberalization period. The work provides an empirical analysis for the period 1980-2010, where the inequality legacy of the import substitution epoch is included as a variable.
2
TABLE OF CONTENTS
1. Introduction………..……...3
2. Theory Discussion………..……5
2.1 Pre-independence and export-led growth period……….….5
2.2 Import substitution industrialization period……….9
2.3 Trade liberalization period……….….15
3. Methodology and Data……….….…..21
3.1 Compilation of agricultural salaries……….…..21
3.2 Construction of upgrading ratios……….…...22
3.3 Data description and model……….……...26
4. Empirical Results………30
5. Conclusions and Limitations of the Work………..36
6. References………..39
3
1. INTRODUCTION
The distribution of income is one of the most debated issues in the world today. Such is the dimension of this topic that in recent years its discussion goes far beyond purely economic literature. There is no doubt that the term inequality generates a bigger media impact than the vast majority of economic terms. To achieve the most equitable distribution is, at least on paper, one of the main objectives that virtually every government is pursuing today, regardless of its economic ideology.
This work addresses the income distribution in Latin America, the region in the world that most acutely feels the impacts of this problem. It tries to determine the sources of inequality for each of the three economic periods identified: the export-led growth period, the Import Substitution Industrialization period (ISI) and finally, the trade liberalization period.
Literature extensively covers the inequality issue in Latin America. Researchers strongly agree that income inequality increased in Latin America for the first and third period. Nonetheless, as it will be shown later on, there is no clear consensus for the import substitution period. Undoubtedly, the trade liberalization is the period that most often is covered in literature. The influence of liberalization policies over income inequality has been deeply studied in Latin America and in other developing regions in the world. This work makes an empirical approach for the trade liberalization period (1980-2010) but tries to contribute with a different explanation of the phenomenon. In such a context, the main contribution of this essay is the link among two of the weaknesses of the development in Latin America; the inequality and the failure to successfully industrialize the economic activity. It is argued that the inequality legacy of the import substitution period goes beyond the period itself. In such a way, the incapacity of the industrialization period to successfully upgrade the industrial base caused deindustrialization after the adoption of the trade reforms. This led to rising inequality.
4
The essay also introduces an upgrading ratio in order to assess whether Latin America has upgraded its manufacturing industry. In addition, export upgrading is also measured with the purpose of comparing both kinds of upgrading and determine its relationship. Some interesting conclusions are drawn from the research. Firstly, the descriptive statistics lead to the conclusion that Latin America has not experienced industrial upgrading and has been closer to industrial downgrading. In addition, the empirical part reveals that the inequality legacy of the ISI period appears to be an important predictor of the rising inequality experienced since the liberalization reforms.
5
2. THEORY DISCUSSION
According to the present work the economic history in Latin America can be divided into three different periods, with each period having its particular sources of inequality. In table 1, it can be seen in a clear and visual way what these periods are, and what are their sources of inequality. Each of the three economic models presented is analyzed in depth in the next sections.
Table 1. Summary table
Economic model Years Income
inequality Source
Social group involved
Export-led growth Independence- WW1
Land rents Land owners Industry
Substitution Industrialization
End WW2- 1980 Manufacturing labor
Workers that engaged in ISI industries Trade
liberalization
1980-present “ISI legacy” Low-skilled workers displaced by deindustrialization
Source: own elaboration
2.1 Pre Independence and Export-Led Growth Period
The inequality in Latin America has deep historical roots. Numerous authors have produced research about the colonial origins of the uneven income distribution in Latin America (LAC). For instance, Reynolds (1996) mentions that in the pre-independence period the system of royal grants established the great latifundios1 as the main characteristic of land distribution. The latifundios concentrated the land in few hands creating unequal distributions since early times.
In the beginning of the post-colonial period the inequality persisted in the region. The economic and political power remained in the hands of the European descendants, who used their power to shape the policies and institutions perpetuating the inequality (Ferranti, 2003).
In the new scenario of independence, the export-led growth was the economic model that dominated the region. As Abad (2008) states, this period started just a few decades after
6
independence2 and was based in the export of primary products (raw materials and agricultural products).
Table 2. Income distribution in Latin America. Gini index estimates.
1870 1913 1929 1938 1950 1960 1970 1980 1990 ARGENTINA 39.1 61.8 49.3 50 39.6 41.4 41.2 47.2 47.7 CHILE 47.2 65.5 49.2 40.5 41.7 48.2 47.4 53.1 54.7 BRAZIL 32.9 29.5 47.2 46.6 55.4 57 57.1 57.1 57.3 COLOMBIA 46.8 40.2 45 51 54 57.3 48.8 56.7 COSTA RICA 30.7 50 44.5 48.5 46 MEXICO 27.8 24.3 30.4 55.5 60.6 57.9 50.9 53,1 URUGUAY 29.6 45.9 36.6 34.9 37.9 37 42.8 43.6 40.6
Source: Prados de La Escosura (2007)
The effects of this growth model on the income distribution were clearly negative3. In table 2, it can be appreciated how the Gini indices rose sharply from 1870 to 1913 in the countries that form the so-called South Cone, Argentina, Chile and Uruguay. These countries were those that most intensively followed this growth model due to their high land abundance and raw materials. Nevertheless, it can be appreciated that the inequality reduced in Brazil for the same period. Williamson (1999) argues that the reduction in the Gini index in Brazil could have had its origins in the emancipation of the slaves. Therefore, the negative trend in the inequality during the Belle Époque4 is more a characteristic of the South Cone rather than a common pattern in the entire region. It would be wrong to speak about increasing inequality in Latin America for the period 1871-1914. However, it is accurate to say that those countries highly involved in the export-led growth model experienced a sharp increase in the inequality for that period. Hence, rising inequality can only be associated with the export-led growth model and the South Cone, but not with the entire Latin American region or the period 1871-1914.
2 Most of LAC countries reached independence from Spain between 1808 and 1826. Brazil declared its independence from Portugal in 1822.
3 Williamson (1999) states that is only after 1870 when the boom of the primary products brought negative consequences to the income distribution. Before that time, in the decades right after
independence, the inequality in Latin America was not higher than in Occidental Europe or United States, Williamson defends.
7
Under the export-led growth, the labor income was coming mainly from agriculture, and to a lower extent from other economic activities such as traditional workshops. Under this scenario, there was not an economic activity which was considerably more productive than the other and therefore it can be deduced that the labor income was equally distributed. However, on the other hand, the capital income5 obtained by the landowners of the latifundios was huge and thus it can also be understood that the capital income was the main source of ascending inequality in Latin America during the export-led growth time. The huge exports of raw materials, brought large sums of capital gains for the landowners. Nevertheless, as the returns from labor did not increase as fast, the income of landowners (land rents) significantly grew relative to the workers, bringing an uneven income distribution. As Prados de la Escosura (2007) stresses, the export led-growth of LAC is a good example of the Stolper-Samuelson prediction since the rewards of the abundant factor, natural resources, grew faster.
Interestingly, in graph 1 (see appendix) it can be seen how the inequality followed along this period the trends of globalization. The reduction of tariffs and transport cost clearly benefited landowners. Once World War I started, the moment considered as the end of the first wave of globalization, the increase in the protectionism and transport prices that took place was not beneficial for the export of primary products, and therefore the rents of the landowners decreased. The outcome of this was a reduction in inequality from 1913.
The years between the outbreak of WW1 and the end of WW2 do not belong to any of the period described. It was a time in which the model that dominated the first period collapsed with the starting of the international confrontations and the subsequent end of the first wave of globalization. Therefore, as the main source of growing inequality (land rents) became considerably lower during this period, which can be named as “adjustment period”, the distribution of income became more even. Williamson (1999) considers this period as the great equalization of the 20th century. Nevertheless, it exclusively affected the South Cone of Latin America. The “adjustment period” in which there was not a lead economic model in Latin America, came to its end with the starting of the industrialization period.
8
In graph 1 can be observed that once the World War II reached its end and the second wave of globalization started, the inequality again increased in LAC. However, the inequality increases cannot longer be associated with the strengths of globalization (tariffs and transport cost). As is going to be explained in the next section, the next economic model that LAC adopted marginalized the land owners and capital income stopped being the source of an unequal income distribution.
Bearing in mind all the theory explained above, the research question for the export-led growth period is as follows:
- Research question 1: the increase in the capital income of landowners (land rents) relative to the labor income (wages) of the rest of the economy was the main source of inequality.
Using secondary data extracted from other authors the above research question is shown with descriptive statistics.
Graph 2. Rental/wage ratio and TOT for Argentina and Uruguay Period 1870-1919. 1911=100
9
Graph 3. Capital/labor ratio and TOT for Chile. Period 1860-1930
Notes: Capital/trabajo refers to the capital/labor ratio. Términos de Intercambio refers to TOT. Numbers in the left-vertical axis refers to capital/labor ratio and in the right-vertical axis to TOT. Horizontal axis refers to years.
Source: Rodriguez Weber (2007)
Graphs 2 and 3 are a graphical representation that captures all the forces affecting inequality in the export-led growth period. Graph 2 includes the rent/wage ratio for Argentina and Uruguay, and Graph 3 includes the capital/labor ratio in Chile. In other words, two different names that essentially capture the same trend, the increase of the capital income relative to labor income. Both graphs also include the evolution of the Terms of Trade (TOT). This upward trend of the TOT represents the advantageous situation of primary goods exported over the goods imported. It can be seen that the evolution of TOT and the capital/labor ratio follow the same trend. It is important to notice that a positive TOT does not immediately mean higher inequality. However, the large gains obtained as a consequence of the successful trade were not evenly distributed among landowners and workers as can be seen in the evolution of the capital/income ratio.
2.2 Import Substitution Industrialization period
It would be wrong to consider industrialization in Latin America before the WW1. Until then, the bulk of manufacturing goods in LAC were produced in small and local workshops (Baer, 1972). However, the industrial activity after WW1 was still tiny and it is after the WW2 when the region started a strong ISI (Import Substitution Industrialization) strategy.
10
primary products was a mistake because that caused a deterioration of the TOT for Latin America. Prebisch (1963) states that if the volume of export of primary products of LAC remained stable, the countries would lose purchasing power in relation to the value of industrial goods imported. In the same report, he defends the use of the income coming from the primary exports to modernize the economy through the development of a manufacturing sector. According to him, LAC countries would be able to develop an inward-growth model6 based on import substitution without sacrificing primary exports. Nevertheless, other authors have shown that during the ISI period the primary sector and its exports were highly harmed (see next page).
The ideas of Prebisch materialized and the inward-growth model became a broad and ambitious policy tool for industrial development in LAC after WW2 (Baer, 1972). It is widely believed that almost every LAC country adopted this strategy. Nonetheless, Bulmer-Thomas (1995) argues that in only 6 countries the ISI strategy was the main policy tool for development. These countries are Brazil, Chile, Argentina, Mexico, Colombia and Uruguay, the so-called LAC6. In the remaining countries, even though the ISI became important, the export-led growth continued to be the main economic strategy. Within the ISI framework, manufacturing became the leading sector for growth in the 50s in LAC (table 3, see appendix). In the 60s, the LAC6 countries reached a manufacturing share over GDP similar to the developed countries and warranted the label of semi-industrial economies (Bulmer-Thomas, 1995). In addition, table 4 (see appendix) shows the LAC share of World exports during the ISI years. In 20 years, the share reduced to the half in part due to the lower promotion of the exports during the period.
There are several authors that consider the ISI period as a successful one. Hira (2007) stresses that the period was successful in terms of growth. Cardenas, Ocampo and Thorp (2000) point out that the ISI strategy accomplished a great industrial progress, notwithstanding the macroeconomic volatility and the rising inequality that it brought. On the other hand, many are the critical arguments about the ISI strategy in Latin America. Bulmer-Thomas (1995) says that the industry was inefficient in every sense and high cost. Baer (1972) emphasizes that the autarkic characteristics of the ISI7 strategy were very detrimental especially for the capital intensive industries that had high fixed
6 ISI, inward-looking model and inward-growth model are used interchangeably
11
cost and required large production to enjoy increasing returns to scale. Baranson (1969) found that the costs in a capital intensive industry like the automobile industry, were in Argentina 60 to 150% higher than in United States.
Likewise, the agricultural sector was also negatively affected by the ISI policies. Overall, the ISI strategy neglected other sectors of the economy. Baer (1972) points out that most public resources were often allocated to the industry, leaving few available for agriculture. Moreover, the overvalued exchange rates, promoted to benefit the imports of intermediate inputs for the industry, damaged the international competitiveness of the export of primary products.
There is a clear consensus in the literature defending that the export-led growth model brought negative outcomes for the income distribution. Similarly, researches strongly agree that the inequality also worsened after the trade liberalization reforms (see next part). Conversely, there is no clear consensus in the research literature about the inequality trends during the ISI decades. There are many essays that defend the ISI period as a period that brought calm to the income distribution8. Bertola et. al (2008) defend that the state-led industrialization has been the only period in Latin American economic history during which growth was relatively fast and inequality was reduced. Similar to this, Prados de la Escosura (2007) stresses that the inward-looking model brought stabilization in the income distribution.
Among those essays that stress the opposite, Bulmer-Thomas (1995) says that both groups of countries (those that fully adopted ISI, and those who followed the export-led growth) failed in the same, the distribution of income. Also, Bulmer-Thomas says that the unequal distribution had a negative impact in the ISI by reducing the effective market for the industrial goods. Since the exports were almost non-promoted9, the local market was the main one, however, the uneven income distribution made it impossible to create a large local market to support the industry. This shows why the inequality was not only a consequence of the ISI period, but also a cause of its underperformance.
In addition, the absorption of workers in industry was smaller than the population growth, and hindered the income distribution, which was more concentrated (Baer, 1972). Finally,
8 The literature referred here deals about inequality in its aggregate effects in Latin America. Papers focus on individual countries are skipped when is purely about inequality conclusions.
12
Hira (2007) supports the negative impact over income distribution of ISI arguing that the tax system was inefficient and reduced the government ability to promote income redistribution, which led to inequality.
It is important to emphasize that papers sometimes use different time periods to evaluate the inequality outcome. For instance, Bertola, Castelnovo and Willebald. (2008) consider the ISI period starting in 1920. Nevertheless, according to the theory the inward-looking model did not start until the end of the World War II. As mentioned earlier, the period from World War I until the end of World War II is considered as a transition period that does not belong to any particular economic model.
Furthermore, the relationship between industrialization and inequality is commonly supported in literature. As Cheong and Wu (2012) defend, Industrialization is vital for the economic growth, nevertheless, the unevenness in the distribution of industrialization can exacerbate regional disparity greatly. This is more or less what Kuznets (1955) advanced by saying that inequality always increases with industrialization in the early stages of economic development.
This essay defends the idea that inequality increased during the ISI decades. However, as the period lasted almost 40 years, there are some fluctuations within it. But, in most of the cases when taking 1938 as a starting point10 and 1980 as an end point, the evolution of the Gini indices supports the worsening in inequality in most of the cases (table 2). Some authors could argue that the inequality trend was not as negative in the South Cone countries as it was in the export-led growth model. Nevertheless, by having a look at the data available, and also considering that the authors that find this period as detrimental are more numerous than those who opposite, it is reasonable to conclude that way. From the LAC6 countries, Argentina can be consider as the only exception where the income distribution remained highly steady during the period. In the case of Brazil, there is a clear increase in the Gini. In table 2 can be seen that the negative trend belongs to the 40s and 50s, at the opening of the ISI period. However, other sources as Jain (1975) locate more in the 60s the increase in the Gini. For the case of Chile and Colombia, can as well be appreciated a clear increase in the index for the ISI decades. In Mexico, the Gini doubles from 1938 up to 1960. Finally, in Uruguay there is also a slow, but positive trend
13
for the Gini. Overall, with the exception of Argentina, there is not a single country where the value of Gini is lower in 1980 than in 1938. It is important to notice that the negative trends belong almost entirely to the first two decades of the ISI period. But, in any case the reductions of the index in the 70s do not fully counterbalance the previous negative trends.
Effects of the ISI period in the sources of inequality
With the adoption of the ISI policies the economic activity of some Latin American countries completely changed and this implied a modification in the income sources. Now we also have in the economy the income coming from the manufacturing activities that as explained below altered the income distribution.
Capital income: the main idea is that the capital income remained highly stable during the ISI period and hence did not affect to the income distribution. On the one hand, the inward-looking strategy harmed other economic activities, including primary activities. The capital income coming from the export of commodities decreased. As stated earlier, to benefit the industry the exchange rate was overvalued, what greatly harmed the competitiveness of the agricultural exports.
On the other hand, the capital income coming from manufacturing activities appeared. According to Reynolds (1996) the capital owners obtained large profits from industrialization in LAC because the value added of the wage costs was low in comparison with other countries. In short, there is an increasing source of capital income (industrial profits) and a decreasing source of capital income (rents of landowners). Both strengths more or less counterbalance each other what makes the capital income to remain relatively stable throughout the import substitution epoch.
Labor income: clearly was the source of income that suffered more disturbances over the period. The apparent equality among workers present during the export-led growth was altered due to the blooming of the industry sector. The manufacturing sector is characterized by higher salaries than agriculture and other activities because it enjoys high productivity. Also, Baer (1972) says that the wages in the industry were relatively higher due to the labor legislation.
14
large supply of unskilled workers acts to keep earnings low for those at the bottom of the skill ladder. Altogether, the industrialization brought wages gaps among workers in the primary-traditional sector and workers in the industry-modern sector. Moreover, brought gaps among workers of the same industry (white collar workers vs blue collar workers). This fact implied a new situation of unequal labor income that increased the inequality. This work tries to focus on the increasing gaps among traditional and modern sector as a source of inequality for the ISI period. There are already some essays that focus in the intra-industry differences as a source of inequality. For instance, Frankema (2008) analysis the wage differentials within the different ISI subsectors as a source of increasing inequality. Nonetheless, there is not much study concerning the wage gaps that appeared among workers in the primary sector and the new manufacturing workers.
Assuming the theory explained above, the research question regarding the import substitution period can be stated as follows:
- Research question 2: the income differences that appeared among traditional jobs and the new manufacturing jobs worsened the income distribution.
To properly show the wage gap among the traditional and modern sector, agricultural and industrial salaries for the ISI period are needed. In this case, as well as in the first period, also secondary sources obtained from other authors are used.
15
Graph 4. Average agricultural wage vs average industrial wage in Chile. Period 1934-1973
Notes: Agricola refers to agricultural. Numbers in the vertical-axis refer to wages. Numbers in the horizontal axis refer to years.
Source: Reyes Campos (2015).
2.3 Trade Liberalization Period
As a consequence of the severe debt crisis that Latin America suffered in the 1980s and that brought an entire decade of slow economic growth, the region modified its entire economic system.
16
comers, Uruguay. Table 5 (see appendix) shows the reform years for each of the countries studied.
In this section the sources of inequality during the trade liberalization period will be studied, putting a special emphasis on the industrial upgrading as a possible strength that has motivated inequality.
The main line of reasoning (explained in depth below) is as follows: Latin America did not succeed in promoting industrial upgrading during ISI period. This meant that when the continent opened up to trade it lacked competitive advantage in manufacturing. As a consequence, the industrial sectors that developed during ISI were outcompeted by foreign companies and many were force to close down. In aggregate terms, this caused deindustrialization, and eventually increased inequality. With deindustrialization many Latin American countries started again to focus on the export of raw materials, in other words a return to the export-led growth time.
Industrial upgrading refers to the process by which economic actors (nation and firms) move from low value to relatively high-value added activities (Gereffi, 2005). In general terms, LAC has failed to promote such upgrading. For instance, Hirschman (1992) points out that one of the failures of the ISI strategy is that LAC failed to upgrade the industrial structure. Also, Sanchez-Ancochea (2009) states that the economic reforms of the 1990s failed as well to promote upgrading in the industry. There is maybe one exception in the continent, Costa Rica. According to Sanchez-Ancochea (2009) Costa Rica has moved up in the global value chain, creating a comparative advantage in high tech sectors.
But, why was it so important to promote industrial upgrading in Latin America during the ISI period? As Sanchez-Ancochea (2009) stresses, economic development requires a sustained process of industrial upgrading. Such upgrading is perhaps not key to develop a comparative advantage but it is key to sustain the comparative advantage. For the industry to drive economic growth it must move to activities with higher values that make the economy more dynamic.
17
based were outcompeted. As Gonzalez et al. (2010) argues for the Argentinean case, the sectors that lead the growth during the ISI epoch were those most damaged with the open up to trade. Those countries without strong competitive advantage in manufacturing became net importers of manufactured goods with the onset of trade liberalization (Rodrik, 2016). As stated earlier, the capital-intensive industries were inefficient and with high costs because the local markets were not big enough to reach economies of scale and hence LAC firms were not able to face the foreign competition. For the labor-intensive and low technological component industries the situation was similar. When the integration of the Latin American countries in the global economy took place, Asian countries were highly specialized in that kind of goods. The Asian countries were more competitive with a much stronger comparative advantage, since their resources (mostly labor) were considerably cheaper. In short, after trade liberalization, LAC was not competitive in the industry scale where it had been mostly specialized during the ISI epoch.
In a different scenario, where LAC had succeeded in promoting industrial upgrading during the import substitution strategy it could have developed a sustainable comparative advantage in medium or high technological component goods and avoid the Asian competition. Nevertheless, the ISI policies were indiscriminate, not concentrated in any sector that could have enjoyed comparative advantage. This setting clearly differs from the ISI policies in Asia where the government focused the resources in concrete sectors, a policy more commonly known as the “picking winners strategy”.
The close-down of many local industries led to the deindustrialization of the region. Boggliaccini (2013) states that the trade liberalization caused deindustrialization through the shrink of the industrial employment. He also says that as most of the industrial workers in LAC were low-skilled, the industrialization destroyed an important source of income for the low social groups. This destruction of manufacturing labor was bigger than the opportunities brought by the trade liberalization. The deindustrialization trend of LAC can be seen in graphs 5 and 6 (see appendix).
18
rising informal economy reduced the scope of the income redistribution policies, this also led to increasing inequality.
This process of deindustrialization reshaped the economic model of LAC. Countries moved their comparative advantage towards the exports of raw materials and agriculture products, a return to the beginning, which shows the general failure of the ISI period in Latin America.
Overall, this essay understands the failure of LAC to sustain a competitive manufacturing industry as a legacy from the ISI period. It is after the trade liberalization when the deficiencies of the inward-looking model were better appreciated. The ISI period was successful in terms of growth, but unable to develop an industrial base able to lead economic growth after the trade liberalization took place. Therefore, the inequality legacy of the ISI decades would not be only within the period itself (1945-1980), but also in its industrial organization that ended up in deindustrialization and raised inequality.
The influence of industrial upgrading over the distribution of income is rarely studied in literature. However, there is extensive literature about the concept named social upgrading and its relationship with industrial upgrading. Social upgrading refers to the improvement of the status of the workers throughout better salaries and higher rights. Therefore, this concept of social upgrading does not include income distribution as a social consequence of upgrading. Barrientos, Gereffi and Rossi (2011) point out that social upgrading might occur for some workers but not for others in the same factory or industry. If this happens, then inequality would rise through the differences occurring in the labor income.
How the sources of inequality were affected during this period?
Traditional trade theory based on the Heckcsher-Ollin model and the Stolper-Samuelson theorem expects trade liberalization in developing countries to reduce inequality by reducing the wage gap of high-skilled and low-skilled workers. The factors rewards of the workers of the factor-abundant sectors (labor-abundant in Latin America) should increase relative to the other sectors.
19
did not hold for Latin America due to two reasons. The first, the technological change, the second, the higher competition that LAC countries had to face when they joined the global economy from China and other Asian countries. This work argues that competition coming from Asia could have been partially reduced if companies have moved up in the global value chain during the ISI years.
Labor income: the shrink in manufacturing brought difficulties for the workers of that sector. However, it is reasonable to think that this pattern mainly affected the blue-collar workers. The white-collar workers, due to its high qualification, have less problems in finding a job in case of displacement. Whereas, the displaced blue-collar workers ended up in activities with worse salaries such as agriculture, unemployed or in the informal sector, which is not protected job and thus worsens the worker´s rights. Rodrik (2016) finds in its empirical work that the low-skilled workers of manufacturing made up almost the entire employment losses in the deindustrialization process in Latin America. So, as the labor income for the low social classes decreased, but the labor income for other classes with higher human capital did not, the inequality increased.
Capital income: the two main sources of capital income are affected during this period. On one hand, the capital income coming from industrial activities very likely shrunk with the entire sector. On the other hand, it is believed that the capital income of the landowners increased since many countries of the region started again to focus on the production and exports of raw materials and agriculture products. Then, the capital income probably remained stable after trade liberalization.
The hypothesis that this work introduces as the main source of inequality for the trade liberalization period is as follows:
- Hypothesis: The ISI period legacy has motivated rising inequality forces in the trade liberalization period.
20
Then, what is exactly considered as the legacy of the ISI period?
The key concept here is industrial upgrading. The failure to promote industrial upgrading during the ISI epoch led to the LAC manufacturing industry being outcompeted once the continent opened up to trade. This legacy of a “poor industrial base” ended up in deindustrialization, what reduced the labor income of the manufacturing workers and increased inequality. Therefore, deindustrialization is the mechanism through which inequality increased, but industrial upgrading is the key concept and the origin of the issue.
Bearing in mind these concepts, the inequality legacy of the ISI period is the part of the next period (trade liberalization) inequality that increased as a consequence of what was done in the previous period, the import substitution period.
21
3. METHODOLOGY AND ECONOMETRIC MODEL
The quantitative analysis of this work exclusively focuses on the third economic period described, this is the trade liberalization period. The reasons for this is the scarce data available for the other two periods.
As discussed earlier, not all the countries in LAC applied the liberalization reforms at the same time. The time period chosen for the quantitative analysis is 1980-2010.
The regression analysis, as most of the descriptive statistics shown earlier, only includes the group called LAC6 plus Costa Rica. As the main goal of the work is to assess the inequality legacy of the ISI strategy, only those countries that strongly pursued the industrialization of the country and thus intended to upgrade are included in the source. The inclusion of countries in the sample that during the ISI period continued to rely on the exports of raw materials as the main source for growth would not allow to properly capture the legacy of the ISI period. There is one exception; Costa Rica. Due to its high manufacturing value added over GDP and its export structure it is decided to include it in the sample.
3.1 Compilation of Agricultural Salaries
The compilation process needed to obtain agricultural salaries is not an easy task. Only by doing a search at national level accurate figures can be found. As Reyes Campos (2015) suggest, there is not a compilation work for the primary sector salaries in the ISI period as there was for the industry. As previously has been mentioned, most of the country resources during the ISI period (including labor and capital) were located in industry. Hence, it looks that the national statistics compilation work followed a similar pattern and censuses for agriculture were not extensive in their research.
Besides the limited data available, researchers face other issues in order to construct reliable time-series of agricultural salaries. Reyes Campos (2015) mentions that part of the agricultural salary was in kind and hence work is needed to convert that part of the salary to monetary units. In addition, in order to estimate real salaries the rural prices index is also needed, what is complicated. Reyes Campos (2015) solves all these issues and elaborates consistent agricultural sectors for the ISI period for Chile.
22
3.2 Construction of Upgrading Ratios
In such a context where the industrial upgrading appears as a key issue in the increasing inequality after the trade liberalization, would be interesting to show descriptive statistics that measure it for Latin America. Unfortunately, due to data constraints the measurement of industrial upgrading cannot be calculated for the ISI period. Nevertheless, this easy introduces an industrial upgrading ratio calculated from 1980 onwards. Therefore, this work does not present statistical proof about upgrading for the ISI period, and assumes that such upgrading was poor. However, it is believed that the upgrading figures introduced subsequently, and that show upgrading for the LAC6 countries for the period 1980-2010, are a fair representation of what industrial upgrading looked like in the import substitution decades. In the case that the inward-growth model was successful to develop a dynamic industrial base it is expected to find industrial upgrading in the years after. In the opposite scenario, where LAC6 countries did not manage to create competitive industrial bases, it is not expected to find upgrading from the 80s onwards. In short, the ISI period set the roots for the subsequent upgrading or downgrading.
Measurement of industrial upgrading
23
This measure of industrial upgrading is considered very adequate since it relates both, the value added and the export performance. As it will be shown later on, the exclusive consideration of export to measure industrial upgrading is misleading. Likewise, measuring upgrading only with value added data per capita would also not be totally recommendable since many authors consider the export performance key to fairly represent upgrading. This is the main reason why upgrading cannot be calculated for the ISI period itself. As has been said, the industrialization of LAC countries was highly autarkic with few manufacturing exports. The exports were poorly promoted and are not an accurate representation of the production capacity and competitiveness of the manufacturing industries during ISI. To put it in another way, the ratio proposed in this work is only applicable for an open-economy and therefore cannot be applied to the LAC6 countries for the ISI11 epoch.
Besides the LAC6 countries and Costa Rica, with the motivation of benchmarking the Latin American continent with other development region of the world that also pursued a state-led industrialization strategy, some Asian countries are also included.
Table 6 (see appendix) presents the main results. The outcomes obtained are completely different for both continents. On the one hand, the upgrading ratio of Latin America is slightly above 0, with some countries experiencing downgrading as is the case of Colombia, but others as Mexico or Brazil very close to 0. In short, upgrading in LAC6 has been almost inexistent from 1980 to 2010. Milberg and Winkler (2011) only determine that upgrading has been existent when the ratio is higher than 1/3. Thus, under this frame, Chile and Uruguay are the only countries that have promoted upgrading. Notwithstanding, this upgrading is quite poor when compared to Asian countries. On the other hand, the scenario for Asia completely differs from that in Latin America with an average value which is three times higher than in LAC. The average value of 0.4 indicates that the continent has experienced economic upgrading.
11An evolution of a ratio during the ISI decades that shows the production percentage of the high tech
24
The upgrading diagnosis can be graphically seen in graph 7 (see appendix). Those countries located below the line drawn are those that have failed to promote industrial upgrading. Conversely, those located above the line have promoted upgrading.
Data used to measure industrial upgrading
Concerning data, the growth rate of merchandise exports is obtained from United Nations Conference on Trade and Development (UnctadStat, 2017). The figures of manufacturing valued added as well as the manufacturing employment are obtained from the 10 sector Database of the Groningen Growth and Development Center (Timmer, de Vries and de Vries, 2015).The manufacturing value added is at constant 2005 national prices. Regrettably, this database does not contain data for Uruguay. The data on MVA for Uruguay comes instead from the World Bank and the Data for manufacturing employment comes from the database Indstat2 (United Nations Industrial Development Organization).
Other ways to measure upgrading
Following the upgrading ratio, in this section of the work is introduced the export upgrading as a potential proxy for industrial upgrading. The export upgrading is calculated as the export share of the medium-high and high technology (MHT) goods over the total merchandise exports. The reasoning is as follows: If LAC6 countries had upgraded during the ISI decades, it is reasonable to think that the export share of these products would represent an important part of the total exports. And, what is more important, a growing share from the moment of the trade liberalization polices onwards would be expected. The traditional exports of these countries were until the moment of the liberalization reforms composed by raw materials and agricultural products. However, after the development of the industry during the ISI period and with the integration in the global economy, the exports of manufactures should become important and progressively gain share to the detriment of the primary exports products.
Therefore, another way to measure in a descriptive way the upgrading might be by checking the evolution of the share of MHT manufacturing exports from the 1980 onwards.
25
and Costa Rica are the only countries that according to the evolution of the share have upgraded their export structure considerably, and Argentina to a lesser extent. Second, with the exception of Costa Rica, only the large countries (Mexico, Brazil and Argentina) have now a representative share of MHT exports in their export structure. Finally, in Brazil, Chile, Colombia and Uruguay there is a negative trend since the 2000s. This trend could be related to the increasing competition from China that probably became more intensive when China started to be part of the World Trade Organization in 2001.
Besides the conclusions described in the previous paragraph, the strongest conclusion is that there is not much relationship among export upgrading and industrial upgrading. To put it in another way, export upgrading does not necessarily mean industrial upgrading in the case of Latin America and therefore it is not consider a good proxy for industrial upgrading
For a discussion about the correlation of both upgrading measures see appendix 3. Data used to construct export shares
26
be excluded (ships in this case) are not very significant in terms of export value it is decided to include the entire codes at the 2 digit level.
3.3 Data Description and Model
Dependent variable:
Unfortunately there is not a dataset that provides consistent data of income inequality for LAC starting in 1980.
The database chosen in this work is the Standardized Income Inequality Database (SWIID), Solt (2016). This dataset includes the largest sample of countries and years, being better than any other existing database. The SWIID satisfies the main requirements, frequently mentioned in literature, that every inequality dataset should satisfy. It does not treat observations with different welfare definitions and unit observations as equal. In addition, excludes the observations that only refer to the rural or urban part of the population. As Berhman, Birdsall and Szekely (2001) argue, this is important because in some regions the effects of trade liberalization can be positive and in other regions negative. Hence, it is preferable to have data based on surveys at national level. The only use of rural or urban samples could lead to wrong conclusion (Berhman, Birdsall and Szekely, 2001). Nevertheless, there are two exceptions concerning this issue, Argentina and Uruguay. For these two countries only urban surveys were made until recent years. However, since both countries are characterized by huge rates of urbanization, Solt (2016) keeps the figures even though surveys were not made at national level. Most of the research papers use the same argumentation and include urban data for Argentina and Uruguay. See Berhman, Birdsall and Szekely (2001) or Babones and Alvarez-Rivadulla (2007).
27
Probably, the largest disadvantage of the SWIID is that only provides Gini indices, and not income shares. Nonetheless, those databases that do provide income shares, only incorporate them from the 1990s onwards in the LAC case. This is the case, for instance, of SEDLAC (Socio-Economic Database of Latin America and Caribbean). If together with this database, the income shares provided by the WIID are used, and some interpolations are made, it is possible to have consistent data of income shares from 1990 onwards, than can be used as a dependent variable in a regression starting in 1990. In the model the version 4.0 of the SWIID is the one selected. The newest version of the dataset, the 5.1 version contains some gaps in the indices for the 1980s. Conversely, the version 4.0 and older versions do not contain gaps in the series. In short, from the versions without gaps the newest version is selected.
The reason why the 5.1 version of the SWIID has lower observations is because some observations are no longer considered as reliable by Solt. This is the case for the observations coming from the University of Texas Inequality project (UTIP) that only have a 90% confidence interval. Thus, it is assumed that the data used is not completely reliable, however, is still considered as the best option.
Main independent variable
As has been previously stated, the empirical section covers the main hypothesis stated in this work. It is intended to measure the influence of the ISI legacy on the rising inequality that has taken place in the trade liberalization period. In other words, determine which part of the rising inequality experienced since the application of the trade liberalization reforms has been caused by the previous period legacy.
Under such a framework, it is necessary to include a variable that captures the ISI legacy. Ideally, this variable would be the industrial upgrading in the ISI years, since it is claimed that the absence of a dynamic upgrading during the ISI period is the key of the issue. Unfortunately, this variable cannot be estimated. Therefore, it is decided to directly use the inequality legacy as a variable. However, another issue appears here about how to calculate the inequality legacy.
28
the countries initiated its trade reforms at different points in time, the inequality legacy of the ISI period did not affect all the countries at the same time. As has been said, it is considered in this work that ISI period finishes the year the trade reforms start, with this year being different for every period (table 5, see appendix). Therefore, the year the reforms start it is considered to be the beginning of the inequality legacy.
In the model inequality legacy is represented with the following formula: 𝐿𝑜𝑔 𝐺𝑖𝑛𝑖𝑖𝑡
𝐺𝑖𝑛𝑖𝑖|𝑆|
This representation allows to the value of the inequality legacy to vary every year, in function of the size of the current Gini.
𝐺𝑖𝑛𝑖𝑖𝑡 refers to the value of Gini index for country 𝑖 in year 𝑡. This is for every year its actual value of the period. Therefore it is the same value than the dependent variable. 𝐺𝑖𝑛𝑖𝑖|𝑆| refers to the value of Gini in the previous period (ISI). This value is a constant, calculated as the average Gini of the ISI years within the period 1980-2010. For instance, within the period 1980-2010, Brazil followed an import substitution model until the year 1988, year in which the trade reforms began. Therefore 𝐺𝑖𝑛𝑖𝑖|𝑆| is the average Gini of the period 1980-1987.
For those countries that at the beginning of the period included in the empirical analysis (1980-2010) had already started the trade liberalization period, 𝐺𝑖𝑛𝑖𝑖|𝑆| is the Gini value of some of the years prior to the end of the ISI period. This is the case of Argentina, Chile and Uruguay. As can be seen in table 5, the beginning of the liberalization reforms is for these 3 countries a year prior to 1980.
For Chile the Gini of 1971 is the 𝐺𝑖𝑛𝑖𝑖|𝑆| value. Year 1967 for Uruguay and finally for Argentina, is an average of the period 1971-197612.
Control variables
The control variables used in literature to measure inequality widely differ. As Atkinson (1997) stresses, there is not a baseline model to measure inequality. Therefore, the
12 Preferably, to estimate 𝐺𝑖𝑛𝑖
29
selection process of variables to include in the model in order to explain the process of income distribution is complicated and highly demanding in terms of information needed. Bearing in mind that complications, the model includes the following control variables: education, as the average years of educational attainment per capita for people over 15 years. Data comes from the Institute for Health Metrics and Evaluation (IHME).
To capture the effects of trade liberalization over the income distribution, the trade (exports plus imports) as a percentage of GDP, is included. The data on trade comes from the World Bank (2017). Public debt over GDP is included to measure the influence of the macroeconomic instability. The public debt data is obtained from the Historical Public Debt Database (Abbas et al. 2010). To capture the influence of deindustrialization over inequality, the model includes manufacturing output over GDP (World Bank, 2017). Finally, the export share of MHT as a share of total manufacturing exports is also included as a control variable.
Econometric approach and empirical specification
The model contains a panel dataset. It is decided to include country fixed effects. This allows to control for all the individual time-invariants characteristics (e.g. cultural factors) that could bias the impact of the predictors. The choice is supported by the result of the Hausman test.
Furthermore, several tests are computed in order to detect misspecification issues that could affect to the consistency of the results. The results reveal that the model suffers from heteroskedasticity and serial correlation. Consequently Huber-White Sandwich standard errors are used to correct for it
Equation 1 is the baseline model in the work. Where 𝐺𝐼𝑁𝐼𝑖𝑡 represents the Gini coefficient for country 𝑖 in year 𝑡. 𝐺𝑖𝑛𝑖𝑖𝑡
𝐺𝑖𝑛𝑖𝑖|𝑆| represents the inequality legacy of the ISI period, 𝐸𝐷𝑈𝐼𝑇
30
This model shows some important differences when compared to other models that also try to capture inequality in LAC for the trade liberalization period. The most important difference is the fact that the trade openness (measured here as trade over GDP) is not the main variable in this case, and is used as a control variable that also could affect the inequality. Once again, the inequality legacy inherited from the ISI period is considered as the key component in the model. This inequality legacy somehow reflects the failure of LAC6 countries to upgrade the industry during the ISI years. A failure that ended up in deindustrialization and increased the inequality. This increased inequality (inequality legacy) is the variable that the model includes.
𝐿𝑜𝑔𝐺𝐼𝑁𝐼𝑖𝑡 = 𝛼 + 𝐵1𝑙𝑜𝑔𝐺𝑖𝑛𝑖𝐺𝑖𝑛𝑖𝑖𝑡
𝑖|𝑆|+ 𝛽2𝑙𝑜𝑔𝐸𝐷𝑈𝑖𝑡+ 𝛽3𝑙𝑜𝑔𝑇𝑅𝐴𝐷𝐸𝑖𝑡+ 𝛽4𝑙𝑜𝑔𝐷𝐸𝐵𝑇𝑖𝑡 +
𝛽5𝑙𝑜𝑔𝐸𝑋𝑃𝑂𝑅𝑇𝑖𝑡+ 𝛽6𝑙𝑜𝑔𝑀𝐴𝑁𝑈𝑖𝑡+ 𝛿𝑖+ 𝜀𝑖𝑡 (1)
4. EMPIRICAL RESULTS
Table 8 shows the summary statistics. Some interesting figures can be appreciated in the table. For the variable that accounts for the export share, the standard deviation shows a high figure (16.65). The minimum value (3.32%) corresponds to the export share in Chile for the year 1983, whereas the maximum value of 66.43% accounts for Mexico in the year 2001. These differences depict a fair explanation of the different trade patterns among Latin American countries. While Chile13 is highly specialized in the export of mineral products, Mexico is highly specialized in manufacturing exports.
Additionally, the standard deviation of the openness variable (trade as % of GDP) is also quite high. The importance of trade in the LAC economies strongly differs among nations. In small and open economies such as Chile or Costa Rica (maximum value belongs to Costa Rica) the percentage of trade over GDP is substantially higher than in other larger economies such as Argentina, Brazil or Colombia.
Finally, the variable measuring macroeconomic instability, captured as percentage of public debt over GDP, also shows a high standard deviation of 28.28. This instance is
31
particularly striking because the high differences are not only among countries, but also, and mainly, within countries. Without going any further, both the minimum and maximum value belong to Chile14.
Table 8: Summary statistics
Variable Obs Mean Std. Dev. Min Max
Gini 217 46.41412 3.877008 37.90635 56.07789 Inequality legacy 217 1.0279 0.1167 0.8030 1.3177 Education 217 6.7464 1.7179 3.18 10.46 Export Share 217 19.8572 16.6587 3.3299 66.4325 Trade as % GDP 217 42.8576 20.6321 11.5 92.5 Public Debt 217 50.5824 28.282 3.9 165.5 Manu (% GDP) 217 20.71 4.95 11.8 34.6 Richer 40 147 75.2 3.27 68.4 82.2
Notes: Statistics are presented without logarithm transformation. Data units are as follows: Inequality legacy is represented as a ratio. Education as the mean years of educational attainment. Export share refers to the % of medium-high and high technology manufactures over total manufacturing exports. Trade as the value of exports plus imports over GDP. Public debt as % of GDP. Manu (%GDP) as the share of manufacturing activities over GDP. Richer 40 as the income share of the richest 40% of the population. To see a deeper explanation of the variables see methodology section.
Table 9 shows the correlation table with the significance levels. It can be appreciated that there is a significant linear relationship between the dependent variable (Gini) and four of the independent variables (Inequality legacy, education, public debt and manufacturing over GDP). Overall, the correlations between the dependent variable and the different independent variables are not very high, being education with a negative correlation of 0.3605 the highest value. Moreover, as expected, the linear correlation between Gini and the variable “Richer 40” is considerably high and significant. “Richer 40” accounts for the 1st and 2nd quintiles of income and it is used as dependent variable in the robustness test. The independent variables “Trade” and “Export share” do not show a significant linear correlation with the dependent variable. Nevertheless, due to theoretical reasons it is believed that these variables are important to explain inequality in Latin America and therefore it is decided to include them in the model.
On the other hand, there is not any correlation among the different independent variables considered as high. In case the correlation among two of the independent variables was
32
reasonably high, one of the variables was dropped from the model in order to mitigate the problems associated with collinearity.
Table 9: Correlation table
Gini In. Legacy Edu. Trade Public debt Export Share Manu (%GDP) Gini In. Legacy 0.1313*** Edu. -0.3605** 0.4500*** Trade -0.1277 -0.1867*** 0.4914*** Public debt -0.1432** -0.1182*** -0.1835** -0.0882 Export Share Manu (% GDP) -0.011 -0.2523*** -0.4936*** -0.0443 -0.1802** -0.4944*** -0.0833 -0.3408*** 0.1524** 0.2969*** 0.007 Richer 40 -0.6458*** -0.0652 0.4644*** 0.259*** 0.0039 0.088 0.0878
Notes: ***p<0.01, **p<0.05. All correlation values refer to logarithm values.
Table 10 provides the main empirical results. It is important to bear in mind that a positive coefficient means an increase of the Gini index and therefore a higher rate of inequality. Conversely, a negative sign represents a reduction in inequality.
The expected signs of the coefficients are as follows: concerning the main variable, it is expected that a higher inequality legacy from the ISI period increases inequality for the trade liberalization period.
In addition, for the control variables, it is assumed that a higher level of education reduces the size of the Gini Index. Thus, it is expected a negative coefficient for education. Likewise, it is expected that the variable that captures the industrialization level of the country (Manufacturing over GDP) appears also with a negative sign. This essay defends that deindustrialization forces reduced an important source of income for the low-skilled workers, increasing the inequality. Therefore an increase of the industry sector is considered as to have a beneficial effect in the income distribution in Latin America. The openness variable (trade as % of GDP) is expected to appear with a positive sign. Also, an increase in the macroeconomic instability, captured as public debt over GDP, should affect positively to the Gini index15. Finally, it is assumed that a higher export upgrading
15 This work does not treat in depth the relationship of income distribution with trade and
33
will be reflected with a positive sign and hence increases inequality. To see a deeper explanation of export upgrading and income inequality see appendix 3.
Column 1 of table 10 depicts equation 1. It presents interesting results. First, the inequality legacy of the ISI period is as expected a significant predictor of inequality in the trade liberalization period. A higher inequality legacy is related with a higher inequality index in the trade liberalization period. Furthermore, the coefficient is considerably high. A 1% increase in the inequality legacy is related to an increase of 64% in the Gini index (see limitations section for a discussion of the coefficient size).
TABLE 10: Main empirical results
Dependent variable Gini Index Richer 40
[1] [2] [3] Education -0.09958*** (0.01624) -0.0962*** (0.0256) -0.1264 (0.9030) Inequality legacy 0.6428*** (0.2563) 0.6325*** (0.1579) 0.4727** (0.1446) Trade as % of GDP 0.0375** (0.01373) 0.0360** (0.0118) 0.04587 (0.0314) Macroeconomic instability -0.0004 (0.0059) -0.0010 (0.0019) -0.0011 (0.0086) Export share 0.0387 (0.0161) 0.0384** 0.5184 -0.00567 (0.0122) Manufacturing as % of GDP -0.0284** (0.0086) -0.020 (0.0131) 0.0029 (0.0267) N 217 217 147 R-sq. (within) R-sq. (between) R-sq. (overall) 0.6346 0.0257 0.1283 0.6217 0.0224 0.1339 0.2057 0.0471 0.0824 Prob > F 0.0000 0.0000 0.0001
34
On the other hand, education, expected as the variable with the highest power to improve income distribution appears as strongly significant. A 1% increase in the mean of years of schooling is associated with a decrease of 9.9% in the Gini index. Likewise, manufacturing value over GDP also has a negative and significant coefficient. A 1% rise in the industrialization level brings a reduction of the Gini index of 2.8%
Furthermore, trade as a percentage of GDP is also significant at the 95% level and has the expected positive sign. An increase of 1% in the trade is associated with an increase in the inequality level of 3.7%.
The remaining two control variables are both insignificant. On the one hand, the macroeconomic stability (public debt % over GDP) has not effect on inequality in this model and appears with an unexpected negative sign. On the other hand, the export share appears with the expected positive sign. Although is an insignificant variable in the model, the low p-value of the variable (0.053) indicates that is close to be a significant explanatory variable of inequality.
Robustness of the Results
To check the robustness of the empirical results obtained, the same regression is made with some modifications.
A new regression is run using in this case the income share of the richer 40 percent of the population. It is important to remember that the data for the income shares does not come from the SWIID. In addition, unlike in the other case, this regression captures the relationship for the 1990-2010 period, since data for income shares was not available for the 1980s.
35
expected (export share). Overall the exclusion of the 1980s reduces the explanatory power of the model. Nevertheless, this is somehow expected since in the 1990s and 2000s some of the variables suffered important modifications and high volatility. With the reduction of the time period measured and thus with a lower number of observations, the regression in column 3 loses significance compared to the main regression. As can be seen, the R-squared values are considerably lower when the richer 40 is used as dependent variable (column 3). For instance, the within R-squared16 figure for the regression with the richer 40 as dependent variable is significantly lower than the within R-squared in the baseline regression (column 1).
Furthermore, another robustness test is carried out. In this case the dependent variable is also the Gini index and the time period remains as the original one (1980-2010). Nonetheless, one of the control variables is modified. The macroeconomic instability is now reflected by the inflation rate17. Column 2 of table 10 depicts the empirical results of this regression. As can be appreciated, three of the four variables that are significant in the main results (column 1), inequality legacy, education and trade, retain the significance level with the same sign and very similar coefficients. The p-values of the variables export share and manufacturing over GDP slightly change in a way that the export share becomes statistically significant, but manufacturing over GDP loses its significance. Therefore it can be concluded that inequality legacy, education and trade are not sensitive to changes in the definition of the independent variables and are important predictors of inequality. All in all, with these robustness test can be appreciated that on the one hand the explanatory power of inequality legacy remains when the dependent variable changes. On the other hand, the modification of one of the control variables does not bring important changes, keeping the influence of inequality legacy unchanged.
16 Within squared refers to how much of the variance within countries the model explains. Between squared refers to how much of the variance between the countries the model explains. Overall R-squared is the weighted average of both.
36
5. CONCLUSIONS AND LIMITATIONS OF THE WORK
This essay represents a comprehensive and different view of the income inequality problem in Latin America. It introduces a framework that attempts to identify the different sources of inequality for the different economic periods. It addresses the increasing inequality in the ISI period from a new perspective. It not only considers the income inequality that appeared within the ISI period itself, but also considers its inequality legacy.
Furthermore, it introduces powerful descriptive statistics that introduce a comprehensive comparison among export upgrading and industrial upgrading that brings interesting findings. Such a clear comparison, with the detailed descriptive statistics used, has not been previously introduced for Latin America.
Throughout the work, several are the conclusions that can be drawn. Not only has the empirical specification brought interesting conclusions, but also the descriptive statistics contribute with no less interesting results, besides being less biased.
In this context, from the theoretical part the following main conclusions can be drawn: 1. The sources of inequality are not the same in the export-led growth model and in
the import substitution industrialization model. In the first period, the increasing capital/labor ratio constituted the main source of income inequality in the so-called South Cone (Research question 1). It is also clear that the worsening of the income distribution in this period is only a characteristic of the South Cone countries and not of the entire Latin American Region. On the other hand, in the second period the differences that appeared among the labor income of those engaged in the modern industrial sector and those engaged in the rural agricultural sector constituted the main source of income inequality ( Research question 2)
37
another way, a successful export upgrading does not mean industrial upgrading for the Latin American scenario.
From the empirical analyses the following main conclusion is extracted:
The inequality legacy inherited from the ISI period appears to be a significant source of rising inequality for the trade liberalization period, this confirms the main hypothesis of the work. Therefore, and at least according to these empirical results, the industrialization strategy that LAC6 countries adopted in Latin America has brought negative consequences in the distribution of income. Consequences that not only have affected the ISI period itself, but also the subsequent economic period.
In addition, this work shows that a non-successful industrial development not only brings negative consequences for economic growth but also for the distribution of income. In this context, it is important to stress the key role of industrial upgrading in economic development. Furthermore, the recent economic history shows the importance of the state to guide this industrial development. Nevertheless, the involvement of the state in the process does not guarantee a successful path. The application of wrong policies might bring serious constraints for the industrial growth.
In this context, one of the main differences between the import substitution strategy adopted in Latin America and in Asia is the export promotion. It is clear that the non-promotion of manufacturing exports in Latin America subsequently harmed the industrial base. Thus, the gradual openness of the economy, as it took place in Asia, appears as a much more effective path that the abrupt change that LAC countries adopted.
Limitations of the work
38
The data issues and the quality of the empirical technique also limit the results. The Gini coefficient selected for the empirics, although, it is considered to be obtained from the most comprehensive database, presents its limitations, especially for the data of the 1980s, as earlier in methodology was stated.
Furthermore, this work does not include very complex econometric techniques. Probably, the large coefficient that the inequality legacy variable has in the model is the biggest limitation. This coefficient is not very reliable. It is believed that the variable used to measure the legacy of the ISI period overstates its influence. Ideally, other variable that properly captures the ISI legacy but that does not include in the variable the Gini index would have been used. Unfortunately, such a variable has not been found.
