«Abstract We examine the effects of precolonial and colonial legacies on the current economic growth rates of ex-colonies. We ﬁnd that precolonial ...»
We now compute the marginal contribution of the technology level AD 1500 variable to current economic growth by focusing on the Korean and Sub-Saharan African economies. The other variables are assumed to be the same, but the technology level AD 1500 variable is assumed to be the average technology level of all ex-colonies, namely 0.38. The actual Korean technology level was 0.85 in AD 1500. By this modiﬁcation, the economic growth rate is lowered to 2.94% and 3.01% according to the estimates in Model 7 of Table 3 and Model 6 of Table 4, respectively.11 These predictions are similar and lower than the realized economic growth rate of 5.78%. This implies that the precolonial legacies played signiﬁcant roles in the Korean economic growth rate. In other words, the high economic growth rate in Korea cannot be explained without the country’s precolonial legacies. As another example, we consider the Sub-Saharan African countries. Their average economic growth rate and technology level AD 1500 are about 0.59% and 0.27%, respectively. If their technology level AD 1500 values are modiﬁed to 0.85 (i.e., the level of Korea), the predicted economic growth rate becomes 3.79%.12 This prediction is similar to that of Malaysia and ranks 5th, implying that the low economic growth rates of the Sub-Saharan African countries are mainly 10 The ﬁrst-stage least squares estimation shows that the instrumental variables are signiﬁcant at the 5% level.
11 These ﬁgures are obtained as follows: 5.54 + 5.51×(0.38 − 0.85) and 5.54 + 5.39×(0.38 − 0.85), respectively.
12 This ﬁgure is obtained as 0.57 + 5.51 × (0.85 − 0.27).
13 due to the low precolonial legacy levels. Figure 1 shows the partial effect of the technology level AD 1500 variable on economic growth using the estimates in Model 5 of Table 4, which show a high positive correlation between the two.
Next, we examine the effects of the colonial legacy variables on the current economic growth rate. From the estimation results, the effects of the colonial legacy variables are uncertain. According to the estimation results of Model 6 in Table 3, the coefﬁcient of the log of the immigrant ratio is positive, as expected. On the other hand, the coefﬁcient of the colonial duration variable is negative. This means that the colonial duration variable captures the negative effects of colonial experiences. To examine these two opposite aspects in more detail, we compare Models 4 and 6 in Table 3. The coefﬁcient of the colonial duration variable is −0.59 in Model 4, and −0.64 in Model 6, in which we control for the immigrant ratio. That is, the negative effects of the colonial duration variable are aggravated by controlling for the immigrant ratio. In the same way, the positive effects of the immigrant ratio are ampliﬁed after controlling for the colonial duration variable.
Although Model 5 estimates the coefﬁcient of the immigrant ratio to be nonsigniﬁcant at the 5% level, it becomes signiﬁcant after simultaneously controlling for the colonial duration and the year of independence variables. This is evident if we compare the estimation results in Models 4 and 5. Therefore, we conclude that the colonial legacies must have had two opposite effects on the current economic growth. Figure 2 shows the partial effect of the immigrant ratio on economic growth, obtained by controlling for the other variables in Model 5 of Table 4. In the same vein, Figure 3 shows the partial effect of the colonial duration, obtained by controlling for the other variables in Model 6 of Table 3 and the continent dummies. These ﬁgures afﬁrm the positive and negative effects of the immigrant ratio and colonial duration, respectively.
These estimation results imply that the overall effect of the colonial legacy variables is determined more by a particular circumstance in each country. For example, the immigrant ratios of Neo-European countries are high. Thus, their current economic growth may have beneﬁted from their colonial legacies. On the other hand, the Sub-Saharan African countries, with few immigrants and long colonial durations, could not exploit the legacies, as pointed out by Heldring and Robinson (2012).
Although these estimation results are similar to those of Acemoglu, Johnson, and Robinson (2001, 2002) and Easterly and Levine (2012), our estimations are not identical to theirs. They do not separate the negative and positive effects, although they admit that the colonial experience has opposite effects on economic growth, as shown here.
14 We further afﬁrm the estimation results by estimating the counterfactual economic growth rates of the ex-colonies. If the United States had not been a colony, its counterfactual economic growth rate is 0.71%13, which is obtained by letting the colonial duration and the log of immigrant ratio be zero in Model 6 of Table
4. This is about 28∼29% level of the realized economic growth rate. On the other hand, if the Sub-Saharan African countries were not colonies, their counterfactual average growth rate becomes 0.90%14 using Model 7 of Table 3, which is higher than the realized growth rate by about 1.6 times. We further compute the counterfactual economic growth rate of Korea in a similar manner. According to Model 7 of Table 3 and Model 6 of Table 4, the counterfactual economic growth rates are about 5.51% and 5.37%, respectively.
These ﬁgures are smaller than those predicted by the model estimation results by 0.02% and 0.14% points, respectively.15 This result implies that the portion of the colonial legacies that have contributed to the Korean economic growth rate is between 0.34% (=0.02/5.78) and 2.42% (=0.14/5.78) from 1960 to 2000,16 which is not signiﬁcantly different from zero. In other words, in the case of Korea, the high economic growth could be achieved even without the colonial experience, which is different to the United States and the Sub-Saharan African countries.
4 Robustness Tests
In this section, we conduct robustness tests for the models estimated in Section 3. These tests are motivated by the concerns that there are possibly omitted variables in the economic growth rate model or that the technology level AD 1500 variable may not be a proper proxy for the precolonial legacies. To handle these problems, we employ possibly omitted variables that affect the precolonial and colonial legacies, such as continent dummies, colonial ruler dummies, geographical variables, macroeconomic variables, and the income and wealth distributive index. After adding these variables, we estimate the models again and examine whether our estimation results show signiﬁcant changes. Furthermore, we consider other variables 13 This ﬁgure is obtained as 2.54 + 0.49 × 1.56 − 0.58 × 4.47, where 2.54 is the model predicted growth rate, 1.56 is the colonial duration years/100, 4.47 is the log (immigrant ratio), and the others are the estimated coefﬁcients from Model 6 of Table 4.
14 This ﬁgure is obtained as 0.57 + 0.66 × 0.87 − 0.43 × 0.56, where 0.57 is the model predicted growth rate, 0.87 is the colonial duration years/100, 0.56 is the log(immigrant ratio), and the others are the estimated coefﬁcients from Model 7 of Table 3.
15 By Model 7 of Table 3, the counterfactual economic growth rate is obtained as 5.53 + 0.66 × 0.35 − 0.43 × 0.59, where
5.53 is the model prediction, 0.35 is the colonial duration years/100, and 0.59 is the log(immigrant ratio). The others are model coefﬁcients. Model 6 of Table 4 computes the counterfactual economic growth rate as 5.54 + 0.49 × 0.35 − 0.58 × 0.59. The ﬁgures are deﬁned in the same way as for Model 7 of Table 3.
16 These ﬁgures are computed as 0.02/5.78 and 0.14/5.78, respectively, and do not accommodate the possible effects of the human capital that were accumulated in Korea during the colonial period. According to Furukawa (1990), the primary school enrollment rate of Korea was 51% just before becoming independent from Japan, and increased to 95% from 1956 to 1960. The Japanese colonial government paid far less attention to raising human capital in Korea than did other colonial rulers, given that the average primary school enrollment rate of the ex-colonies was 66% in the 1960s. Refer to Woo and Kahm (2011) for empirical investigations of the Korean primary school enrollment rates between 1930 and 1960.
4.1 Robustness Tests for Possibly Omitted Variable Bias In this subsection, we correct for a possibly existing omitted variable bias. First, Model 1 of Table 6 modiﬁes the estimation results by adding continent dummies to Model 6 of Table 3. Here, we consider the continents of Asia, Africa, the North and South America. Model 2 adds the colonial ruler dummy variable instead of the continent dummies. The colonial ruler dummy is deﬁned as one if the United Kingdom, France, or Spain are the colonial rulers, and zero otherwise. Model 3 includes the colonial ruler dummy and continent dummies together. Model 4 replaces the year of independence with the year of colonization.
Insert Table 6 around here.
These modiﬁcations affect our estimations, and we summarize their implications as follows. First, none of these modiﬁcations change the implications of the previous estimations. The estimated coefﬁcients for the technology level AD 1500 and colonial duration variables maintain the same signs and remain signiﬁcant. Second, the colonial ruler dummy might have affected the model estimation in a different way. More speciﬁcally, North (1990) and Landes (1998) point out that the ex-colonies ruled by the United Kingdom show higher economic growth rates than those ruled by France and Spain. They explain that this is mainly because the United Kingdom had more favorable institutions for private entrepreneurship, trade, and innovation than did the other countries. On the other hand, Grier (1999) notes that the educational aspects of the United Kingdom were superior to those of the other countries, which contributed to economic growth more than the qualities of institutions did. However, Acemoglu, Johnson, and Robinson (2001), criticize these views and emphasize that the disease environments of the ex-colonies ruled by the United Kingdom attracted more Europeans than did other ex-colonies. As a result, more Europeans emigrated to the UK ex-colonies.17 Note that the colonial ruler dummies are signiﬁcant for Models 2 to 4. We further exclude Korea, Brazil, and Indonesia from the samples and estimate the models again. These countries are eliminated because most ex-colonies were ruled by the United Kingdom, France, and Spain (and mainly the United Kingdom and France). Thus, the estimation results may be different if the ex-colonies were ruled by these countries. Furthermore, Korea showed the highest growth rate, so it might be an outlier in terms of the colonial ruler dummy. Nevertheless, Models 5 and 6 deliver the same qualitative estimation results, and we ﬁnd no signiﬁcantly different coefﬁcient values among the colonial ruler dummies. Models 7 and 8 17 The average immigrant ratio of 23 French ex-colonies is 0.96%, whereas that of 37 UK ex-colonies is 2.29%. Here, the United States, Canada, Australia, and New Zealand are excluded from the 37 samples.
16 show the estimation results obtained after excluding the United States, Canada, Australia, and New Zealand.
These countries are different to the others because they were developed as settler colonies. The estimation results in Models 7 and 8 are not signiﬁcantly different. Third, existing literature has different views on how the legal origin affected the economic growth rate. According to La Porta, Lopez-de-Silane, Shleifer, and Vishny (1997), La Porta Lopez-de-Silane, and Shleifer (1998, 2008) and Mahoney (2001), the common law system of the United Kingdom brings about more favorable environments for economic activities than does the French civil law system. On the other hand, Acemoglu and Johnson (2005) and Klerman, Mahoney, Spamann, and Weinstein (2011) argue that there is no direct interrelationship between the legal origin and economic growth. We examine these different views by estimating Model 9 and by restricting the samples to the common law and civil law systems only.18 According to our estimation of Model 9, we ﬁnd no evidence that the common law system is superior to the civil law system in terms of economic growth. Furthermore, the legal origin dummy does not modify the interpretations of our prior estimation results.
Insert Table 7 around here.
Next, we estimate the model using other proxies for precolonial legacies. The results are reported in Table 7. Model 1 replaces the technology level AD 1500 variable with the urbanization index AD 1500 proposed by Chandler (1987). Models 2 and 3 are estimated using the population density AD 1500 variable provided by Acemoglu, Johnson, and Robinson (2002) and Bandyopadhyay and Green (2012), respectively.
Acemoglu, Johnson, and Robinson (2002) estimate the population density using the population in McEvedy and Jones (1978) and dividing it by the area of total arable land in McEvedy and Jones (1978). Bandyopadhyay and Green (2012) divide the same population by the area of potentially arable land. Models 4 and 5 use the state antiquity index AD 1950 suggested by Putterman (2007) and the product of the state antiquity index AD 1950 and the population density as another proxy variable, respectively. These estimations show that the economic growth rate is positively and signiﬁcantly related with these proxy variables at the 1% level. Furthermore, the interrelationship between economic growth and the colonial duration variable is negatively signiﬁcant, even after these replacements. Except for Model 2, these negative relationships are signiﬁcant at the 1% level, and Model 2 reports a signiﬁcant relationship at the 10% level.
One of the interesting aspects of these model estimation results can be found in Models 6 and 7 of Table
7. These two models use the state antiquity index AD 50 and the technology level 1000 BC as proxies for the precolonial legacies, respectively. From these, we can see that the proxies still signiﬁcantly affect the economic growth rate, even though they were formed more than 2,000 and 3,000 years ago, respectively.
18 Korea adopts the German law system. Therefore, we omit Korea from this sample.
17 Figures 4 and 5 show the partial effects between the growth rate and these two proxy variables, respectively.