«Abstract We examine the effects of precolonial and colonial legacies on the current economic growth rates of ex-colonies. We ﬁnd that precolonial ...»
8 with a lower settler mortality and a lower population density. Furthermore, they argue that the number of European settlers affected institutions in ex-colonies. The ex-colonies with larger numbers of settlers created “inclusive institutions,” similar to those of European countries, that emphasized private property rights. On the other hand, the ex-colonies with fewer immigrants maintained “extractive institutions” that were more oriented to extracting natural resources. Moreover, these different ruling systems affected economic activities even after the ex-colonies were politically independent from their ruling countries. The ex-colonies with institutions favorable for private property rights could continue promoting a high level of economic activities. By this hypothesis, they place the United States, Canada, Australia, and New Zealand at one extreme, and Peru and Bolivia at the opposite extreme. Note that the latter countries belonged to the Inca civilization and enjoyed the returns of a higher standard of civilization than was the case in the former countries. In essence, the ex-colonies with few settlers became poor, whereas the ex-colonies with a greater number of immigrants became rich. Acemoglu, Johnson, and Robinson (2001, 2002) point out this change of status, and refer to it as the reversal of fortune.
We accommodate this hypothesis by including the immigrant ratio as one of the regressors for the current economic growth rate. Acemoglu, Johnson, and Robinson (2001) provide most of the data observations needed for this study, measured in 1900. Korean data are collected from Etemand (2000). Here, the immigration ratio is deﬁned as the ratio of Japanese residents in Korea to the total population of Korea in 1913.
According to Table 1, the average immigrant ratio in the 1900s is about 11%. The so-called Neo-European countries, such as the United States, Canada, Australia, and New Zealand, had immigrant ratios greater than 87% (87.5% for the United States). On the other hand, 55 of 102 countries had very small immigrant ratios of less than 1%. Most ex-colonies had immigrant ratios less than 2%, including Korea.6 On the other hand, Argentina, Brazil, and Venezuela had immigrant ratios as high as 60%, 40%, and 20%, respectively.
Next, we estimate the correlation coefﬁcients between the log of the immigrant ratio and the precolonial legacy variables. The estimation results are reported in Table 2. Here, we can see that the log of the population density is negatively correlated with the log of the immigrant ratio. This result reﬂects what Acemoglu, Johnson, and Robinson (2001, 2002) point out, namely that more Europeans immigrated to the ex-colonies with lower population densities. The log of the immigrant ratio is also negatively correlated with other precolonial legacy variables. According to the hypothesis, the ex-colonies with fewer immigrants developed political ruling systems to exploit natural resources. The estimated negative correlation coefﬁcients imply that the negative effects of colonial rule may dominate the positive effects of the precolonial legacies.
6 There are several different statistics on the ratio, although they are more or less similar. The statistics yearbook of the colonial government in Korea reports that the immigrant ratio was 1.76% and 2.00% in 1913 and 1920, respectively. According to Etemand (2000), the ratio is 1.8% in 1913. Kim (2006) estimates the ratio and reports 1.58% and 1.78% in 1913 and 1920, respectively.
9 Furthermore, the immigrant ratio alone cannot distinguish between the positive and negative effects of the colonial legacies, although immigration must have been a signiﬁcant channel through which positive and/or negative legacies were left to the ex-colonies, as noted by Acemoglu, Johnson, and Robinson (2001, 2002) and Easterly and Levine (2012).
Therefore, we include the colonial duration variable as our next colonial legacy variable, and use it to supplement the immigrant ratio in measuring the effects of colonial legacies. By including the colonial duration variable, we intend to control for possibly existing negative effects. For the ex-colonies with fewer immigrants, a longer period of colonial rule means that the negative effects are more likely to be revealed by the colonial duration variable.
The current study is not the ﬁrst attempt to relate the colonial duration variable to current economic growth. Prior studies report different empirical ﬁndings. Grier (1999) empirically shows that current economic growth is positively related with the colonial duration variable, but without controlling for the number of immigrants and the other variables that distinguish between ruling systems. Note that the immigrant ratio is likely to increase if the colonial duration increases, as can be veriﬁed by a simple regression.7 On the other hand, Price (2003) empirically ﬁnds that the current economic growth rate is negatively or nonsigniﬁcantly correlated with the colonial duration variable by restricting data observations to African ex-colonies only.
This ﬁnding differs to that of Grier (1999) and implies that a more precise estimation result is obtainable by controlling for precolonial and colonial legacy variables at the same time, and an omitted variable bias can be avoided.
We report the descriptive statistics of the colonial duration variable in Table 1. Most colonial duration data observations are available in Grier (1999). We use her data for all countries except Australia and New Zealand, which were ofﬁcially and politically independent from the United Kingdom in 1901 and 1907, respectively (e.g., Chanda and Putterman, 2007; Olsson, 2009). We let these independence years be our data observations by following Chanda and Putterman (2007) and Olsson (2009).8 The ex-colonies experienced 158 years of colonial rule, on average, but with much variation. The longest duration is 513 years (Cape Verdi), and the second longest is 498 years (Guinea-Bissau). On the other hand, the shortest duration is Iraq, with 15 years. The colonial duration of Korea is 35 years. Another aspect of the colonial duration is revealed by sorting them according to the continents to which the ex-colonies belong. The North and South American ex-colonies averaged 293 and 248 duration years, respectively. The next longest duration occurred in Africa (107 years on average). Asian countries experienced 56 years on average. In terms of 7 The coefﬁcient is estimated as 0.64 with a t-statistic of 8.29.
8 This modiﬁcation does not yield different conclusions.
10 colonial ruling countries, the countries with the longest and shortest ruling periods are Portugal and France, respectively. This additional information may also be used to distinguish between the positive and negative effects of colonial legacies.
As the ﬁnal colonial legacy variable, we consider the year of independence. Easterly and Levine (2003,
2012) note that the sooner the independence year, the higher is the expected economic growth, because an established political and economic system is built over time. Thus, earlier independence years are likely to lead to higher levels of economic growth. Table 1 reports the descriptive statistics of the independence year. The earliest independence year is 1776, when the United States became independent from the United Kingdom. The latest independence year is 1997, when Hong Kong became independent from the United Kingdom. Most ex-colonial countries were independent in the early 1960s after World War II, and the colonial history started when Christopher Columbus discovered the new continent. The main interest here lies in examining how different independence years affect the current economic growth rate after controlling for the other precolonial and colonial legacy variables.
All data sources and detailed information are provided in the Appendix.
3 Models and Estimation Results
We specify a linear model for the economic growth following the work of Barro (1991):
where Yi denotes the economic growth rate of the i-th country; Ai is the set of variables that are conventionally included for the regression of economic growth; Bi is the set of variables that are included to examine how they affect the economic growth rate; and Ci is the set of variables that are already well known to potentially affect the economic growth rate. More speciﬁcally, we let A include the per capita GDP in 1960 and the primary school enrollment rate or schooling years.9 These variables are typically used to explain the economic growth rate using Barro’s (1991) conditional convergence hypothesis and the human capital theory. We follow the same convention. The variable set B includes one of the precolonial legacy variables in Table 1 and other colonial legacy variables, namely colonial duration years, the immigrant ratio, and the year of independence. Since our main interests here are in examining whether these variables affect the curAlthough these two variables are used as proxies for human capital, they have different properties. The primary school enrollment rate has an upper bound of 100%. If this bound is reached, it can no longer discriminate the quality of human capital. On the other hand, schooling years does not have this restriction. Furthermore, the sample size of the school enrollment rate is greater than that of the schooling years. In this study, we use both, although this does not change our ﬁndings.
11 rent economic growth rate, the variables in B are the core explanatory variables in terms of the goals of this study. Finally, the variable set C includes continent dummies, colonial ruler dummies, legal origin dummy, geographical variables such as latitude, and other macroeconomic variables such as investment rate, income, or asset distributive index. These are other factors that may affect economic growth or are used to control for the other factors. A number of prior studies use these as factors that affect economic growth (e.g., Levine and Renelt, 1992). Thus, we use them to verify whether our estimations are robust.
Insert Tables 3 and 4 around here.
The estimation results are reported in Tables 3 and 4, which are obtained by separately estimating the model using the average of schooling years and the primary school enrollment rate as proxies for human capital, respectively. The precolonial legacy is measured by the proxy variable, technology level AD 1500.
We brieﬂy summarize the estimation results. First, Barro’s (1991) conditional convergence hypothesis ∗ ∗ is afﬁrmed by our estimations. The signs of β1 and β2 are obtained as predicted. Second, the R2 value of each model is relatively high. About 3/4 of the total variation in the economic growth rates is explained by these models. Third, all estimated parameters are signiﬁcant, except the independence year in Table 4. The coefﬁcients of the technology level AD 1500 and colonial duration variable are signiﬁcant at the 1% level.
Insert Table 5 around here.
Before examining the estimates in more detail, we predict the economic growth rate using the model estimates. Table 5 shows the prediction results of the Korean economic growth rate. According to the estimates in Tables 3 and 4, the predicted growth rates are 5.54% and 5.53%, respectively. The realized economic growth rate was 5.78%. Thus, the model estimates imply that the Korean economic growth rate is not an outlier. This further implies that there was no secret method that led to the high economic growth in Korea. This interpretation can also explain the low economic growth rates of the Sub-Saharan African countries. Table 5 shows that their economic growth rate is well predicted by the model estimates in Tables 3 and 4. These predictions show that our models efﬁciently explain the high economic growth rate of Korea and the low growth rate of the Sub-Saharan African countries.
We now examine the model estimation results in more detail. We focus on Table 3 because Table 4 can be examined in a similar way. First, there is a positive relationship between the economic growth rate and the technology level AD 1500 variable. The difference between Models 2 and 3 lies in the addition of the technology level AD 1500 variable to the GDP per capita and schooling years. After this addition, the associated R2 value more than doubles, from 0.35 to 0.65. Then, Model 4 adds the colonial duration 12 and independence year variables to the explanatory variables in Model 3, and Model 5 adds the log of the immigrant ratio to the variables in Model 4. In all cases, the estimated coefﬁcient of the technology level AD 1500 variable is positive and signiﬁcant at the 1% level. This aspect is more evident if we estimate the model using the two-stage least squares (2SLS) estimation method. Here, we employ the state antiquity index AD 500 and the date of agriculture as our instrumental variables. This enables us to correct any existing bias resulting from the endogenous variables and to avoid potential problems of reverse causality.
Model 7 shows the estimation results. The estimated coefﬁcient of the technology level AD 1500 variable is greater than those in the previous models. The F -statistic of the ﬁrst-stage regression is signiﬁcant, and there is no evidence of over-identiﬁcation based on Hansen’s (1982) J-test statistic. This implies that the two instrumental variables do not affect the economic growth rate, other than through the technology level AD 1500 variable.10 Table 3 also estimates Model 8 using the worldwide samples. Here, the estimated coefﬁcient of the technology level AD 1500 variable decreases signiﬁcantly, although it is still positive. The ex-colonies must have beneﬁted more from the precolonial legacies than did those colonies without colonial experiences. In summary, the precolonial legacies must have signiﬁcantly affected the current economic growth of the ex-colonies: the precolonial legacy variables cannot be ignored in explaining the economic growth rates of the ex-colonies, even after these countries were decolonized.