Enclaves and Development: An Empirical Assessment

For instance, a search of JSTOR indexed economics journals reveals just two references in general economics journals (once in an article written by political scientists, and one brief mention in a survey of “radical” literature) and only a handful of other references in field journals.

Though slavery was important in the South, many of the key US institutions were formed in the seventeenth century before slavery became more significant and at no time did slaves form more than 20% of the entire population. In contrast, indigenous peoples formed 80–90% of the populations of Peru, Bolivia, or Mexico, while slaves constituted more than 90% of the population in most Caribbean islands (Engerman and Sokoloff 1997).

We are aware, of course, of an existing literature that attempts to relate various measures of “dependency” to economic performance outcomes, for instance, the work of Chase-Dunn (1975) and Evans and Timberlake (1980). Aside from working with more recent datasets, one key advantage to our approach of using pre-1930 enclave status is that it can be more plausibly considered exogenous, or at least pre-determined, relative to present day outcomes compared to the more contemporaneous measures of ‘dependency’ or foreign capital penetration used in these other studies. Also, the book of Cardoso and Faletto makes clear that it is in the period prior to 1930 that the enclave status of a country was determined.

There seem to be many interpretations in the literature about what “control” means. Many argue that what Cardoso and Faletto “really meant” was linkages in the sense of Hirschman (1958). Yet this word is not mentioned in the book, despite the fact that Hirschman’s work was well known by the time the first version was written, and we do not find the arguments in the book closely linked to the notion of linkages.

See Hojman (1983) for a further and more critical exposition of the theoretical and empirical difficulties of defining and measuring enclave status.

The quality grades are determined as follows. One for observations (a) where the underlying concepts are known and (b) where the quality of the income concept and the survey can be judged as sufficient according to the criteria described above. Two for observations where the quality of either the income concept or the survey is problematic or unknown or we have not been able to verify the estimates (the sources were not available to us); the country information. For further details, see http://​62.​237.​131.​23/​wiid/​WIID2c.​pdf.

For a couple of countries in Sub-Saharan Africa, there is no observation that is rated either 1 or 2. In this case, we took the most recent observation available. The results we report are robust to dropping these observations.

It is not ideal to treat enclave status as exogenous since there may be omitted variables that influence whether or not a country becomes an enclave and are correlated with institutions. If this is so, then the estimated coefficient on the enclave variable will be biased. Ideally, we would also like to have an instrument for D _i and estimate a model with two first stages along the lines of Acemoglu and Johnson (2005). There are several potential candidates for such an instrument. Though Cardoso and Faletto do not dwell on why a country is or is not an enclave, their discussion suggests that the presence of a mining sector made a country more likely to be an enclave, and also, small countries were more likely to be enclaves, other things equal. Unfortunately, the presence of mines is not a satisfactory instrument for enclaves since it would not be excludable from the second stage. We did experiment with country size, which is negatively correlated with the propensity to be an enclave. Unfortunately, however, we did not find robust results using this strategy and, thus, do not report results using it here.