Government Size and Trade Openness: Some Additional Insights

A body of influential research has suggested that there is a positive association between trade openness and government size. Cameron (1978), one of the first to establish econometric evidence on the topic, noted that trade openness in 1960 was a strong predictor of the increase in government tax revenues between 1960 and 1975. He pointed out that more open countries tend to be more unionised, with collective bargaining leading to greater demand for social protection accommodated by increasing tax revenues. This pioneering version of the compensation hypothesis – by which more open countries tend to have bigger public sectors – was reappraised and further articulated by Rodrik (1998). While challenging the collective bargaining explanation, Rodrik argued that government spending might serve as an indirect insurance against external (and undiversified) risk. His most influential result was to find a positive association between government consumption and trade integration in a large sample of countries that qualifies openness both as a determinant and as a predictor of government consumption levels across countries (Rodrik, 1998; 1004). (Note 1) This conclusion would suggest a strong complementarity between markets and governments, with a more powerful role for government consumption in those economies that are subject to larger external risks.

In an influential work, Alesina and Wacziarg (1998) (henceforth AW) have challenged the Rodrik’s hypothesis, by arguing that the positive relation between openness and government size could be mediated by country size. The first reason is that country size is negatively correlated to government size, as the costs of certain (non-rival) public goods grow less than proportionally to the size of population. This is typical, for example, of infrastructures, roads, libraries (at least up to the congestion limit) and implies that the per capita cost of public goods declines in larger countries. The second reason is that country size is also negatively correlated to trade openness, as small countries have less opportunity for autarky. As argued by AW (p. 306), these two facts taken together imply that more open countries may have bigger governments.

This has cast some doubts on the existence of a Rodrik-type direct link between openness and government size. Consistently with the two hypotheses, AW – by running OLS on 1980-84 averages for the same set of countries used by Rodrik (1998) – actually find a negative relation between government consumption and population (taken as a proxy of country size) and a negative relation between trade openness and population. In both cases, the log of population exhibits a highly significant negative coefficient, and the result appears robust not only to a parsimonious specification of explanatory variables, but also to an extension of the basic model to control for possibly omitted variables.

Then, in order to capture the impact of country size on the co-variation between government consumption and trade openness, the authors move to the estimation of the basic Rodrik’s specification, where country size is not included among the explanatory variables (as in Table 1 in Rodrik, 1998) and replicate the Rodrik’s result of a positive association between government consumption and trade openness. By omitting trade openness and including country size the negative relation between country size and population is also confirmed. When including both (trade openness and country size), the positive impact of trade openness persists, that confirms the Rodrik’s result. However, AW impute this persistence to the high degree of collinearity between openness and country size. Thus, they experiment a version of the regression where variables calculated as ratios are included in levels and not in logs. In this case, the positive relation between openness and government consumption disappears, showing that the original Rodrik’s result might be driven by the omission of country size. (Note 2)

More recently, Ram (2009) (henceforth R) has challenged the outcome of AW mainly on the econometric ground. Considering 154 countries for the period 1960-2000, R shows that while pooled OLS regressions replicate the results of AW, a fixed effect estimation that takes into account cross-country heterogeneity would not lead to a significant negative co-variation of country size and either trade openness or government size (p. 213). Thus, the estimates by R would be consistent with a direct link between openness and government size along the lines suggested by Rodrik (1998), instead of being mediated by country size as argued by AW. In R, the compensation hypothesis would indeed be supported by the positive sign of trade in all specifications.

This paper sheds additional light on these issues. In particular, it will compare the results by AW and R with those obtained by an updated panel analysis in the period 1962-2009, with data taken from the Penn World Tables 7.0 (PWT) and from the World Development Indicators (WDI). We show that the sign of the relationship between government size and economic openness is not necessarily driven by country size. More importantly, the results obtained by fixed effects as in Ram (2009) show that the compensation hypothesis strictly depends on the inclusion of African countries, further weakening the general validity of the compensation hypothesis.

For full text: click here

(The original author: Paolo Liberati

Published by Sciedu Press)