On the Size of Government

It is well-known to students of international economics that wealthy economies typically operate under the rule of large governments. Given the conventional wisdom that large governments not only demand high rates of distortionary taxation, but also supply a plethora of distortionary subsidies, this is puzzling. One way to reconcile these observations is to suppose that large governments have a parasitic relationship with otherwise-efficient economies: it is simply not feasible for poor economies to maintain large governments—still, a large government is usually not, ceteris paribus, conducive to wealth creation.

To evaluate this hypothesis, it is appropriate to restrict our sample of economies so as to include only the wealthy. Doing so controls for many irrelevant sources of economic (in)efficiency. Based upon the IMF's list of developed economies, IMF data on GDP (PPP) per capita, and Heritage Foundation/Wall Street Journal data on public sector share of GDP, the relationship for 2011 is displayed below:

A simple linear regression yields a decrease of $385.30 of per-capita income for every percentage-point increase in the relative size of government (R-squared: 0.099; p-value: 0.069). The fit is on the cusp of statistical significance, but the chart makes it seem that the relationship (or lack thereof) is primarily driven by a set of influential points. On the extreme left, we find the four Asian Tigers (Hong Kong SAR, Korea, Singapore, and Taiwan Province of China); on top we find Luxembourg. To determine the extent of their influence, first consider the fit excluding Luxembourg:

A second simple linear regression yields a decrease of $327.50 of per-capita income for every percentage-point increase in the relative size of government (R-squared: 0.125; p-value: 0.044). The fit is statistically significant, but why exclude Luxembourg from our sample? Wikipedia reports the following:

In April 2009, concern about Luxembourg's banking secrecy laws, as well as its reputation as a tax haven, led to its being added to a "grey list" of nations with questionable banking arrangements by the G20. In response, the country soon after adopted OECD standards on exchange of information and was subsequently added into the category of 'jurisdictions that have substantially implemented the internationally agreed tax standard.' In March 2010, the Sunday Telegraph reported that most of Kim Jong-Il's $4bn in secret accounts is in Luxembourg banks.

At a minimum, it is safe to say that Luxembourg is unusual. What if we instead retain Luxembourg in our sample, but exclude the four Asian Tigers?

A third simple linear regression yields a decrease of $306.60 of per-capita income for every percentage-point increase in the relative size of government (R-squared: 0.026; p-value: 0.392). The relationship disappears when we exclude the four Asian Tigers. Finally, consider the fit excluding both:

A fourth simple linear regression yields a decrease of $16.90 of per-capita income for every percentage-point increase in the relative size of government (R-squared: < 0.001; p-value: 0.948). Nothing to see here; just statistical noise.

This analysis suggests a few tentative conclusions. First, whatever signal may be extracted from the noise, it pertains exclusively to the extreme four Asian Tigers on the one hand, and unusual Luxembourg on the other. Second, the relationship may stem from other features of these economies, related to but distinct from the size of government (e.g., benign regulatory environments may be correlated with small governments). Third, to the extent that size of government is causally significant, it seems that its (independent) effect is only tangible when contrasting opposing extremes.

In much of the developed world, politics leaves little room for reducing the relative size of government. Small victories on this front, in light of the preceding considerations, offer negligible hope for improving the quality of life in society. Growth enthusiasts would be wise to reconsider their policy priorities.