Category Archives: Institutions – Macro Evidence

DO INSTITUTIONS CAUSE GROWTH?

DO INSTITUTIONS CAUSE GROWTH?

E.L. Gleaser, R. La Porta, F. Lopez-de-Silanes & A. Shleifer

Journal of Economic Growth, 9 (2004)

A SUMMARY

In a Nutshell

Broadly speaking there are two schools of thought regarding institutions. Firstly, as epitomized by Douglas North and latterly Acemoglu, Jonson and Robinson, are those that argue the political institutions of limited government cause economic growth. Putting the right institutions in place will therefore create investment, human and physical capital and hence growth. The second school argues that the institutions of limited government are themselves caused by human capital development as essentially more educated people are more likely to resolve disputes by negotiation as opposed to violence, and literacy etc. means that people are better able to engage with politics, understand political wrongs etc. According to this thesis which has received support from Prezworski and Lipset, relatively unconstrained leaders pursue policies which alter the stock of human and social capital, and then as that stock increases the institutions of democracy etc. develop. This argument is related to the Olken & Jones article summarized above. The article supports the latter of these arguments.

It does so by seeking to address flaws in the literature relating to the former hypothesis. Firstly it argues that the measures of institutions as used in various institutional studies are not capturing real institutions, but rather the outcomes of political choices. Secondly they attack the instrumental variable analyses of amongst others AJR, by showing that the instruments used are correlated with human capital stocks both past and present, and thus exogeneity conditions are violated. In other words it appears the instruments are isolating not institutions but human capital. They run a variety of correlation analyses which all indicate that political institutions properly measured have little significant effect on growth, whereas human capital does.

Measurement on Institutions

Institutions as defined by North are the rules of the game that place constraints on individuals [executive politicians etc.] which are relatively durable and unchanging. This implies that it is constitutional arrangements, not policies that fall within the definition of institutions. However, the data used in institutional papers is largely survey data based on subjective assessments of expropriation risk, government effectiveness etc. which are outcome measures of policy choices not permanent institutional arrangements. This can be seen when it is shown that the measures employed both rise with income levels, and have a high variance. If institutional measures improve with income then they are not a good basis for establishing causality due to endogeneity problems.

They go on to look at correlations between true institutional variables – plurality, proportional representation, judicial independence and constitutional review, and find that they are not generally correlated with income, and they are only weakly correlated with the outcome indices traditionally used.

The bottom line is thus that the measures usually used are not suitable for establishing causation as they do not reflect constraints on governments nor permanent features of the political system, but rather volatile outcome measures of policies pursued by those in charge.

Human Capital

In regression analysis initial levels of human capital in 1960 is a much better predictor of subsequent growth than the true institutional variables noted above which all enter insignificantly. The traditional measures all still enter positively and significantly most likely reflecting the reverse causality issues. They conclude that in OLS regressions the evidence that institutions cause growth rather than the other way around is non-existent, and true measures of institutions have no predictive power even when averaged.

Politics and Growth Since 1960

Strong association between dictatorship and low levels of education. Countries with high human capital levels in 1960 have grown twice as fast as those with low levels.

Instrumental Variables

AJR use settler mortality in their groundbreaking paper. However, it is not clear that what the Europeans brought with them was their institutions rather than their human capital. Thus if settler patterns influence growth through channels other than institutions, the approach is fundamentally flawed.

They test the correlation between settler mortality and the constitutional measures of institutions as defined above and find only very weak correlations. i.e. settler mortality is not strongly correlated with the rules and procedures that actually govern the societies in question. Additionally the instrument appears to be correlated with the disease environment today [suggested by Sachs and friends to be a prime cause of underdevelopment], so the issue of the modern disease environment is real and should not be overlooked. The killer point is that settler mortality is strongly correlated with years of schooling both in 1960 and 2000, and the correlations are even stronger than between mortality and measures of institutions. Thus the instrument is not exogenous and is most likely proxying for human capital. They run a 2SLS regression of log GDP per capita in 2000 on human capital levels in 1960 and 200, as well as executive constraints, both instrumented for settler mortality etc. (which produce strong first stages). Whilst years of schooling are significant in the second stage, executive constraints are not which indicates that human capital measures are more important than institutional measures. In conclusion, they are not saying that human capital is all that is important, only that patterns of colonization have effected growth through a variety of channels. The IV approach does not tell us what causes growth.

From Schooling to Institutions

They supply indicative evidence that schooling leads to institutions rather than the other way around, by lagging school variables, which are significantly correlated with subsequent institutions. There is no significant correlation when the institutional variables are lagged as against the subsequent schooling variable.

Conclusion

It appears that communities face a set of institutional possibilities that are defined by levels of human capital. These possibilities increase with human capital. The institutions are persistent, but they can improve also with income. But in this framework institutions have only a second order effect, as it is human capital that shapes both institutional landscape and productive possibilities of a society.

Policy implications are that a focus solely on democratic institutions is flawed when pursuing growth. The right mix of policies can equally be chosen by a dictator. For countries to emerge from poverty it may be more important to increase the human capital that will subsequently cause institutions to improve.

 

Advertisements

DO LEADERS MATTER?

DO LEADERS MATTER? NATIONAL LEADERSHIP AND GROWTH SINCE WORLD WAR II

B.F. Jones & B.A. Olken

Quarterly Journal of Economics, Vol. 120, No. 3 (2005)

Principal Research Question and Key Result Are growth rates influenced by individual leaders? By exploiting exogenous variation in leadership the authors conclude that leaders do in deed matter for growth. This result holds only for changes in leaders in autocratic regimes, not in democratic regimes.
Theory Leaders may have some agency, but they are constrained by a historically determined set of choices from which they must make decisions according to Marx. A conflicting view is the “great man” view whereby outcomes are determined by highly idiosyncratic leadership. A more moderate view is that of Weber who claims that charismatic leadership can be deterministic in certain settings, but only when the institutional norms etc. do not constrain the actions of the leader. In other words it is the interaction between leaders and institutions that will determine the ability of the leader to influence outcomes.
Motivation Countries experience dramatic reversals of fortunes, either experiencing huge increases/decrease in growth. Given that factors such as institutions and human capital that are widely thought to be important determinants of growth are likely to be largely persistent over time, then there must be some other factor that can explain these observed growth shocks. If institutions are important then it could be that leaders in some sense substitute for, or partly control institutions and therefore exert personal influences on growth.
Data Growth indicators from Penn tables in a panel from 1945-2000 for countries who experienced a leadership transition due to natural/accidental death of the leader. This amounts to 57 leadership transitions. However, the estimation is made with all years and all countries as this helps to estimate the time fixed effects.
Strategy Leadership transitions are not exogenously determined as economic outcomes can help leaders to maintain power. Death is however an exogenous source of variation in leaders once assassinated leaders have been stripped out, so the authors use growth data in countries in which a leader has died of natural/accidental causes.They include dummies for PRE and POST which are true in the T years before/after the death of the leader and will pick up average growth rates in those years. Then Post/Pre represents that change in growth around the death of the leader, and the variance of that measure represents the average volatility of differences in growth outcomes surrounding leadership deaths. It is the variance in which we are most interested. Time and state fixed effects are included.

T is 5 years, and the actual year of death is not included so as to exclude turbulence only related to leadership succession.

Results
  • Variance in the POST-PRE coefficients is 31% higher around leadership transitions that it would be normally.  The actual difference in growth however does not seem to systematically increase nor decrease.
  • Using the Polity IV data set they codify countries into autocracies and democracies as it may be the case that the degree to which leaders matter is a function of their context i.e. different institutional structures allow for greater/lesser idiosyncratic effect of leaders. The results indicate that autocratic leaders have a significant effect on growth at all timings, whereas there is no significant effects for the death of democratic leaders.
  • Of the autocracies there are strong effects where there are no political parties, and no effects where there are political parties, and similarly for legislatures which indicates that leaders have more effect where institutions are weak. These distinctions are not simply being driven by income. If the poorest states could not afford good institutions then the effects would be larger for poorer states as institutions would be less of a constraint. This is not the case; the autocrat effects are most pronounced for middle income states indicating that extremely low or non-existent institutions in poor states may limit the ability of the leader to influence national outcomes.
Robustness
  • They check correlations between econ variables and the death of the leaders to check that death is not caused in some way by economic outcomes and find no convincing link. They also find that the dead leaders were an average of 8 years older than other leaders which is consistent with death being caused by complications due to ageing.
  • T is allowed to be 3 and 7  years
  • The timing T is also allowed to be t+1 and t+2 years after the actual death of the leader.
  • Pre and Post are shifted 5 and 6 years backwards as a placebo test and the null hypothesis that there is no change in the variance of Pre-Post is not rejected.
  • They run similar regressions for changes in monetary policy (measured by inflation rate), growth of government expenditures, growth in international trade, and the state of conflict in the country. Only the inflations rate is significant, and it appears to change most immediately around the death of the leader, and again the results only hold for autocrats, not democrats. Thus it appears that leaders can strongly affect monetary policy but no other growth oriented policies.
Problems
  • There is little theory to suggest exactly why leaders may matter for growth. They identify monetary policy as changing significantly around the death of a leader which could imply that it was through this channel that the leader was influencing growth, however, the implication is not proved. Also it might be questioned as to whether monetary policy really is the only channel through which leaders matter. For example, perhaps new leaders seek too cement approval by improving health or education benefits, and this is why growth is found to increase slightly after the death of an autocrat. Ultimately the paper tells us that leaders are important, but it is not able to state exactly why.
  • There are minor concerns about the exogeneity of the death if developing countries have worse healthcare/infrastructure and hence leaders are more likely to die in those places.
  • Other things could be happening in the transition period especially in autocratic states. If one leader is replaced with a successor more/less favourable to the West then the state may receive more/less overseas aid and other assistance, and this could increase growth and be driving the results.
  • What are the policy implications? The growth differentials are presumably because leaders are pursuing different policies, but what are they? Monetary policy is identified, but why could they not have tested for health, education, credit markets, R&D and all the other policy variables that are thought to determine growth outcomes.
Implications
  • The Solow growth model maintains that growth is determined by population growth, labour market growth, capital and technology. This article challenges that in some respects by showing that leaders also matter.
  • If the specification is correct it suggests that whilst institutions such having political parties and legislatures [which are arguable better institutional variables rather than property rights etc. which are outcomes rather than institutions – see Gleaser] are important for growth they are deterministic only to the extent that they limit the action of leaders. What is important for determining growth, holding constant those institutions are the policies pursued. The implications are therefore hard to assess, as what is really needed for growth are good policies imposed by benign dictators, but there are clearly logical and moral problems with this approach.