The paper examines how much governance matters for long run economic development in poor countries. Answering this question confronts four methodological challenges. First, since growth equations are likely to be underspecified, unobserved effects are likely to render standard estimators biased and inconsistent. Second, measurement error, time-varying unobserved effects, and possible feedback from growth into governance will violate the exogeneity assumptions of many estimators. Bias and inconsistency is again the consequence. Third, use of cross-country evidence opens the possibility that the interaction between governance and growth shows considerable country-specificity. In the presence of heterogeneity, estimation under a specification that imposes homogeneity on the governance – growth link, will again result in bias and inconsistency. Finally, an assumption of linearity in the growthâ€governance relationship across large ranges of per capita income and differences in institutional structure is at least questionable. The paper addresses these problems by contrasting results obtained under pooled OLS, fixed effects, GMM, country specific time series and PMG estimators. Results confirm that all four methodological concerns are valid. Taking account of the problems renders estimates of the impact of governance more robust, and serves to increase its impact. Our best estimate of the impact of improving rights on the level of real per capita output is that this differs between countries with good, mid-range and poor rights. The implied elasticities of a benevolent impact on the level of real per capita output of improving rights ranging from 0.28 to 0.22 for countries with the worst rights, and 0.07-0.02 for countries with the best rights. Countries with midrange rights have a perverse association between improving rights and output, though the estimated elasticity range is relatively weak over the 0.03-0.02 range. Improving rights have an indirect impact on output as well a direct one. Improving rights serve to increase the productivity of investment, with the elasticity of output with respect to investment rising from 0.29 to 0.45 between countries with the worst rights, and those with midâ€range rights (thus improving its productivity by a factor of 1.6). The impact of economic policy on output similarly improves under improved governance – though more weakly than the productivity of investment – with both the openness and the anti-inflationary policy stance of the economy proving to be significant. Investment in human capital by contrast has an impact on output that is invariant to the level of rights. The estimated elasticity of education on output reaches 0.44 at an average of eleven years of schooling, and unitary elasticity at an average of 25 years of schooling. These results find a significant echo, but also some nuance in equations specified in terms of economic growth. Substantively, our estimation results confirm an increasing productivity of investment for growth purposes under rising governance, and they are consistent with rising levels of investment under improving governance. Further evidence in support of the impact of good governance comes from the fact that physical capital, human capital, openness of the economy, foreign direct investment and anti-inflationary economic policy all further spur growth – often dramatically so. By contrast, for midâ€range rights countries only investment in physical capital raises economic growth (though strongly so), while for poor rights countries empirical results prove mixed at best. The implications are twofold. Under good governance policy makers have the gamut of standard policy handles at their disposal in promoting growth. Their task is correspondingly easier. And on a methodological note, the finding of strong heterogeneity across countries serves to offer an explanation of why the literature may have struggled to isolate particularly robust results under cross sectional and insufficiently sophisticated panel estimators.