"... In PAGE 9: ... Accordingly, we focus the discussion on the joint relevance of each group of variables.25 Table1 , column (i), shows our baseline specification. Our dependent variable is a de facto fix dummy (LYSFIX) that takes a value of one if a country is classified as a de facto fix and zero otherwise.... ..."

"... In PAGE 22: ... Variables used in the productivity regressions] 4.3 Empirical results Table5 presents different specifications of a baseline equation in which MFP growth is regressed only on the industry leader, the technology-gap variable and human capital. From the discussion above, all specifications control for country and industry fixed effects.... In PAGE 23: ...51 The results also suggest a positive effect of human capital on MFP, as would be expected. [ Table5 . MFP regressions: selection of the baseline specification] In Table 6, we extended the analysis to check the influence of economy-wide privatisation patterns and product market regulations, looking at both their direct impact and their indirect influence through the rate of technological catch up.... ..."

"... In PAGE 10: ... Such added variables are potentially necessary in order to identify the independent effects of tax rates on reported AGI shares. We augment our baseline 1960-1997 specification with additional variables to investigate the extent of omitted variable bias, and Table2 presents the findings from this experiment. Our first candidate for an additional variable is a measure of worker productivity (see Column 1 of Table 2), which may lead to increased profits and increased income-reporting at the top end of the AGI distribution.... In PAGE 10: ... We augment our baseline 1960-1997 specification with additional variables to investigate the extent of omitted variable bias, and Table 2 presents the findings from this experiment. Our first candidate for an additional variable is a measure of worker productivity (see Column 1 of Table2 ), which may lead to increased profits and increased income-reporting at the top end of the AGI distribution. We also examine the importance of globalization by adding (separately, in Column 2) the international share of GDP, defined here as the sum of imports and exports divided by GDP.... In PAGE 10: ...7 Each of these three variables may have an impact on movements in the top AGI share above and beyond changes in tax rates. As observed in Table2 , these additional variables have very little if any impact on our overall findings. While the variables themselves are somewhat 6 Somewhat surprisingly, these results are virtually unchanged in the absence of the lag and lead differences in the tax rate variables.... In PAGE 35: ...159-160). The Likelihood Ratio Test for VARs of various lag lengths are given below in Appendix Table2 , and indicate that the optimal lag length is three for both specifications. However, an analysis of these tests in conjunction with the residuals suggests that the optimal lag length for models containing real GDP should instead be five.... In PAGE 35: ... However, an analysis of these tests in conjunction with the residuals suggests that the optimal lag length for models containing real GDP should instead be five. Appendix Table2 : Likelihood Ratio Test Results C(t), T(t), RGDP, AGI C(t), T(t), AAA, AGI Lag Length Likelihood Ratio Test Likelihood Ratio Test 1 230.26 167.... ..."

"... In PAGE 12: ... For this taking each significant variable in turn, we added interaction terms between the variable and the phase of conflict dummies. Only one of these produced significant results: secondary school enrolment at the start of the conflict ( Table4 ). We found that while the higher is initial secondary school enrolment the shorter is the conflict, this effect gradually erodes as the conflict persists.... In PAGE 26: ...26 Table4 : Continued Model [1]: Without controls for unobserved heterogeneity Model [2]: With controls for unobserved heterogeneity Log likelihood -75.39 -71.... ..."

"... In PAGE 21: ... All the simulations are generated with the same initial seed, so that the realized sequence of shocks is the same for all economies. The parameters of the various experiments we perform are presented in Table1 . These include the preference parameters (risk aversion and discount factor) and the parameters that characterize the endowment processes (point of supports and transition matrixes).... In PAGE 22: ...We start our discussion by looking at a baseline specification, whose features are reported in the first columns of Table1 and Table 2. The discount factor is equal to .... In PAGE 24: ... Even in the simple example we propose here this does happen. For the parameters in the first column of Table1 , on average, the system finds itself 3.5% of the time in a situation in which private transfers flow form the relatively poorer to the relatively richer individual.... ..."

"... In PAGE 21: ... In the Table we also report the implied properties of the income processes, such as its mean, standard deviation and first order autocorrelation. In the first three rows of Table2 , we report, for each experiment, the average level of utility achieved in the absence of enforceability constraints (which we denote first best), the utility under autarchy, and that achieved with enforceable contacts. Next, we report the standard deviation and autocorrelation of consumption in the three sets of equilibria.... In PAGE 22: ...210 .0001 We start our discussion by looking at a baseline specification, whose features are reported in the first columns of Table 1 and Table2 . The discount factor is equal to .... In PAGE 26: ...14 The effects of such a scheme are immediately apparent in the model we considered above, in that they will enter both the value of autarchy and the continuation value of insurable contracts. As the introduction of the government scheme is equivalent to a reduction in the variance of the aggregate shock, we know already, from Table2 that the effect of such a scheme will be a reduction in the amount of risk sharing that happens in equilibrium. In 14One way to think about such a scheme is that the government is smoothing shocks across villages maintaining a balanced budget at each point in time.... In PAGE 31: ...Table2 , a transfer that moves the mean of the endowment causes a decrease in the level of private transfers. The results we obtain, therefore, are consistent with the implications of the model with imperfect enforceability.... ..."

"... In PAGE 23: ... All the simulations are generated with the same initial seed, so that the realized sequence of shocks is the same for all economies. The parameters of the various experiments we perform are presented in top panel of Table1 . These include the preference parameters (risk aversion and discount factor) and the parameters that characterize the endowment processes (points of support and transition matrices).... In PAGE 23: ... In this part of the Table we also report the implied properties of the income processes, such as its mean, standard deviation and first order autocorrelation. In the bottom panel of Table1 , we report the standard deviation and autocorrelation of consumption in the three types of equilibria (enforceable, first best and autarky). Finally, we report three statistics that are indicative of the amount of risk sharing that is achieved in the equilibrium with self enforceable contracts: the average level of private transfers as a percentage of per-capita income and as a percentage of the transfers in the first best equilibrium and, finally, the fraction of consumption cross sectional variance (after removing aggregate consumption) over the variance of endowments.... In PAGE 25: ...Table1 . The discount factor is equal to .... In PAGE 25: ... In this particular example, the intertemporal allocation of resources is closer to first best than to autarky. This can be seen by looking at the ratio of the cross sectional variance of consumption to the cross sectional variance of endowments (last row of Table1 ): the value of .27 is closer to 0 than to 1.... In PAGE 25: ... Even in the simple example we propose here this does happen. For the parameters in the first column of Table1 , on average, the system finds itself 3.5% of the time in a situation in which private transfers flow form the relatively poorer to the relatively richer individual.... In PAGE 28: ...17 The effects of such a scheme are immedetiately apparent in the model we considered above, in that they will enter both the value of autarky and the continuation value of insurable contracts. As the effect of the government scheme is equivalent to a reduction in the variance of the aggregate shock, we know already, from Table1 that the effect of such a scheme will be a reduction in the amount of risk sharing that occurrs in equilibrium. In particular, we would have a reduction in the average size of private transfers and an increase in the ratio of the variance of consumption (net of aggregate consumption) over the variance of endowments.... In PAGE 32: ... This causes the amount of individual risk sharing to increase. For instance, if the World Bank introduces an aggregate insurance scheme that reduces the variance of aggregate shocks by an amount equivalent to the reduction in aggregate variance discussed in column 5 of Table1 , the ratio of the variance of consumption to endowments (which is inversely related to the amount of risk sharing that is sustainable in equilibrium) which increased from 0.27 in column 1 to 0.... ..."

"... In PAGE 8: ... 4.5 Error Analysis Table5 shows the comparison of the number of er- rors for various models with that for the original model in parsing the GENIA corpus. For each of the methods, the table gives the numbers of common errors with the original Enju model and the ones specific to that method.... ..."

"... In PAGE 27: ...Table9 is a representative sample of various alternative specifications that were tried. As was mentioned earlier, size (as measured by population) may affect migration either as a scaling factor on the payoff shocks, or as a variable affecting the cost of migration.... In PAGE 27: ...6 Sensitivity Analysis Our empirical results are inevitably based on some more or less arbitrary model specification choices. Table9 explores the robustness of the results with respect to some of these choices. The general conclusion is that the parameter estimates are robust.... In PAGE 27: ... If these cost of living differences merely capitalize the value of amenity differences, then unadjusted wages should be used to measure the incentive to migrate. Results for this specification are given in the second column of Table9 : the estimate of quot; is substantially reduced, without much effect on the other coefficients, and the likelihood is lower. Thus in practice the theoretical ambiguity as to whether wages should be adjusted for cost of living differences does not change the qualitative empirical results: either way, income significantly affects migration decisions.... In PAGE 27: ... Thus in practice the theoretical ambiguity as to whether wages should be adjusted for cost of living differences does not change the qualitative empirical results: either way, income significantly affects migration decisions. The last three columns of Table9 show the relevance of the climate variables. Including these variables gives a substantial improvement in the fit, without changing the conclusion that income significantly affects migration.... In PAGE 27: ... If either climate variable is included on its own, the result is that warmer places are preferred. The other specifications in Table9 are concerned with sensitivity of the estimates to the discount factor ($), and the horizon length (T). Reducing $ to .... ..."