Online appendix to Mark My Words: Information and the Fear of Declaring an Exchange Rate Regime

Size: px

Start display at page:

Download "Online appendix to Mark My Words: Information and the Fear of Declaring an Exchange Rate Regime"

Margaret Dickerson
6 years ago
Views:

Online appendix to Mark My Words: Information and the Fear of Declaring an Exchange Rate Regime Pierre-Guillaume Méon and Geoffrey Minne In this online

for electoral competitiveness, law and order, and central bank independence; we drop countries with high capital controls, and we run a bootstrap

JEL Classifications: F33, F41, F53, F59, L82.

1 Online appendix to Mark My Words: Information and the Fear of Declaring an Exchange Rate Regime Pierre-Guillaume Méon and Geoffrey Minne In this online appendix, we provide extra evidence complementing our paper Mark My Words: Information and the Fear of Declaring an Exchange Rate Regime : we control for electoral competitiveness, law and order, and central bank independence; we drop countries with high capital controls, and we run a bootstrap estimation. Keywords: Official exchange rate regime, de facto exchange rate regime, press freedom, information, fear of floating. JEL Classifications: F33, F41, F53, F59, L82. CEB Working Paper N 13/040 November 2013 Université Libre de Bruxelles - Solvay Brussels School of Economics and Management Centre Emile Bernheim ULB CP114/03 50, avenue F.D. Roosevelt 1050 Brussels BELGIUM ceb@admin.ulb.ac.be Tel. : +32 (0)2/ Fax: +32 (0)2/

2 Online appendix to Mark My Words: Information and the Fear of Declaring an Exchange Rate Regime November 5, 2013 Pierre-Guillaume Méon Université libre de Bruxelles (ULB) Centre Emile Bernheim CP-114/03 Avenue F.D. Roosevelt, Brussels, Belgium Phone: pgmeon@ulb.ac.be Geoffrey Minne Université libre de Bruxelles (ULB) Centre Emile Bernheim and ECARES CP-139 Avenue F.D. Roosevelt, Brussels, Belgium Phone: gminne@ulb.ac.be Abstract: In this online appendix, we provide extra evidence complementing our paper Mark My Words: Information and the Fear of Declaring an Exchange Rate Regime : we control for electoral competitiveness, law and order, and central bank independence; we drop countries with high capital controls, and we run a bootstrap estimation. JEL Codes: F33, F41, F53, F59, L82. Keywords: Official exchange rate regime, de facto exchange rate regime, press freedom, information, fear of floating. 1

3 1. Controlling for the broader institutional context In Table 1 we add the legislative index of electoral competitiveness from the 2009 Database on Political Institutions (Beck et al, 2001) to the set of control variables. That variable is based on the way in which the legislature is appointed and on the number and share of political parties in the legislature. It ranges from 1 to 7, and increases with competitiveness. It can be used to control for the impact of politics in addition to democracy. Table 1: Baseline logit model controlling for electoral competitiveness. (1.1) (1.2) (1.3) (1.4) (1.5) (1.6) Freedom ** ** House Index (0.010) (0.020) (0.011) Internet users *** ** ** (%) (0.011) (0.032) (0.014) GDP per capita * ** ** (log) (0.158) (0.286) (0.206) (0.161) (0.379) (0.204) Population (0.001) (0.022) (0.001) (0.001) (0.012) (0.001) Trade openness (0.003) (0.006) (0.004) (0.003) (0.005) (0.004) Democracy ** ** dummy (0.363) (0.761) (0.459) (0.319) (0.719) (0.381) Financial Openness (0.106) (0.337) (0.120) (0.100) (0.270) (0.117) Legislative ** ** ** ** competitiv. (0.105) (0.322) (0.118) (0.089) (0.244) (0.105) Constant * (1.411) (2.998) (1.870) (1.438) (3.212) (1.848) Countries Years sample Observations 1, ,172 1, ,282 Correctly classified 65.58% 89.05% 62.71% 64.52% 85.34% 62.17% Robust standard errors are clustered by and are shown below coefficients. All regressions include year fixed effect. The sample in columns 1.1 and 1.4 is the whole sample. The sample consists of de facto s only, in columns 1.2 and 1.5. The sample consists of de facto s only, in columns 1.3 and 1.6 *** significant at 1%, ** significant at 5%, * significant at 10%. 2

4 To control for the broader de facto institutional environment, we add the law and order index from the International Country Risk Guide published by the PRS group to the set of explanatory variables. The law and order index assesses the strength and impartiality of the legal system. The results of those regressions are displayed in Table 2. Table 2: Baseline logit model controlling for law and order. (2.1) (2.2) (2.3) (2.4) (2.5) (2.6) Freedom House ** ** Index (0.010) (0.024) (0.012) Internet users *** *** * (%) (0.011) (0.034) (0.014) GDP per capita ** * ** (log) (0.183) (0.354) (0.252) (0.189) (0.371) (0.265) Population (0.001) (0.017) (0.001) (0.001) (0.009) (0.001) Trade openness (0.003) (0.007) (0.005) (0.003) (0.006) (0.005) Democracy ** *** dummy (0.382) (1.152) (0.491) (0.300) (0.633) (0.374) Financial Openness (0.112) (0.339) (0.129) (0.106) (0.265) (0.125) Law and order ** *** (0.164) (0.318) (0.194) (0.139) (0.226) (0.171) Constant ** *** ** *** (1.321) (2.502) (1.923) (1.416) (2.578) (2.081) Countries Years sample Observations 1, ,045 1, ,163 Correctly classified 65.96% 91.57% 63.35% 64.19% 89.59% 65.18% Robust standard errors are clustered by and are shown below coefficients. All regressions include year fixed effect. The sample in columns 2.1 and 2.4 is the whole sample. The sample consists of de facto s only, in columns 2.2 and 2.5. The sample consists of de facto s only, in columns 2.3 and 2.6 *** significant at 1%, ** significant at 5%, * significant at 10%. 3

5 In Table 3, we control for central bank independence, measured by the updated version of Cukierman s index used by Acemoglu et al. (2008). This index, which captures how much independence a central bank officially has, is based on legal features as determined in the bank s charter. Table 3: Baseline logit model controlling for central bank independence. (3.1) (3.2) (3.3) (3.4) (3.5) (3.6) Freedom ** House Index (0.020) (0.104) (0.024) Internet users ** ** (%) (0.019) (0.083) (0.030) GDP per *** *** * * capita (log) (0.339) (3.494) (0.377) (0.425) (3.455) (0.701) Population * (0.001) (0.012) (0.002) (0.001) (0.009) (0.002) Trade openness (0.005) (0.069) (0.008) (0.005) (0.113) (0.008) Democracy dummy (0.703) (0.822) (0.473) (0.550) Financial ** Openness (0.230) (1.335) (0.307) (0.175) (0.528) (0.266) CBI (1.254) (4.563) (2.250) (1.090) (7.135) (1.735) Constant *** ** * (3.919) (29.713) (5.594) (4.495) (45.206) (7.551) Countries Years sample Observations Correctly classified 78.89% 94.31% 77.89% 75.60% 95.15% 77.75% Robust standard errors are clustered by and are shown below coefficients. All regressions include year fixed effect. The sample in columns 3.1 and 3.4 is the whole sample. The sample consists of de facto s only, in columns 3.2 and 3.5. The sample consists of de facto s only, in columns 3.3 and 3.6. *** significant at 1%, ** significant at 5%, * significant at 10%. 4

6 2. Dropping countries with high capital controls Capital controls may allow some countries to declare a fixed exchange rate regime, keep the official rate pegged, and believe they have accurately declared their official regime, while their black market rate actually floats. In Table 4, we drop countries that exert stringent capital controls. 1 Table 4: Baseline logit model where countries with high capital control are excluded. (4.1) (4.2) (4.3) (4.4) (4.5) (4.6) Freedom * ** House Index (0.009) (0.018) (0.011) Internet users *** ** ** (%) (0.011) (0.041) (0.013) GDP per ** * capita (log) (0.152) (0.316) (0.196) (0.162) (0.524) (0.197) Population (0.001) (0.036) (0.001) (0.001) (0.015) (0.001) Trade openness (0.003) (0.005) (0.004) (0.003) (0.006) (0.004) Democracy * * ** dummy (0.354) (0.792) (0.433) (0.282) (0.574) (0.333) Financial * Openness (0.110) (0.376) (0.122) (0.105) (0.307) (0.119) Constant * * ** (1.207) (2.640) (1.577) (1.310) (4.339) (1.623) Countries Years sample Observations 1, ,297 1, ,404 Correctly classified 66.92% 90.61% 64.61% 68.10% 88.89% 63.39% Robust standard errors are clustered by and are shown below coefficients. All regressions include year fixed effect. The sample in columns 4.1 and 4.4 is the whole sample. The sample consists of de facto s only, in columns 4.2 and 4.5. The sample consists of de facto s only, in columns 4.3 and 4.6. *** significant at 1%, ** significant at 5%, * significant at 10%. 1 We dropped observations with a Chinn and Ito (2006) index belonging to the three categories signaling the most stringent restrictions to capital movements. Those observations represent approximately 12% of Chinn and Ito s (2006) dataset, but only 3% to 4% of our baseline samples. The problem related to capital controls is therefore less severe than one may think. We also performed the same robustness check by dropping successively the extreme value of the index, then the most and second extreme values of the index, and we obtained similar results. Those two extreme categories account for no more than 11% percent of Chinn and Ito s (2006) sample, while dropping the fourth category would imply dropping near 45% of their sample. Dropping the three extreme categories is therefore a reasonable compromise, making it possible to drop significant restrictions to capital movements without reducing the sample size to a point that would prevent meaningful regressions. 5

7 3. Bootstrapping As de facto exchange rate regimes, and consequently our dependent variables, are estimated, standard errors may be underestimated, thereby biasing statistical inference and leading to spuriously significant coefficients. We therefore reran all our estimations again using bootstrapped standard errors. The results are reported in Table 5. Table 5: Baseline logit model with bootstrapped standard errors. (5.1) (5.2) (5.3) (5.4) (5.5) (5.6) Freedom ** ** House Index (0.008) (0.029) (0.012) Internet users *** ** ** (%) (0.010) (0.036) (0.014) GDP per * * capita (log) (0.165) (0.689) (0.208) (0.170) (0.458) (0.218) Population (0.002) (0.044) (0.012) (0.002) (0.029) (0.013) Trade openness (0.003) (0.016) (0.006) (0.004) (0.008) (0.005) Democracy * ** dummy (0.351) (1.327) (0.466) (0.253) (0.871) (0.353) Financial Openness (0.134) (0.467) (0.134) (0.089) (0.312) (0.123) Constant 2.293* ** (1.238) (4.323) (1.517) (1.340) (3.963) (1.904) Countries Years sample Observations 1, ,332 1, ,441 Correctly classified 65.64% 89.37% 64.41% 66.04% 85.53% 63.15% Robust standard errors are estimated using 50 bootstrap replications. They are clustered by and are shown below coefficients. All regressions include year fixed effect. The sample in columns 5.1 and 5.4 is the whole sample. The sample consists of de facto s only, in columns 5.2 and 5.5. The sample consists of de facto s only, in columns 5.3 and 5.6. *** significant at 1%, ** significant at 5%, * significant at 10%. 6

8 4. Hyperinflation In previous regressions, we neutralized the effect of episodes of hyperinflation by dropping the corresponding observations. As a robustness test, we rerun our baseline specifications including episodes of hyperinflation, considered as flexible regimes. The results are displayed in Table 6. Just as in benchmark findings, increased press freedom and access to information decrease the propensity to announce a regime that is different from the de facto regime. Table 6: Baseline logit model including episodes of hyperinflation (6.1) (6.2) (6.3) (6.4) (6.5) (6.6) Freedom House Index * ** (0.009) (0.018) (0.011) Internet users (%) *** ** ** (0.010) (0.024) (0.013) GDP per capita (log) ** * * * (0.145) (0.276) (0.197) (0.150) (0.376) (0.198) Population (0.001) (0.014) (0.001) (0.001) (0.009) (0.001) Trade openness (0.003) (0.006) (0.004) (0.003) (0.005) (0.004) Democracy dummy ** * *** (0.346) (0.723) (0.434) (0.263) (0.528) (0.331) Financial Openness (0.103) (0.272) (0.116) (0.093) (0.247) (0.113) Constant * (1.464) (2.710) (1.937) (1.413) (3.482) (1.840) Countries Years sample Observations 1, ,332 2, ,441 Correctly classified 66.72% 87.84% 64.41% 67.58% 86.15% 63.15% Note: Robust standard errors are clustered by and are shown below coefficients. All regressions include year fixed effect. The sample in columns 6.1 and 6.4 is the whole sample. The sample consists of de facto s only, in columns 6.2 and 6.5. The sample consists of de facto s only, in columns 6.3 and 6.6 *** significant at 1%, ** significant at 5%, * significant at 10% 7

9 5. Lagged explanatory variables We lag the variables of interest by one year. 2 The results are displayed in the Table 7. They still show a negative impact of greater press freedom and access to information on a s propensity to dissimulate its exchange rate regime. Table 7: Baseline logit model with lagged press freedom and access to information (7.1) (7.2) (7.3) (7.4) (7.5) (7.6) Freedom House Index * ** (0.009) (0.019) (0.010) Internet users (%) *** ** ** (0.012) (0.046) (0.014) GDP per capita (log) ** ** ** * (0.151) (0.301) (0.199) (0.158) (0.399) (0.198) Population (0.001) (0.022) (0.001) (0.001) (0.012) (0.001) Trade openness (0.003) (0.005) (0.004) (0.003) (0.006) (0.004) Democracy dummy ** ** *** (0.347) (0.773) (0.431) (0.277) (0.610) (0.333) Financial Openness (0.105) (0.313) (0.118) (0.098) (0.275) (0.114) Constant 3.126*** * 4.308*** (1.184) (2.379) (1.576) (1.698) (3.910) (1.840) Countries Years sample Observations 1, ,268 1, ,378 Correctly classified 65.38% 89.75% 65.14% 66.24% 86.36% 64.37% Note: Robust standard errors are clustered by and are shown below coefficients. All regressions include year fixed effect. The sample in columns 7.1 and 7.4 is the whole sample. The sample consists of de facto s only, in columns 7.2 and 7.5. The sample consists of de facto s only, in columns 7.3 and 7.6 *** significant at 1%, ** significant at 5%, * significant at 10% 2 We also considered a lag of two years. The results were unchanged. 8

10 Probability of Probability of 6. Predicted probability by decile To have a better representation of the results across our sample, we plot the predicted probability of the exchange rate regime for each decile of the key explanatory variable (baseline model). The probability is strictly decreasing for both press freedom and access to information. The negative effect related to access to information seems to be reinforced from the 7 th decile on. Figure 1: Predicted probability of one s exchange rate regime according to press freedom deciles. 0,7 0,6 0,5 0,4 0,3 0,2 0, deciles Figure 2: Predicted probability of one s exchange rate regime according to Internet access deciles. 0,6 0,5 0,4 0,3 0,2 0, Internet access deciles 9

11 7. References Acemoglu D., Johnson, S., Querubin, P., Robinson, J.A., When does policy reform work? The case of central bank independence. Brookings Papers on Economic Activity, 39 (1), Beck, T., Clarke, G. Groff, A. Keefer, P. Walsh, P., New tools in comparative political economy: The Database of Political Institutions. World Bank Economic Review 15 (1), Chin, M.D., Ito, H., What matters for financial development? Capital controls, institutions, and interactions. Journal of Development Economics 81 (1),

What Type of Microfinance Institutions Supply Savings Products?

What Type of Microfinance Institutions Supply Savings Products? Anastasia Cozarenco, Marek Hudon and Ariane Szafarz Recent evidence shows that the poor desperately need access to savings products. But