American trade policy towards Sub Saharan Africa –- a meta analysis of AGOA



Table 1: MST—Test of Authentic effect

(1)
log of T-stat

(2)

T-stat

(3)
log of T-stat

(4)
log of T-stat

(5)

Corrected T-stat

log of degrees of freedom

degrees of freedom (square root)

N (in logs)

precision

0.175***

(0.026)

0.013

(0.008)

0.176***
(0.026)

0.001***
(0.000)

-0.010

(0.013)

Constant

-1.059***

(0.224)

1.400*

(0.549)

-1.078***

(0.227)

0.066

(0.092)

Observations

172

173

172

172

173

R2

0.218

0.061

0.220

0.227

0.008

F-test

45.10

2.362

45.57

30.28

0.515

Robust Standard errors in parentheses, Absolute values of T-stat used
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001

Pooled Meta-Regression Analysis

In Table (2), our initial FAT/PET results are presented using a random effects model. Precision is insignif-
icant in all three columns of the table, indicating the absence of a genuine effect. The constant (
β0) in
Equation (8) is significant at 5% in column (2) indicating the presence of publication bias. In the remaining
two columns the
FAT is not passed. This is not however, indicative of the absence of publication bias, given
that the funnel plots are consistent with the result from column (2). We believe it is the nesting of several
coefficients per study which is driving this. In terms of the remaining variables in columns (2) and (3),
it is observed that the number of explanatory variables, annual data and logged dependent variable con-
tribute to higher t-statistic values, all things equal. On the contrary, holding all other variables constant, the
number of countries, cross-sectional data, robust standard errors, product fixed effects, GMM and Heck-
man/Tobit type estimators relative to other estimators reduce the reported t-statistics. These are all in line
with expectations—for example robust standard errors tend to reduce the bias in reported t-statistics, while
increasing the number of countries (increases the sample size) and dilutes the effect of the reported
agoa
coefficient. Nevertheless, product fixed effects, GMM and Heckman/Tobit estimators tend to reduce the
bias in OLS coefficients. Further analysis to investigate the publication bias and the evidence of a genuine
effect are presented in the remaining tables to enable us reach a more definitive conclusion on the presence
of publication bias and genuine effects.

11



More intriguing information

1. Opciones de política económica en el Perú 2011-2015
2. On the origin of the cumulative semantic inhibition effect
3. Towards a framework for critical citizenship education
4. The name is absent
5. Lending to Agribusinesses in Zambia
6. Explaining Growth in Dutch Agriculture: Prices, Public R&D, and Technological Change
7. Correlation Analysis of Financial Contagion: What One Should Know Before Running a Test
8. Evaluating the Success of the School Commodity Food Program
9. The name is absent
10. AN EMPIRICAL INVESTIGATION OF THE PRODUCTION EFFECTS OF ADOPTING GM SEED TECHNOLOGY: THE CASE OF FARMERS IN ARGENTINA