In: Statistics and Probability
Table 2 below shows regression results from the study of Schularick and Taylor (2012).1 The dataset comprises annual panel data for 12 countries between 1870 and 2008. The study asks a simple question: does a country's recent history of credit growth help predict a financial crisis? The dependent variable is the probability of a financial crisis event pit in country i in year t.
Ordinary Least Squares | |
Country fixed effects | |
Explanatory variables | dependent variable in year t: pit |
credit growth in the previous year (t-1) | -0.0273 |
(0.0815) | |
credit growth in year t-2 | 0.302 |
(0.0872) | |
credit growth in year t-3 | 0.0478 |
(0.0853) | |
credit growth in year t-4 | 0.00213 |
(0.0814) | |
credit growth in year t-5 | 0.0928 |
(0.0752) | |
Observations | 1,272 |
Countries | 14 |
R-squared | 0.023 |
Standard errors appear in parentheses. |
Answer to the following questions (use the |t|>2 rule of thumb in your answers to indicate significance at the 95% level):
a.
The estimated regression equation is,
Probability of a financial crisis in year t = -0.0273 credit growth in year t-1 + 0.302 credit growth in year t-2 + 0.0478 credit growth in year t-3 + 0.00213 credit growth in year t-4 + 0.0928 credit growth in year t-5
b.
Test statistic, t = Coefficient / Std Error
t for credit growth in year t-1 = -0.0273 / 0.0815 = -0.3349693
t for credit growth in year t-2 = 0.302/ 0.0872 = 3.463303
t for credit growth in year t-3 = 0.0478/ 0.0853 = 0.5603751
t for credit growth in year t-4 = 0.00213/ 0.0814 = 0.02616708
t for credit growth in year t-5 = 0.0928/ 0.0752 = 1.234043
At 95% significance level, the coefficients are statistically significant for which |t| > 2
Thus, only the coefficient for credit growth in year t-2 is statistically significant.
c,
The results indicates that only credit growth in year t-2 is statistically significant in predicting the probability of a financial crisis in year t.