Model
We conduct a country-by-country time series analysis of the government reaction function using impulse response of the country risk (Measured by the EMBIG) to an increase in de Federal Funds Rate. For this purpose, we part from a simple reaction function that we estimate for each country:
(1) st = α + Bit-1 + et where et ~ i.i.d.
Where s and i are, respectively, the country risk and the U.S. Federal Funds Rate. The latter is exogenously defined by the Federal Reserve.
Data and Methodology
The country risk variable is proxies with the EMBIG index, as explained in the main article. For this variable, the source of the data is the World Bank Economic Monitor. For the Federal Funds Rate, the source is FRED. We first test the stationarity of each variable. Tests reject non stationarity of all the first differenced series at the 95% confidence level.
The VAR lag order selection criteria indicates that 1 lag is optimal. Since for small samples, the AIC is more appropriate, we report its results for each country for different lags (Table 1). Both the Trace and maximum Eigenvalue tests in Johansen Full information Maximum Likelihood test suggests one cointegrating equation for different model assumptions concerning deterministic trends. We select the one with intercept and no trend in VAR.
Next, we proceed to build a Vector Error Correction Model (VECM), country by country, with one lag determined optimally as described above:
(3) Δyt = -Πyt-1 + ∑i=1 ΦjΔyt-i + ui,
Where yt =
Results
In Table 2 we report the long-run cointegration equations for each country. For all of them, an increase in the Federal Funds rate generates an increase in the EMBIG. How do these results hold in the short run? To answer this we turn into impulse response functions, which are displayed in the second graph of the main article. The short-run results are consistent with the long-run ones: an increase in the Federal Funds Rate generate, on average, during the following next quarters, an increase in the country risk.
Table 1. AIC for each country 1 to 3 lags.
| Country | AIC 1 lags | AIC 2 lags | AIC 3 lags |
| Argentina | 19.83 | 20.15 | 20.10 |
| Brazil | 17.35 | 17.50 | 17.48 |
| Colombia | 13.73 | 14.74 | 15.82 |
| Mexico | 14.31 | 14.62 | 15.50 |
Table 2. Vector Error Correction Models.
| Cointegrating Eq: | CointEq1 |
| FEDRATE(-1) | 1.000000 |
| EMBIG_ARG(-1) | -0.002028 |
| (0.00082) | |
| [-2.46806] | |
| C | 0.381888 |
| Cointegrating Eq: | CointEq1 |
| EMBIG_BRA(-1) | 1.000000 |
| FEDRATE(-1) | -143.4935 |
| (39.1876) | |
| [-3.66171] | |
| C | -144.2223 |
| Cointegrating Eq: | CointEq1 |
| EMBIG_COL(-1) | 1.000000 |
| FEDRATE(-1) | -102.1662 |
| (20.8519) | |
| [-4.89961] | |
| C | -88.09853 |
| Cointegrating Eq: | CointEq1 |
| FEDRATE(-1) | 1.000000 |
| EMBIG_MEX(-1) | -0.037714 |
| (0.00541) | |
| [-6.97417] | |
| C | 6.930402 |
Note: The table displays the estimation results of the cointegrations equation for, respectively, Argentina, Brazil, Colombia and Mexico.
