Is the US current account deficit stationary over time? To answer this question, we will use Unit Root tests. Go to FRED, and download the series “Balance on current account, NIPA’s” as well as the series “Gross Domestic Product, Quarterly, Seasonally Adjusted Annual Rate” symbol GDP. Import the data from Excel into your econometric software. Keeping the data at quarterly frequency, calculate the ratio of the current account balance to GDP.
a. Run the Augmented Dickey-Fuller unit root test on the current account balance/GDP ratio for all of the observations available. Make sure to test the level, not the first difference. Include an intercept but do not include a trend. What do the test results mean about the stationarity of the current account/GDP ratio?
b. Reset the sample to 1947Q1 until 1980Q1. Rerun the Augmented Dickey-Fuller test for this period. Was the current account/GDP stationary then? Explain your results from examination of the data and other sources.
c. Reset the sample to 1981Q1 until 1991Q1. Rerun the Augmented Dickey-Fuller test for this period. Was the current account/GDP stationary then? Explain your results from examination of the data and other sources.
d. Finally, reset the sample to 1991Q2 to 2022Q2 and rerun the test. What about stationarity in this period? Again explain your results from examination of the data and other sources.
e. A possible explanation of all of these results is that the US current account deficit has undergone two “level shifts”. Before 1981, the US had a very small positive current account balance on average, but with ups and downs. Between 1981 and 1991, the current account shifted to deficit, but fluctuated around a new level. After 1991, yet another level was reached.
To try out this hypothesis, create 3 dummy variables:
Pre1981=1 from 1947 Q1 to 1980 Q4, 0 from 1981 Q1 on.
Dum1981q11991q1=1 from 1981 Q1 until 1991 Q1, 0 all other times.
Dum1991q22022q2=1 from 1991 Q2 until 2022 Q2, 0 all other times.
Run the following regression:
CAGDP = a*Pre1981 + b* Dum1981q11991q1 +c* Dum1991q22022q2 +u where u is the stochastic error term.
Do your regression results support the level shift idea?
Provide your answers to the questions in a short write-up. Provide some brief background about the impact of the Bretton Woods Agreement and its abandonment in 1971-3, the Plaza Accords of 1985, and any other major policies, events or trends affecting the evolution of the current account over the years covered in these estimations. Briefly discuss how your results fit with that background information.
