by Cyril Morong

My guess is that it has been studied quite a bit, with probably alot of papers on it. Alot of people talk about trade deficits and how they affect jobs, wages and the unemployment rate.

But I tried something myself anyway. The data was the U.S. from 1965-2000.

I ran a regression in which the yearly percentage point change in the unemployment rate was the dependent variable (UE). The independent variables were

BAL = the yearly percentage point change in the trade balance (as a percentage of GDP). For example, in 2000 it was -3.8% and in 1999 it was -2.8%. So for 2000, the change was -1.0.

LF = the yearly percentage point change in the labor force participation rate.

PR = the yearly percentage change in productivity.

GDP = the yearly percentage change in the real per capita GDP (with the labor force used instead of the population).

WAGE = the yearly percentage change in real hourly wages

The OLS regression equation was

UE = .49 + .368*BAL - .87*LF +
.087*PR - .358*GDP + .05*WAGE

The r-squared was .854 and the standard error was .391

The t-values were

BAL 2.65

LF –3.00

PR 1.48

GDP –9.82

WAGE 1.07

So holding the other factors constant, as the yearly percentage point change in the trade balance gets more positive (negative), the yearly percentage point change in the unemployment rate gets bigger (smaller). This says that as we get bigger trades deficits, the lower the unemployment rate (at least compared to the previous year).

It looks like as the labor force participation rate rises, the unemployment rate goes down, ceteris paribus. But the bigger the productivity improvement, the higher the unemployment rate. I thought that if productivity goes up, firms could increase output without adding workers. But that might be only one explanation. And of course, the bigger the increase in GDP, the lower the unemployment rate.

So if GDP goes up 4.0%, then the unemployment rate would fall 1.43 percentage points. If the trade balance gets 1.0 percentage points more negative, then the unemployment rate would fall .368 percentage points. If productivity rises 3.0%, then the unemployment rate would rise .261 percentage points. If the labor force participation rate rise 1.0 percentage points, unemployment would fall .87 percentage points.

There may be time series issues like serial correlation which I have not tested for. There may also be other factors like minimum wage laws, regulations, etc that might affect this.

Comments welcome.

The correlation between WAGE and PR was high, .72. So I removed both from the regression and added the difference (PR – WAGE). So it is the productivity increase above the wage increase. Sort of the increase in the net benefit of hiring workers. That regression equation was

UE = .85 – .02*(PR – WAGE) + .216*BAL – 1.28*LF - .34*GDP

The r-squared was .813 and the standard error was .436. The t-values were

(PR – WAGE) -.398

BAL 1.50

LF –4.47

GDP –8.48

But the unemployment rate still goes down when the trade deficit gets more
negative since the coefficient on BAL is still positive (which says if the
trade balance gets more positive, the unemployment rate goes up).

I also tried lagging the BAL variable one year. The coefficient became
.087. But that still shows that as the trade balance gets more positive, the UE
rate goes up. And as the trade balance gets more negative, the UE rate goes
down. Regressions with only GDP growth, BAL and productivity growth have pretty
much the same results.

As for wages, they too seem to go up when the trade deficit gets
bigger. This covered the 1965-2004 period.

In the regression below

WAGE = the annual percentage
increase in the hourly wage – the annual percentage increase in the CPI

So it is the increase in real wages.

WAGE = -1.96 + .616*PR +.165*GDP -.70*LF - .022*BAL

The r-squared was .40 and the standard error was about 1.7 (which seems
high). So, as the trade deficit gets bigger (or more negative), wages go up. If
I lagged the trade deficit (BAL) one year, it got even more negative. The
t-value on productivity was 3.24. So it was significant. None of the other
variables had t-values close to 1.96.

If I took out the labor force participation variable (LF).

WAGE = -2.25 + .679*PR +.151*GDP + .042*BAL

So, if trade deficit gets smaller, wages go up since the coefficient on
BAL is positive. If we went from having imports equal to exports to a trade
surplus equal to 1% of GDP, wages would go up .04%. This is extremely slight
and the t-value was .07, so it was not significant. If I lagged the trade
deficit (BAL) one year, it turned negative. Again meaning that a bigger trade
deficit meant higher wages. Notice how productivity is the dominant force.

So it looks like reducing the trade deficit does not increase wages.