Matterwave said:
Perhaps it's best if you also included a regression on the two variables. Instead of graphing them over time, plot the data points for each year on a graph with deficit spending on the x-axis and unemployment on the y-axis. This should give you some better indication if there is correlation (on an instantaneous basis). You can then expand your analysis by plotting say, the unemployment rate at time t and the deficit spending at time t+3 or something to see if there is supposedly lagged correlation.
This is all correlation...it's very hard to come up with causation arguments without first correcting for every omitted variable...omitted variable bias is incredibly hard to get rid of in something as complicated as this haha.
THIS. .. THIS. THIS. THIS!
Yes!
Although such a regression will be basically nonsense. As matterwave said,
omitted variable bias. This means there's an X2 that's causing both X1 and Y1, and this X2 isn't in your model. This makes it look like you have very neat correlation, but you actually don't! (E.G., when comparing student test scores to class size, an example of X2, X3, X4 would be income, geographic area, and whether there are computers in the classroom). An example of that X2 in what we're talking about are real world events, such as recessions, etc. Recessions are causing correlated directional changes in both GDP and unemployment, but it's not in you rmodel, and so it makes unemployment look like it's having a huge causal effect when it's not!
I don't know how you're going to account for this, frankly.
----------------------------------------------------
In addition,
There is some reason to expect this kind of negative correlation after all!
Inflation and employment are negatively correlated. It's not unwise to suspect that the government would risk a deficit during times of non-inflation. This will make it seem like a deficit is causing unemployment. Another omitted variable!
