2slugbaits Uses a VAR to Look at the Effects of M2 and the President’s Party on the Economy

This one is by 2slugbaits…

Here’s a little extension of cactus’ Toy Model. The math can be a bit nerdy, but the pictures are instructive. As I mentioned in a comment to cactus the other day, I ran a simple Vector Autoregression (VAR) model using the first differences of the natural logs of real GDP and seasonally adjusted M2. Using first differences of natural logs eliminates an econometric problem called “nonstationarity,” but we can ignore all that because it gets kind of nerdy. I used a dummy variable for President with “0” for Democrat and “1” for Republican. For purposes of the model it does not matter which party is assigned which value.

The first chart is an “impulse/response” chart showing how real GDP reacts to a one time shock in the M2 money supply. The horizontal axis is in quarters, so 1 = 3 months, 2 = 6 months, etc. The blue bars represent statistical significance confidence bands. If the range of the blue bars is above the zero baseline, then the shock is statistically significant.

In this simply toy VAR model a shock in M2 has a positive effect that persists for approximately 5 quarters (the bottom of the blue band crosses the x-axis at the 6th quarter. The effect dampens with time, as expected.

Here’s a similar chart showing what happens when the President’s party changes from Democrat to Republican.

Here the response is negative; i.e., real GDP goes down in response to a Republican regime. The damage tends to persist for a long time. The top of the blue band (remember, these are negative responses, so the zero baseline is at the top) doesn’t converge to the zero baseline until 16 quarters after the shock occurred. Put another way, the economy begins to recover from the damage after one full Presidential term.

Now this toy model isn’t something that I’d fall on my sword over. There are plenty of technical problems with it, but that does not mean it isn’t suggestive. One reason that economists use VAR models is that they are both parsimonious and fairly robust in showing the general relationship between variables even if the point estimates aren’t always the best.

This one was by 2slugbaits.

Cactus here: Its no surprise that a shock related to the President’s party persists much longer than a monetary shock. A President – whether in office for four years or eight – has time to put in place a lot of policies, many of which are very difficult or even impossible to reverse.

But we keep coming back to the same point. Here at Angry Bear we’ve had post after post after post on this one point. I’m not kidding, this must be at least the 30th post on this topic alone in the last year or so, and we’ve approached it from all sorts of angles with all sorts of tools, and the results are always the same. For some reason, at least going back to Ike, the economy has generally performed better when a Dem sits in the Oval Office. Or perhaps stating it better, the economy has generally performed worse when a Rep sits in the Oval Office. (I think the issue is at least as much Reps generally doing something wrong than it is Dems generally doing something right.) Growth is slower, and the likelihood of a recession is much higher.

If we can’t (all of us, the data included) agree on this simple point, there really is no hope of making things better in the long run.