# Toy Models – Explaining Growth in Real GDP per capita using the Growth in Real M2 per capita and the President’s Party

I’m always hesitant to include any regressions in blog posts… but I’m gonna do it for this post. Nothing particularly complicated… call it a toy model. But I’m going to touch on two of my favorite topics which also happen to be two of the more controversial topics I’ve touched on over the past year.

Consider… economic theory says that you can move the economy through fiscal or monetary policy. So let’s put together a simple model based on that. We’re going to look at real GDP per capita from BEA’s NIPA Table 7.1. To avoid issues of non-stationarity, we’ll try to explain the percentage change in real GDP per capita.

Monetary policy – I like real M2 per capita better than other measures like, say, nominal interest rates. We can get monthly M2 from the Fed going back to 1959, monthly CPI from the BLS, and quarterly population figures which, like the real GDP per capita, comes from.

I use the March M2 figures for Q1, June M2 figures for Q2, etc.

Fiscal policy – I’m going to use short-hand, and simply create a dummy variable, equal to 1 if the President is a Dem and equal to 0 if the President is a Rep. For those quarters in which there’s been overlap (e.g., Q1 of 2001), I go with the party of the new President since he is in office for the greater part of the quarter.

So, here’s the model:

% change in real GDP per capita = B0 + B1 * % change in real M2 per capita + B2 * Party Dummy

Since the dependent variable is a percentage change, we don’t expect a particularly strong fit – we’re more concerned with the signs and significance of the variables.

Here we go:

Its a dumb little model, but it indicates that the MS is significant. And the President’s party is significant. In fact, having a Dem in the White House is good for an extra 1.1% growth in real GDP per capita.

The fit, frankly, sucks. But then, we didn’t expect much – its a toy model with only two variables, and anyway, explaining a variable that is a percentage change generally is a pain anyhow.

But let’s see what we can do with this simple model. Let’s lag the 1 year % change in real M2 per capita by a quarter. Put another way… let’s see if we can assume that the Fed doesn’t react to changes in the economy as much as it leads them.

Hmmm… the model’s fit rises a bit, the significance of the Fed’s behavior goes up. So yes, it seems the Fed leads the economy more than it follows.

But is a 1 quarter lag the right lag to use? I won’t go through a bunch of variations, but I found I got better results using a 3 quarter lag than anything else:

The fit of the regression is still not that great, but its respectable given that the dependent is a percentage change in something.

And are the variables significant? Well… I’m not one to argue with a P-value of 8.81657298073407E-23. (To those of you aren’t numbers people, the answer to the question “are the variables significant?” amounts to heck yes.) The Fed tends to lead the economy by 3 quarters. And Dem Presidents have produced faster growth. (And no, the Fed isn’t exactly helpful to Democrats.)

So please, once and for all, can we stop arguing about whether these two factors matter? I think a better use of our collective time, frankly, is explaining:

1. Is it a good thing that the Fed is essentially leading the economy considering how partisan it is? (Its pretty clear that Uncle Alan really didn’t like GHW or Clinton, but he bent over backward for GW.)

2. Why does the economy grow faster when Dems are in office.