A response to New Deal Democrat

"New Deal Democrat" at the Boondad blog took the time to examine my latest recession indicator and question some of the methodologies and outcomes. NDD actually contacted me a while ago and expressed some of his opinions over these issues so the fact that he has blogged about it is not only welcome but also allows me a chance to respond.

NDD specifically has an issue with using real 10 year bond rates as a recessionary indicator. The reason is that the relationship between negative real 10 year bond rates and an eventual recession appears to break down in before 1953. Now the reason why I did not use any pre-1953 data in my study is because the 10 year bond rate series (GS10)  available at the St Louis Fed only starts in 1953, while seasonally adjusted inflation (CPIAUCSL) starts in 1947. NDD, however, has pointed out that a discontinued government bond data series called LTGOVTBD has data going back to 1925, and that because both GS10 and LTGOVTBD follow each other closely from 1953 to 2006 (when the series was discontinued) it is therefore a good proxy. Add to the fact that non-seasonally adjusted inflation figures (CPIAUCNS) begin in 1913 and you have the beginnings of a pre 1953 data series that could confirm or deny my assertion that negative real 10 year bond rates will always lead to recession.

And the conclusion that NDD has come up with is that they don't always lead to a recession. And here is the salient graph:

When looking at this graph, understand that whenever the red line is higher than the blue one, then that is an example of negative real 10 year bond rates (or, more specifically, negative real long term bond rates). As you can see, there are seven instances since 1925.

Now of course what I have done is to average out the results over three months. This is what we get in the first instance:

1925 and 1926 see some low rates but the only negative result occurs in January 1926. So what happens after? A recession in October 1926. Hmmm.

The second and third instances occur during the New Deal era:

Actually what is notable from this graph is the sheer height that real long term bond rates hit during the depression - peaking at 14.27% in March 1932. If anything this is pretty solid evidence that real 10 year bond rates can operate as a "window" for the economy to grow: too low and the economy crashes, too high and the economy crashes.

As you can see, real bond rates return to more reasonable levels around 1934. The key date here is March 1933 when the Emergency Banking Act was passed. Another important date was June 1933 when Glass-Steagall was passed. Both of these acts resulted in an increase in the money supply and a subsequent move out of deflation. By December 1933, prices began inflating again. This is important to realise in the data in the graph above since the presence of inflation again reduced real long term bond rates from the stratosphere of 14% to around 3% by December 1933.

Of course the graph then shows that in 1934, inflation had reached a point where real long term bond rates had turned negative. They turn negative again for a short time in 1935, and then negative again 1937... which, um, then turns into another recession.

I suppose I could argue that the instances of negative real long term bond rates in 1934, 1935 and 1937 were the cause of the 1937 recession. I think that is not an unreasonable assumption to make, especially the 1937 instance. We need to remember though that the massive hangover of debt and deflation that typified the great depression's stratospheric real long term bond rates needed a little bit more than a few months of negative long term rates to balance against. Nevertheless, if my "window" theory is true, then it could possibly be said that the US economy might've avoided the 1937 recession had real ten year bond rates always remained positive from 1934 onwards. In short, too much inflation was created.

After all, you could put out a house fire with a fire hose, but if you immerse the burning house in the sea, you still end up damaging it badly.

Okay, so let's move on to the war and post-war years:

So as you can see, the war brought about a huge drop in real 10 year bond rates. A recession occurs in February 1945, and then another huge drop in real ten year bond rates occurs during 1946-1949. In fact that period has the deepest recorded period of negative real long term bond rates (which peaks at -16.96% in April 1947). Then another recession hits in November 1948. Then another period of negative rates hits between 1950-1952, followed by a recession in 1953. Beyond this point my data series kicks in.

The first thing to look at here is the effect of the war, specifically the effect of a Total war economy upon the United States. This graph shows what I'm talking about:

Between Japan's attack on Pearl Harbor and the demobilisation of forces in 1946, the US government effectively quadrupled in size. This was due to massive spending to create a war machine capable of defeating Japan and Germany. The growth in the size of government was not "added onto" the economy, but effectively "crowded out" the private sector. The US government increased taxation and went on a massive borrowing spree to fund the war.

Now this is very important for us to understand. In my analysis of real long term bond rates, we're not just looking at causal relationships but also the effect of such causal relationships to the wider economy. Long term bond rates are an indication of how the market is acting at a certain time. If the government is 10% of the economy then the private sector is 90% of the economy, which means that any negative long term bond rates are being experienced by that 90%. But in the war years we see an increase in the size of government and a decrease in the size of the private sector. This means that the negative real long term bond rates during this period are only experienced by around 55% to 60% of the economy. Thus the effect would be less - hence the lack of recession until 1945.

As for the period of negative real long term bond rates between 1946 and 1949, these end up presaging the 1948-49 recession.

In short while the pre-1953 data is essential to examine, I do not think that it rules out the correlation and causation effect at all. Two huge events - The Great Depression and World War 2 - are enough to affect the quality of the data under examination.

There is another issue: Is LTGOVTBD a good proxy for GS10? Just because they correlate when they are measured concurrently doesn't mean that the same correlation existed pre-GS10. Long term government bond rates hardly move much at all between 1925 to 1956, despite the huge swings in inflation and deflation in this period. They do not exhibit the same ups and downs which typify GS10 rates. this makes me think that LTGOVTBD rates, at least in the early days, were pegged by government order. GS10, especially from the 60s onward, seem to move according to the actions of the market.

And finally an answer to a commenter at Boondad who points out:

You cannot calculate CURRENT "real long term interest rates." It's a fallacy.

To calculate a CURRENT "real long term interest rate," you would need to know what "inflation" (CPI, whatever) WILL BE over the next TEN YEARS while you collect your 10-Year Treasury Coupons. You don't know that.

I think this commentator has got it wrong because the interest rate, while certainly applicable over a ten year period, is determined by what the market wants at the time. So while it has a long term function it also has a short term indicative effect. Bond rates go up when investors see bonds less as less attractive to other investments (such as shares) and go down when the market is running away from something. This can be seen during late 2008 when bond rates plunged to 2.42% during the credit crunch after being 4.1% 5 months before. The idea is that the interest rates of government bonds is a  way of measuring market sentiment and activity. 10 year bonds are thus a good measurement of what is going on at the time that they are invested in.

And my argument is that when investors put money into ten year bonds they are essentially investing in something "safe". But when inflation exceeds that "safe" investment, things go wrong and a recession follows. This will be because a calculation of your return on interest will be less than the increase in consumer prices. In other words, the amount of money you gain from such an investment will be less than your increase in spending. Thus a recession. And if my "window" theory is correct, a recession will also occur when the amount of money you gain from an investment will be more than offset by a decrease in income from other sources.

No comments: