State Space Models

All state space models are written and estimated in the R programming language. The models are available here with instructions and R procedures for manipulating the models here here.

Sunday, April 19, 2015

Is the US Printing Too Much Money?


The Federal Reserve, the central bank of the US, has the power to print money. The US has just been through the Financial Crisis of 2007-2008. As a result of the Financial Crisis, the US Federal government has gone into debt both to maintain operations in the face of decreased tax revenue and to stimulate the economy. The Federal Reserve could simply print money to erase the Federal Debt but the fear is that printing money will lead to inflation

In this post, I look at this issue using statistical models based on Complex Systems Theory and World-Systems Theory. The models show that the US has not printed too much money (but could at some point in the future and has at times in the past) and that the money supply has historically had little to do with inflation as measured but the Consumer Price Index (CPI). Other forces in the world-system are at work here, not just the policies of the US Federal Reserve.

Printing money has been a contentious issue throughout US History and the current episode is no different (if you want to read in more detail type Is the US printing too much Money into the Google search engine). Monetary theory is also a contentious area in macroeconomics. If I tried to summarize  the area, you would instantly stop reading this post. 

Let me just mention one theory that is easy to understand and applies to the question at hand (most monetary theory doesn't). The theory is Milton Friedman's k-percent rule. Simply put, the central government should increase the money supply at some fixed percent, the k-percent. Contrast Friedman's theory to Keynesian counter-cyclical policy: the money supply should be increased during recessions to stimulate the economy and decreased after the recession to prevent inflation. The problem with each of these theories is "how much." How much should k-percent be or how much should the money supply be increased during a recession and decreased afterwards?

The "how much" question could be rephrased in a way that would be understandable to Stock Market Analysts who used technical analysis. The figure above is the US M1 Money supply (the definition of the money supply that is under government control) taken from the Financial Forecast Center (FFC). It includes actual data starting in April 2012 and a forecast that starts in 2015. The forecast is made using artificial intelligence techniques, not economic theory. A simple form of technical analysis would just connect the high and the low points for M1 over a period of time (the dashed green and blue lines). The argument is that if M1 goes outside this range, it is changing too much. Using this form of analysis, what tends to scare analysts (the red arrow in the graph) is when M1 increases rapidly as it did after Dec-2014. A problem with the graph above is its limited historical scope. We'd really like to look further back to set reasonable ranges and decide how M1 has fluctuated historically. In any event, the FFC is forecasting a peak in M1 for 2015.


The figure above shows M1NS (M1 not seasonally adjusted) from the Federal Reserve. We can see that the money supply expanded during the Dot-com Bubble but remained fairly flat until 2009. Why did M1 increase during the Dot-com Bubble and what would have happened had continued increasing (line A) rather than flattening out until 2010? Were the sharp increases in the money supply (lines B and C) after the Financial Crisis justified or something to be feared? And, what are the dashed green, red, and blue lines in the figure?

The dashed green and blue lines are the 98% bootstrap prediction intervals for the dashed red line, which is the attractor path for M1. The attractor path is the simulated time path of M1 derived from a state space model of the US economy. It shows what M1 would have been (a fictional line) without random shocks (the black line is the fact line). The attractor path is the line to which M1 will return without random shocks. The conclusion is that from before 1980 until 2000, M1 was too high. After 2000, until 2012, M1 was too low. As of 2012, M1 was right on the attractor path; if it stays there increasing at k-percent per year, M1 will be just right and it cannot be said that the US is printing too much money.


Now let's look at the US Inflation Rate as measured but the Consumer Price Index (CPI). The graph above is another forecast from the Financial Forecast Center (FFC), this time looking at the rate of change in the CPI. There have been a lot of increases and decreases in the CPI since Apr-12. Each increase (solid red arrow) could have been used by commentators to trigger fears of inflation. Technical analysis shows that the swings are increasing but have never peaked much over 2% while the FFC forecast is for essentially zero inflation after Dec-2014. Had the US been printing too much money and had all that money printing created inflation, we should have seen it here and we don't.


The forecast above is for CPIAUCNS (CPI for All Urban CoNSumers), again from the Federal Reserve. In this case, the model is forecasting the level of the CPI not the rates of change. It's very easy to see that the CPI is on the attractor path and well within the 98% prediction intervals, unlike the M1. You can pick particular blips (for example the red arrow) and become worried about inflation but the blips are random variation, all within probable ranges. 

The fact that the dynamics of M1 and the CPI are very different means they are being driven by different forces. The M1 is best explained by the state of the US economy and the CPI is best explained by the state of the World system. This should make some sense since the US is a globalized economy that controls its currency through the Federal Reserve and is at the same time the hegemonic leader of the World-system. These issues seem to escape most monetary models and economic models of inflation.

NOTE: In case you are wondering how good the state-space models are at predicting M1 one-month into the future (the typical criteria for econometric models), the forecast graph is presented below.


The models do an excellent job with very tight prediction intervals, getting wider of course into the future. The two models used for the forecasts are the USL20 model and the WL20 model. The USM1 models is here and the US CPI model is here. Explanations for how to use the models are available here.

QUESTIONS FOR FUTURE POSTS:
  1. What are the forces in the US Economy and the World System that drive monetary policy?
  2. Why was the M1 too high during the Dot-com Bubble and too low afterwards?
  3. During the Financial Crisis of 2007-2008, M1 growth was pretty flat. Was the US Federal Reserve trying to pop the Subprime Mortgage Bubble?
  4. What would be a reasonable value for Friedman's k-percent? In 2015, the annualized growth rate of the M1 attractor was about 5%. Should the value of k-percent increase, decrease or stay the same in the future?
  5. What are the forces in the World System that drive inflation?
  6. Did the US recently go through a Debt Crisis similar to ones in Europe and Latin America?
  7. Would harsher Austerity Policies produced a better or worse outcome in the US? Are stronger Austerity Policies needed in the future? 
  8. What about the performance of Federal Reserve policy instruments such as the Fed Funds Rate?
  9. What about the behavior of interest rates and the Zero Lower Bound problem?