Note: This article is an excerpt from Abolish Money (From Economics)! (Section 8). The book was published in January 2017, and the text and charts were (mostly) not updated. The fact that updates were not seen as necessary might be viewed as an editorial comment on the information content of money supply growth.
This essay follows up “Instability of Money Velocity,” Section 7. The difference between the analysis in that section and the previous is that the focus here is upon growth rates of money and nominal GDP, while velocity is an attempt to relate the level of the money stock to nominal GDP. We could imagine a situation where velocity is unstable (for any number of reasons), yet we can still draw conclusions from periods of “high” or “low” growth in monetary aggregates. In technical terms, we could hope that there is a correlation between money growth and nominal income growth (or inflation).
(NOTE: The figure above has been updated. The blue line is roughly in line with the end of the available data at the time of the writing of the book. The upward hook in the nominal GDP series was not available at the time of writing. I have left the text below unchanged from the book version.)
For example, the above figure depicts the growth rates of the M2 monetary aggregate and nominal GDP for the United States at the time of writing. It would be quite easy to look at the chart, and tell a story that inflationary risks are building up because of too-easy Federal Reserve policies.
The rest of this essay discusses why I am skeptical about drawing such a conclusion from the money numbers alone (please see the disclaimer in the next paragraph), even though there were periods in which rapid money supply growth did appear to contain useful information. When we look at longer runs of data, we find that money growth can decouple from other data, and so apparent predictive successes at any particular point could be chalked up to luck.
(As an aside, I do not currently disagree with the message of M2 growth at the end of 2016. Although I am not particularly bullish on nominal GDP growth, it is likely to pick up from its current subdued pace because of mean reversion. Furthermore, it appears likely that the new Republican administration will enact tax cuts, and so there should be a fiscal expansion in 2017.)
Advantages of Monetary Aggregates as DataThe beauty of the narrow monetary aggregates is that they are based on highly centralised instruments, and they are (almost by definition) easy to value. The components are either central government liabilities (which are tracked by the central government for obvious reasons), or bank deposit data that are tracked by bank regulators.
The advantage due to centralisation starts to break down once we start looking at wider monetary aggregates. Those aggregates include instruments such as money fund holdings or repos. Assuming the data are available, they are more difficult to assemble, and we would have greater concerns about the quality of the data. (Is the financial sector creating functionally similar products outside the sight of regulators?)
The technical advantages of money versus other economic variables that result from this centralisation and the simplicity of the data include the following points.
- Data can be collected at a higher frequency (weekly instead of monthly or quarterly).
- The lag in data publication can be much reduced.
- The data do not depend upon the manipulation of raw data, such as is required by price index or production data.
- The data generally do not need to be revised.
There are only a few disadvantages with the data, but they mainly revolve around attempts to create generic monetary aggregates that are comparable across countries. The instruments that are included in each aggregate (for example, M2 versus M1) vary, and are not strictly comparable between different regions. Changes in financial practices will mean that there may be structural breaks in behaviour. (This partly explains why economists who focus on money keep inventing new adjustments to the money numbers.)
These technical advantages are allied to some of the theoretical advantages of money. Broad money aggregates include instruments that are created by the act of borrowing. Any realistic analysis of a monetary economy tells us that nominal income growth is going to be at least correlated with increased debt levels. (One needs to be careful about this; the “multiplier” between increased debt levels and nominal income growth should not be expected to be stable.) Even though I am skeptical about money as a theoretical concept, it does give us a partial glimpse into the trends of debt creation, and the data are available in a timely manner.
However, this advantage of timeliness only matters if you are attempting to time the economic cycle at a high frequency – an extremely difficult task. The broad debt statistics that are available in a less timely manner (for example, from the Flow of Funds report) give a much better global overview of financing trends. My view is that we need to have a good idea of the current situation before we worry about trying to get a one-month lead on the rest of the market. Given that it took years for the consensus to grasp the slow nature of growth in the current cycle, I am not particularly concerned about a 1- to 2-month delay to get data that are more reliable. However, if a country lacks a reliable Flow of Funds report, monetary aggregates may be the only information that is available.
Money Growth Itself Should Not be the ConcernI am assuming here that we are not concerned about the money growth numbers by themselves, rather their effect on the economy. This is an assumption that is not shared by some of the more money-centric analysis frameworks; this section discusses this divergence.
Firstly, if a country has some form of currency peg system (such as a Gold Standard), and if the money supply is growing faster than the country’s holdings of the backing commodity (gold holdings), then the backing ratio will decline. This raises the odds of a speculative attack on the currency. This is a very real concern for historical analysis, but is not applicable to modern free-floating currencies.
Secondly, some Austrian economists define “inflation” to be money growth, and argue that everyone who calls rising consumer prices inflation is using the term incorrectly. I would argue that this is a misunderstanding of how languages operate – if everyone except you uses a word in a certain fashion, you are the one who does not understand the contemporary definition of the word.
In any event, even if we redefine the term inflation to be money supply growth, it is unclear why we should care about it. Central banks are typically not just given mandates to target generic “inflation,” their mandates are defined in terms of the movement in a particular consumer price index. The Austrian argument appears to rely on the Quantity Theory of Money, which would imply a constant velocity (which I discussed in the previous section).
Finally, Monetarists enjoy characterising monetary policy in terms of the rate of growth of the money supply. They argue that only unsophisticated bumpkins define monetary policy in terms of the level of interest rates. I disagree, as I believe that the central cannot set the level of the money supply. (This is the endogenous/exogenous money debate that is discussed in Section 10.) However, even if we grant the Monetarists a terminological victory and define monetary policy in terms of money growth, that does not tell us about the linkage between monetary policy and other economic variables. Once again, the object in this essay is to discuss such linkages.
OK, But Does Money Analysis Work in Practice?Although I am open to believing that money growth data could be useful for analysing the economy, I have not found any convincing means to do so. By itself, that is meaningless – my inability to find any useful relationship may just reflect a lack of imagination. That said, if there were investors who did find reliable relationships, we would expect (using market efficiency arguments) that the release of monetary data would move markets. We have not seen such behaviour since the 1980s.
There are a number of ways in which money growth could be useful for forecasting other economic variables (for example, price indices or nominal GDP). I list possibilities in order of their analytical strength.
- There is a strong relationship between the levels of money growth and the target economic variable. (The strongest possible relationship would be if “velocity” were constant.)
- The level relationship is imperfect, but there is a correlation between money growth and the target variable, with money leading the other variable. For example, this means that if money growth accelerates, we will expect the other variable to start rising within a few months – although we might not be able to guess the magnitude of the rise of the target.
- Money growth and the target variable are correlated, but money growth is essentially coincident with the other variable. Although this might not appear to be useful in theory, it is helpful in practice, given the reduced publication lag of money supply numbers.
- You can throw money growth numbers in with other variables into a multiple regression blender and get a model to estimate the target variable.
The rest of this essay examines the U.S. historical experience during most of the post-war era, looking at the linkage between money growth and nominal GDP growth. The analysis is visual, that is, the data are displayed graphically. Although academics may scoff at just eyeballing charts, such a presentation is probably less misleading than regurgitating the results of statistical tests.
Firstly, I do not think there is a reliable statistical link between money growth and nominal GDP growth. However, statistical tests can only determine if a particular relationship between variables is statistically significant (over a sample period). If I find tests that confirm my position (that there is no relationship), all I have done is to test a particular rule relating the variables. It may be that I have just missed another rule that is statistically significant, possibly as the result of my analytical biases. Secondly, I have seen a lot of dubious results “proved” statistically; looking at the actual data in chart form, and understanding the context, is a necessary step. (An example is the relationship between government debt levels and growth, which I discuss in the next section.)
U.S. Experience: Early Post-WarThe analysis here uses the monetary aggregate data that has been adjusted by the St. Louis Federal Reserve Bank staff. The earliest period examined is from 1948-1960, for which only the monetary base is available.
The early post-war era saw a deep recession in 1948-1949, which was the result of demobilisation after World War II. That period was unusual, as the economy was transitioning from a command economy that was managed almost solely with an eye on the war effort. As a result, we should be cautious about drawing conclusions from data for that era. (For example, one of the dubious statistical attempts to prove that high government debt levels reduced growth levels was entirely based upon that singular recession, which coincided with the high government debt levels of the immediate post-war era.)
Therefore, I would not put too much emphasis on the utter disconnect between nominal GDP growth and the growth rate of the monetary base that occurred in early 1951. At the point marked by the vertical line, nominal GDP growth actually peaked close to 20%, while the level of the monetary base was essentially stagnant. Subsequently, the growth rate in the monetary base accelerated while nominal GDP decelerated. Given my previous caution about this era’s data, all we can say is that there is no stable mechanical link between money supply growth and nominal GDP growth.
In the later parts of the sample period (post-1954), one might detect a correlation between the growth rate of the monetary base and nominal GDP growth. The growth rates were markedly different, which presumably reflects the transition away from the conditions of wartime finance. If one refers back to the charts in “Instability of Money Velocity” (Section 7), there was a trend decline in the velocity of money in this period.
U.S. Experience: The Monetarist PeriodDuring the period from 1960 to 1980 the data looked like what Monetarists predicted it would. (As I discussed in “Instability of Money Velocity,” the true end of stable velocity was even later, around 1992. However, the sample period was cut down to two decades in order to make it easier for readers to examine the data.)
The chart above shows the growth rates for the monetary base and nominal GDP for the period 1960-1980. There appears to be a strong correlation between the growth rates. However, the relationship is not perfect; for example, the growth rates diverged strongly in early 1976 (marked with a vertical line).
However, there are reasons to be suspicious of the monetary base as an indicator, as it misses innovations in financial practices. We would expect more success with a wider monetary aggregate. The figure below shows M2 growth, and it appears to have a better fit; in particular, it does not have the divergence that took place in 1976 (also marked with a vertical line). The preferred measure was M3, but the Federal Reserve stopped calculating it as a result of statistical budget cutbacks (which provided fodder for the believers in the Quantity Theory of Money who are also partial to suspicions of the motivations of government agencies).
However, switching monetary aggregates was not able to save money growth as an indicator over a longer time span, as discussed in the next section.
Recent U.S. ExperienceThis historical survey ends with the period from 1980 to 2007. (The period from the Financial Crisis to the present is deliberately ignored, as quantitative easing greatly affected the monetary aggregates, without having much of an impact on the economy. Since the argument in this essay is that money growth is supposed to be an indicator that naturally reflects economic activity – and not that the money supply can be set by central banks to achieve policy targets – this manipulation of the data by policymakers can be legitimately ignored.)
The choice of 1980 as a dividing line for periods was arbitrary, and as previously noted, money growth had almost as good a track record in the early 1980s as in the 1960-1980 interval. However, the experience in the mid-1980s was dubious, unless one takes the line that there can be very long lags between changes in money supply growth and the economy. The vertical line (marking February 1987) highlights the divergence that opened up between nominal GDP growth and money supply growth. The previously mentioned “long and variable lag” story is not particularly helpful when we have continuous business cycle oscillations (as was the case before the long expansions that started in the 1990s) – all sinusoidal signals of the same frequency can be viewed as either leading or lagging each other.
Once we hit the 1990s, it would be hard to see how changes in base money could have been viewed as predicting anything. The growth rate was accelerating from late 1989 until the 1990-91 recession hit, and thus was moving in exactly the wrong direction.
Furthermore, it is hard to square the rapid monetary base growth of the post-recession period with the tepid nominal GDP growth rates, and it lagged behind the acceleration seen during the late-1990s telecom boom.
The experience during the 2000s was similarly weak. The base grew rapidly during the weak post-recession period, while it was decelerating during the 2003-2005 period when the housing bubble was building up steam.
Finally, the monetary base “blew up” as an indicator around the year 2000. Federal Reserve Chairman Alan Greenspan bought into the story that the “Y2K bug” put the financial system at risk, and the Fed “flooded the system with liquidity” to forestall the risks. Once New Year 1999 passed, the Fed returned its balance sheet back to its usual size, creating the trailing downward spike. Many commentators blamed the final blow off in the tech stock bubble – that peaked in March 2000 – on the Fed’s overreaction to the Y2K bug. (My feeling is that the stock market bubble had its own momentum, and Fed policy could do little to deflect it.) Once again, erratic policy decisions wiped out the usefulness of the monetary base as an indicator.
Turning to M2, it was possibly even less reliable than the monetary base as an indicator during this sample period. The growth rate in M2 matched the behaviour of the monetary base in 1987, also diverging from GDP growth then (marked with a vertical line).
During the early 1990s expansion, M2 growth was essentially dead. Although it was a weak expansion, the economy was still growing. Money growth started to catch up with economic growth in the post-1995 period, but then shot off in the other direction – expanding very quickly during the weak growth period after the 2001 recession.
In summary, it was impossible to make any useful economic forecasts using money growth as an indicator during the post-1990 period. In fact, it was almost more effective as a contrary indicator (predicting the opposite of what will happen).
I started working in finance in 1998, and the completely erratic behaviour of the money supply as an indicator had been apparent for some time. My experience was that only a few diehard Monetarists would quote the money growth numbers, while everyone else dropped those charts from their chart packs.
Concluding RemarksMy analysis here was relatively brief and qualitative, and it would certainly be possible to look at the experience of other regions. It is entirely possible that money growth is a reliable indicator elsewhere. That is not contradicted by the experience in the United States that I examined here – it worked in the United States relatively well from 1960 to 1980 (at least). The problem is that you face the same risk as the Monetarists in the 1990s – you have no idea when it will stop working as an indicator.
It might be possible to change the definition of the money supply used, or apply some transformations to create a reliable indicator. (I tried doing that earlier in my career, and failed to do so.) However, you once again run into the same problem – your new series might work on historical data, but we need to wait years (if not decades) to see if it holds up against new data.
(c) Brian Romanchuk 2017, 2018