I have previously noted that Milton Friedman’s debating techniques and attitude towards facts were, erm, slippery to say the least. However, I focused primarily on his public face, and it seemed he could merely have adopted a more accessible narrative to get his point across, losing some nuance along the way. It could be argued that most are guilty of this, and it didn’t reflect Friedman’s stature as an academic.
Sadly, this is probably not the case. Commenter Jan quotes Edward S. Herman on Friedman’s academic record, giving us reason to believe that Friedman’s approach extended through to his academic work. It appears the man was prepared to conjure ‘facts’ from nowhere, massage data and simply lie to support his theories. With thanks to Jan, I’ll channel some of what Herman says, using it to discuss Friedman’s major academic contributions in general, and how his record seems to be rife with him torturing the facts to fit his theories.
The Permanent Income Hypothesis
The Permanent Income Hypothesis (PIH) states that a consumer’s consumption is not only a function of their current income, but of their lifetime income. Since people tend to earn more as they get older, this means that younger generations will tend to borrow and older generations will tend to save. The PIH has been a key tenet of economic theory since its inception, and Friedman won the Nobel Memorial Prize for it in 1976.
When discussing Friedman’s in-depth empirical treatment of the PIH, Paul Diesing (as quoted by Herman) found it wanting. He listed six ways Friedman manipulated the data:
1. If raw or adjusted data are consistent with PI, he reports them as confirmation of PI
2. If the fit with expectations is moderate, he exaggerates the fit.
3. If particular data points or groups differ from the predicted regression, he invents ad hoc explanations for the divergence.
4. If a whole set of data disagree with predictions, adjust them until they do agree.
5. If no plausible adjustment suggests itself, reject the data as unreliable.
6. If data adjustment or rejection are not feasible, express puzzlement. ‘I have not been able to construct any plausible explanation for the discrepancy’…
It does not surprise me that Friedman had to treat the data this way to get the results he wanted. For the interesting thing about the PIH is that it displaced a model that was far more plausible and empirically relevant: the Relative Income Hypothesis (RIH). The RIH argues that individual consumption patterns are in large part determined by the consumption patterns of those around them, so people consume to “keep up with the Joneses“. It was developed by James Duesenberry in his 1949 book Income, Saving and the Theory of Consumer Behaviour.
In his discussion of this apparent scientific regression, Robert Frank lists 3 major ‘stylised facts’ any theory of consumption must be consistent with:
- The rich save at higher rates than the poor;
- National savings rates remain roughly constant as income grows;
- National consumption is more stable than national income over short periods.
The PIH can easily explain the last two phenomena, as it posits that saving (and therefore consumption) is unrelated to current income. However, this same proposition required Friedman to dismiss the first phenomenon outright. He therefore suggested that the high savings rates of the rich resulted from windfall gains rather than income. A neat hypothesis, but unsubstantiated by the evidence: savings rates also rise with increases in lifetime income.
Conversely, Duesenberry’s theory is well equipped to explain all three of the listed phenomena. The RIH implies that poor consume a higher percentage of their income to keep up with the consumption of the rich. As society as a whole becomes richer, this phenomenon will not disappear, as the poor will still be relatively poor. Thus, the apparently contradictory first two points in Frank’s list are reconciled. It is also worth noting that the third point, that consumption is less volatile than income over short periods, can be explained by the RIH because people are used to their current standard of living, so they will sustain it even through hard economic times.
So why, despite fitting the facts without manipulating them, did the RIH fall out of favour? Presumably, it made economists (particularly those of Friedman’s ilk) uncomfortable because of its implications that much consumption was unnecessary and wasteful, that redistributing income might spur consumption and therefore growth, and because it did not rest on innate individual preferences but on the behavior of society as a whole. The idea of a consumer rationally making inter-temporal consumption decisions in a vacuum was just, well, it was real economics. The result is that Friedman’s poorly supported hypothesis shot to fame, while Duesenberry’s well supported hypothesis was forgotten.
NAIRU stands for ‘Non-Accelerating Inflation Rate of Unemployment’, and it implies that past a certain level of unemployment, workers will be able to demand wages so high that they will create a wage-price spiral. Hence, policy should aim for a ‘natural’ rate of unemployment, decided empirically by economists, in order to prevent the possibility of 1970s-style stagflation.
My first problem with the NAIRU is the way it is commonly seen as ‘overthrowing’ the naive post-war Keynesians who insisted on a simplistic trade off between inflation and unemployment. As I have previously noted, the originator of the curve, William Phillips, did not believe this; nor did Keynes; nor were the post-war Keynesians unaware of the possibility of a wage-price spiral. Furthermore, the NAIRU idea was really just a formalisation of a long standing conservative notion that we should keep some percentage of people unemployed for some reason. In this sense, the launch of the NAIRU was more a counter revolution of old ideas than a novel approach.
However, the real issue is whether the NAIRU is empirically relevant, and it seems it is not. First, as Jamie Galbraith has detailed, there is little evidence that unemployment has an accelerating effect on inflation at any level. Furthermore, empirical estimates of the NAIRU seem to move around so much, depending on the current rate of unemployment, that the idea has little in the way of predictive implications. The data simply do not generate a picture consistent with a clear value of unemployment at which inflation starts to accelerate;: we are far better off pursuing full employment while keeping numerous inflation-controlling mechanisms in place.
“OK” you say. “Perhaps the NAIRU does not exist. But what about this was disingenuous on Friedman’s part?” Well, the notion that the interplay between workers and employers is a key determinant of the rate of inflation flat out contradicts Friedman’s oft-repeated exclamation that “inflation is always and everywhere a monetary phenomenon”. If inflation is purely monetary, then the level of unemployment should not affect it at all! However, for whatever reason, Friedman was prepared to endorse both the NAIRU and his position on inflation simultaneously.
The Great Depression
Friedman’s Great Depression narrative was probably his biggest attempt to rehabilitate capitalism in a period where unregulated markets had fallen out of favour. He blamed the crash on the Federal Reserve for contracting the money supply in the face of a failing economy. This always struck me as strange – he was, in effect, arguing that the Great Depression was the fault of ‘the government’ because they failed to intervene sufficiently. This implies that the real source of the Great Depression came from somewhere other than the Federal Reserve, and therefore its sin was more one of omission than commission. Even if we accept the idea that the Great Depression was worsened by the action (or inaction) of central banks, Friedman is being disingenuous when he says that the Great Depression was “produced” by the government.
However, even Friedman’s own figures fail to support his hypothesis: according to Nicholas Kaldor, the figures show that the stock of high powered (base) money increased by 10% between 1929 and 1931. Peter Temin came to a similar conclusion: using the same time period as Kaldor, real money balances increased by 1-18% depending on which metric you use, and the overall money supply increased by 5%. Though base money contracted by about 2% at the onset of the crash, a contraction this small is a relatively common occurrence and not generally associated with depressions.
There is then the issue of causality. In many ways Friedman assumed what he wanted to prove: that the money supply is controlled by the central bank. Yet there are good reasons to doubt this, and believe that movements in income instead create a decrease in the money supply, which would absolve the central bank of responsibility. When economists such as Nicholas Kaldor pointed out this possibility, Friedman reached a new level of disingenuous (the first paragraph is Friedman’s comment; Kaldor responds in the second):
The reader can judge the weight of the casual empirical evidence for Britain since the second world war that Professor Kaldor offers in rebuttal by asking himself how Professor Kaldor would explain the existence of essentially the same relation between money and income for the U.K. after the second world war as before the first world war, for the U.K. as for the U.S., Yugoslavia, Greece, Israel, India, Japan, Korea, Chile and Brazil?
The simple answer to this is that Friedman’s assertions lack any factual foundation whatsoever. They have no basis in fact, and he seems to me to have invented them on the spur of the moment. I had the relevant figures extracted from the IMF statistics for 1958 and for each of the years 1968 to 1979, for every country mentioned by Friedman and a few others besides… Though there are some countries (among which the US is conspicuous) where in terms of the M3 the ratio has been fairly stable over the period of observation, this was not true of the majority of others.
Bottom line? Friedman had to assume his conclusion – that the money supply was in control of the Federal Reserve – in order to reach it. Yet, based on his own numbers, his conclusion was still false, as the money supply increased over the ‘crash’ period from 1929-1931. When Friedman was pushed on these matters, he simply made things up. However, lying hasn’t helped him escape the fact that his theory of the Great Depression is false.
Milton Friedman’s academic contributions do not stand up to scrutiny. Friedman seemed to be prepared to conjure up neat, ad hoc explanations for certain phenomena, simply asserting facts and leaving it for others to see if they were true or not, which they usually weren’t. He selectively interpreted his own data, exaggerating or plain misrepresenting it in order to make his point. Furthermore, his methods should be unsurprising given his incoherent methodology, which allowed him to dodge empirical evidence on the grounds of an ill-defined ‘predictive success’, something which sadly never materialised. In almost any other discipline, Friedman’s attempts at ‘science’ would have been laughed out of the room. Serious economists should distance themselves from both him and his contributions.