Portal de Periódicos da CAPES

Interest rates or quantity of money? Edward Nelson on Milton Friedman

2021; Wiley; Volume: 41; Issue: 2 Linguagem: Inglês

10.1111/ecaf.12467

1468-0270

Tim Congdon,

Economic Theory and Institutions

Edward Nelson, Milton Friedman and Economic Debate in the United States, 1932–1972. University of Chicago Press (2020). Volume 1: 784 pp. ISBN: 978-0226683775 (hb, £40.00); 978-0226683805 (e-book, £34.29). Volume 2: 640 pp. ISBN: 978-0226684895 (hb, £40.00); 978-0226684925 (e-book, £38.00) Milton Friedman was awarded the Nobel prize for economics on 14 October 1976. The prize was “for his achievements in the fields of consumption analysis, monetary history and theory, and for his demonstration of the complexity of stabilization policy”, to cite the commendation from the Royal Swedish Academy of Sciences (1976). That day Friedman was – in his own word – “barnstorming” in Michigan to promote a balanced budget amendment to the state's constitution. He heard about his Nobel prize in a parking lot in Detroit, where he and his hosts were amazed by a large throng of journalists. As I stepped out of the car, a reporter stuck a microphone in my face and said, “What do you think about getting the prize?”. I said, “What prize?”. He said, The Nobel prize”. Naturally I expressed my pleasure at the information. The reporter then said, “Do you regard this as the pinnacle of your career?” or something to that effect, and I said no. I said I was more interested in what my fellow economists would say about my work fifty years from now than about what seven Swedes might say about my work now. (Spencer & Macpherson, 2014, p. 55) Famously or notoriously, depending on one's point of view, Friedman was controversial. But – whether one loves or loathes him – he is acknowledged to have had a double achievement matched only perhaps by J. M. Keynes in the pantheon of great economists. He both transformed the academic discipline to which he devoted his exceptional intellectual abilities, and contributed massively to the public life of his own nation, the United States of America, and to that of others, including the United Kingdom. Nelson recognises the dual nature of Friedman's career, explaining with great skill the blending and interdependence between the scholarly output and the barnstorming. Much of the barnstorming was in the White House and on Capitol Hill, as well as in Midwest car parks, and it had major consequences for US policymaking. From the early 1950s Friedman reasserted the empirical validity of the quantity theory of money, when the consensus of his profession was against him. He then emphasised the potency of monetary policy relative to fiscal policy, and argued that monetary policy should be directed towards the attainment of price stability, not full employment. His themes formed a veritable ‘monetarist counter-revolution’ against ‘the Keynesian revolution’ inaugurated by Keynes's 1936 The General Theory of Employment, Interest and Money. Moreover, Friedman's messages were so effectively delivered that in the 1970s and 1980s they led to a policy refocusing along the lines he favoured. Arguably, this refocusing was responsible for the Great Moderation, a period of benign macroeconomic outcomes (with low inflation and steady growth) that lasted in the USA from the mid-1980s to the Great Recession of 2008 and 2009. Nelson's book is impressively thorough, detailed and well-expressed, and reading it is a pleasurable immersion course in Friedman's activities. All the same, it was written by Nelson, not by Friedman. It is inevitably Nelson's interpretation of Friedman, and reflects a balance between the author and his subject. That is fair enough. But at times the balance is not right, with too much Nelson and not enough Friedman. This review will argue that, despite its author's commendable industry and the depth of his knowledge, the book is far from being the definitive intellectual biography that Friedman deserves. Nelson is renowned for his mastery of the modern journal literature in macroeconomics, a knowledge so encyclopaedic and extensive that it must be unique in the Federal Reserve's research departments, and may be unmatched even in academe. Unsurprisingly, he is in awe of the great names of modern macroeconomics like Hicks, Modigliani and Samuelson, who were able to put together – in his phrase – “fully-specified structural models” (i, p. 182). In macroeconomic models of this sort aggregate demand and aggregate supply are represented by equations, and the economy's equilibrium – notably real output and hence employment – can be described diagrammatically by the intersection of lines for the aggregate demand and supply functions. By contrast, Friedman offered reduced-form models, with one or two equations in which nominal output was determined above all by the quantity of money, in line with his quantity-theory commitments. Unlike structural models with many equations (and often very complex and arcane mathematics), reduced-form models of the MV = PT variety are not altogether kosher in the top-rated journals. Nelson views this as a weakness in Friedman's oeuvre, invoking the authority of Allan Meltzer to support his position. Meltzer, also a prominent figure in American monetary economics in the late twentieth century, was often on Friedman's side in policy debates. He is reported as having stated that Friedman “never developed a model”. According to Nelson, Meltzer had tried to write down what he thought might be the model that Friedman and long-time co-author, Anna Schwartz, “had in mind”, but had struggled. Further, these remarks meant that there was “an implicit model in [Friedman's] empirical and narrative work” (i, p. 182). Nelson has therefore taken upon himself the task of providing the equations for a latter-day Friedmanite structural model. A conceptual framework which might perhaps have been in Friedman's head, but was never properly written down, is therefore to be made explicit by Nelson. In particular, the argument of chapter 5 culminates in an aggregate spending equation which determines real output, in the same fashion as the IS curve in Hicks's IS/LM depiction of Keynes's General Theory. The equation, presented in logarithmic form, makes output dependent on its lagged value, two interest rate terms, and a shock term incorporating “exogenous disturbances” to the components of aggregate demand. In Nelson's judgement, the impact of interest rates would ideally be captured by “a vector of interest rates, rather than a single short-term interest rate” (emphasis in original). But, in the equation as stated, the two interest rates are one short rate and one long rate, and that is all (i, pp. 226–7). Is this equation what Friedman believed, what Nelson believes that Friedman believed, or what Nelson believes? The two-page section (on pages 226 and 227 of the first volume) containing the new IS function relies on three footnotes. The first begins, “In standard, linearized New Keynesian models, the real long-term interest rate does not matter in its own right for spending decisions; only the component of the real long-term rate that corresponds to the expected path of short-term rates does so.” This proposition is attributed to a 1999 paper by Julio Rotemberg and Michael Woodford, and a 2004 paper co-authored by three economists, one of whom is Nelson himself. The following sentence does refer to a 1974 cassette tape by Friedman. But the existence of the cassette tape is news to the current reviewer and probably other economists, and frankly the reference is obscure (i, p. 531).11 The notes are numbered 200 to 202. The text discussion relates to note 200. The second and third footnotes do not mention Friedman. Rational expectations will be used, and adaptive expectations will not be regarded as integral to Friedman's theory. (i, p. 188) The assets considered in spending and portfolio choices other than money and short-term securities will be limited to a single, long-term, fixed-interest nominal security. (i, p. 189) Lags from monetary policy to real GDP and inflation will not be modelled. (i, p. 191) Friedman equivocated about rational expectations. Sure enough, he was always complimentary to Robert Lucas, usually seen as the seminal thinker in the ‘rational expectations revolution’.22 The key paper is usually seen as Lucas (1972). But that is not the same thing as endorsing the idea. Rational expectations are forward-looking and based on the validity of economic theories which are assumed to be held by individuals and to determine their behaviour. Friedman instead preferred the hypothesis that expectations are adaptive and respond to past events. In his words, “the most important single thing [in understanding the short-run non-neutrality of money, and its consequent ability to cause movements in demand and output] is the tendency for expectations to be backward-looking and to be adjusted slowly”. Moreover, in interviews later in his career he found “the notion of ‘correct rational expectations’ … very hard to give much content to” (Samuelson & Barnett, 2007, p. 137).33 The quotations are from Friedman's responses to an interview in 2000 by John Taylor. One author (Rivot, 2016, p. 222) has used the word ‘dismissal’ to describe Friedman's attitude towards rational expectations, although this perhaps goes too far. A fairer way of putting it might be to say that he was curious, but sceptical.44 Milton Friedman and Anna Schwartz (1982, pp. 556–7) identify “a limitation of much recent work on rational expectations”. For Nelson to cast doubt on the relevance of lags to money's impact on the economy is surprising in general terms, as lags are present and obvious in any business context. It is also surprising in this particular instance, as Nelson has penned a superb seven-page account of Friedman's evolving views on the subject in chapter 15 in his second volume. Nelson deems Friedman's initial statements about lag structures – notably in a full chapter ‘The lags in effect of monetary policy’ in his 1969 collection of essays The Optimum Quantity of Money – as “precarious” (ii, pp. 232–9).55 Nelson uses the word ‘precarious’ on page 232. Friedman's chapter ‘The lag in effect of monetary policy’ appears in Friedman (1969, pp. 237–60). The chapter drew on a 1961 paper in The Journal of Political Economy. Nelson's reservations are well-stated and seem justified. All the same, they leave Nelson open to the charge of disloyalty, since Friedman devoted much energy to the analysis of the lags – “the long and variable lags”, as he understood them – between changes in the quantity of money and changes in the price level. The quantity theory of money can be viewed mostly as a theory of the price level, but by common consent the effect of change in the rate of money growth on inflation typically comes after an effect on real output. Friedman said as much on numerous occasions (1991, p. 15). Nelson might insist that he wants to derive an IS function relating to real output, not to make statements about the price level, and that the plethora of comments on lag structures in Friedman's papers is not germane to the IS function he is proposing. Maybe so, but Friedman (1969, p. 140) made another important and distinctive argument here. According to the evidence he had examined, the response of financial markets – notably, of stock market prices – to a change in money growth came even earlier than the response of output. This raises many questions. How do money, financial markets and the real economy interact? What were Friedman's views on the range of assets that matter in the interactions? And how does Nelson report on and handle these views? In his 1936 General Theory Keynes advanced a new theory of the determination of ‘the rate of interest’, by which he meant the yield on bonds. This theory – the so-called ‘liquidity preference theory’ – said that, in equilibrium, the rate of interest had to be such that investors were happy with the balance between money and bonds in their portfolios. When combined with the notion that the quantity of investment was sensitive to bond yields, and that national income was a stable multiple of investment, Keynes had assembled the ingredients for the IS curve of Hicks's IS/LM construction. In The General Theory Keynes had the common sense to recognise that in reality the level of the stock market was vital to business confidence and investment decisions, even if several pages mocked the stock market and its operators. All the same, in the last 15 or so years of his life Keynes made statements to the effect that money had its impact on the economy only through changes in ‘the rate of interest’, where – to repeat – this meant the yield on bonds.66 No economist would dispute that – to the extent that increases in the quantity of money lower bond yields, in line with Keynes's liquidity preference theory – there is a positive effect on aggregate demand. But it is an extraordinary and unjustified leap to go from here to claiming that an increase in the quantity of money boosts aggregate demand only if it lowers the bond yield. Unfortunately, Keynes was bamboozled by his own theory and did sometimes make that leap. In February 1936, in an unpublished letter to The Times, he wrote, “…if the creation of credit [and money] has no effect on the rate of interest, it will have no effect on prices”. The claim was absurd, both then, now, and in all the intervening decades. Bonds are only one asset among many, while fluctuations in the value of all the bonds held in the economy are usually less than those of any of quoted equities, residential and commercial real estate, and unquoted equities and business assets of all sorts.77 The reviewer emphasised the point in his article in the February 2021 issue of this journal (Congdon, 2021; see particularly pages 25–7). (See Table 1 – relating to the American economy – for the smallness of changes in the value of households' bond holdings relative to that of other assets. In the period of 75 years covered in the table, the average annual change in the value of bonds held by households and non-profits was slightly above 0.4 per cent of net worth at the end of the previous year, if the sign is ignored. That average change is less than for any of the other asset categories included.) Nevertheless, Keynes's General Theory had such prestige – particularly after it was converted into textbook format by Samuelson in 1948 – that many economists were persuaded that money had its impact on the economy only through changes in ‘the rate of interest’.88 The discussion of money and interest rates in Samuelson (1948, pp. 287–306) does contain an isolated reference (p. 302) to investors balancing money against equities and real estate, but the argument follows the lines of Keynes's General Theory. It is taken for granted that changes in money matter because, above all, they affect bond prices and yields. Some of the best material in Milton Friedman and Economic Debate in the United States, 1932–1972 is on Friedman's drastic change of view about money's importance in the 1940s, particularly in the three years from 1948 to 1951. Whereas as a young man Friedman was attracted to Keynesian income–expenditure theorising, he became in middle age a forthright and articulate exponent of the quantity theory of money. The major influence on this intellectual upheaval must have been evidence, evidence from statistical series mainly, but also evidence from a large body of real-world facts and figures. Surely Friedman was aware of the relatively minor role of bonds in household portfolios. Given that, it was understandable that he should see money's effects on the economy as not being exclusively via bond yields, as in Keynes's liquidity preference theory and Hicks's IS function. One of the earliest demonstrations of his more spacious and wide-ranging view was in his celebrated introduction, ‘The Quantity Theory of Money: A Restatement’, to Studies in the Quantity Theory of Money (Friedman, 1956). Equities were seen as being distinctive in yielding a positive real return, while physical non-human capital goods offered a return that at least compensated for movements in the price level. Changes in the price level were therefore regarded as having a significant influence, via the yields on equities and physical non-human capital goods, on the quantity of money demanded. Friedman noted towards the end of section 18 that “the attack on the quantity theory associated with the Keynesian under-employment analysis” relied on the notion that “the only role of the stock of money and the demand for money is to determine the interest rate” (1956, p. 17). This narrow understanding of money's role was indeed one of the Keynesian totem-poles that Friedman wanted to knock down. That was the rationale for tackling the subject “from the broadest and most general point of view”. To put the matter another way, in his 1956 ‘restatement’ – often seen as the launching-pad of the ‘monetarist counter-revolution’ – Friedman was opposed to a specification of the demand to hold money in which the only non-income variables were bond yields. His aversion to overplaying the role of bond yields in monetary economics was reiterated in an article on ‘Money and Business Cycles’ (Friedman & Schwartz, 1969), first published in 1963. In this article Friedman and Schwartz offered “a tentative sketch of the mechanism transmitting monetary changes”. They hypothesised two kinds of money injection, one adding to the monetary base held by the banks and through the base multiplier boosting the quantity of money, and the other a direct addition to the quantity of money through open market purchases from non-banks. In their assessment a sequence of portfolio adjustments would ensue. Banks would expand their loan books, while non-banks would shift their attention from shorter-dated, safe bonds to “higher-risk fixed-coupon obligations, equities, real property, and so forth” (1969, p. 230). The higher prices of securities would affect the valuation of all capital assets, and stimulate both consumption and investment. It followed that, “The monetary stimulus is, in this way, spread from the financial markets to the markets for goods and services” (1969, p. 231). The ‘tentative sketch’ did not have much traction in macroeconomic research in subsequent decades. It nevertheless anticipated many papers about ‘the portfolio rebalancing channel’ written about the effects of central bank asset purchases – or ‘quantitative easing’ – after 2008.99 The reviewer noted the affinities between the portfolio rebalancing literature and Friedman and Schwartz's 1963 ‘tentative sketch’ in Congdon (2021, pp. 30–1). Friedman and Schwartz placed emphasis on the need for “a much broader view” of the processes at work than was common among economists at that time, when Keynesianism was ascendant. In their view, the asset price fluctuations relevant to macroeconomic outcomes were not just of “government and private fixed-interest and equity securities traded on major financial markets”, but of “other assets, including consumer durable goods, consumer inventories of clothing and the like” (emphasis added). Moreover, in their words “it is necessary to make [the] ‘rate of interest’ an equally broad construct, covering explicit or implicit rates on the whole spectrum of assets” (1969, p. 231). Alternatively put, Friedman and Schwartz thought that rates of discount applied – often tacitly – to the valuation of income streams from non-bond, unquoted assets were at least as much part of the transmission mechanism as the yields on quoted bonds. Nelson is correct when he says that they favoured a multi-yield approach. But his proposal of an IS curve with two yields on bonds is naughty, as it ignores the point that they regarded yields on all assets as vital to the analysis. Friedman and Schwartz seem to have been prepared to pay attention, as serious macroeconomists, even to the portfolio balance between money and wardrobes of fashion items!1010 A reference to clothes appears again in the 1982 Monetary Trends, in an account of the transmission mechanism that recalls Friedman's 1956 ‘restatement’ and the 1963 Friedman and Schwartz ‘tentative sketch’ (Friedman & Schwartz, 1982, p. 58). Friedman's readers cannot miss frequent mentions of equity markets in his discussions of the transmission mechanism. Stocks and shares are after all more significant in household wealth than clothing! But Nelson wants to diminish the role of equities in Friedman's analysis of macroeconomic instability. He devotes several pages of his second volume to debates between Friedman and Paul Samuelson, using them to cast doubt on the place of equities in the Friedmanite transmission mechanism. Both Friedman and Samuelson had columns in Newsweek for many years, and Samuelson took the opportunity in the very first of these (on 19 September 1966) to insert the famous witticism that the stock market had predicted nine of the last six recessions. Nelson says that, “the quip relayed an important position – the disconnection of equity prices from US economic activity – for which many more economists would join Samuelson as advocates”. He further remarks that Friedman was among these economists, with “doubts about both the dependence of the stock market on economic fundamentals, and the feedback from equity prices to the economy” that would “deepen in the late 1960s and shift him towards a very Samuelsonian perspective on the stock market's significance” (ii, p. 181). These statements are questionable, to put it mildly. Nelson cannot dispute the remarks in the 1956 ‘restatement’ and the 1963 ‘tentative sketch’ which have been quoted above and speak for themselves. If his contention is to have any validity at all, it relies – as he says – on a supposed move towards “a very Samuelsonian perspective” from the late 1960s. But Nelson overlooks two examples of counter-evidence. First, the 1982 Monetary Trends volume (again co-authored with Anna Schwartz) presented a description of the transmission mechanism similar to that in ‘the tentative sketch’ almost 20 years earlier, with this description retaining an important role for equities.1111 A quotation from Monetary Trends says it all. In the view of Friedman and Schwartz (1982, p. 57), the Keynesians “regard spending as affected only ‘indirectly’ as the changed interest rate [due to a change in the quantity of money] alters the profitability and amount of investment spending…We, on the other hand, stress a much broader and more ‘direct’ impact on spending.” Second, one has to wonder whether Nelson has noticed Friedman's last published academic paper, in the 2005 Journal of Economic Perspectives when Friedman was 93. This paper – entitled ‘A Natural Experiment in Monetary Policy Covering Three Episodes of Growth and Decline in the Stock Market’ (Friedman, 2005) – reviewed three periods in which stock markets had risen strongly in celebration of technological breakthroughs, but had then crashed. The first of the stock market booms was in the USA in the ‘Roaring Twenties’, the second in Japan in the 1990s and the third again in the USA in the 1990s. Friedman presented a chart in which he showed that quite strong money growth had continued in the USA after the bursting of the dotcom bubble in 2000 and 2001, whereas in Japan money growth had stalled from the early 1990s and in the sequel to the USA's 1929 stock market crash the quantity of money had collapsed. It is clear from the text of the article that Friedman saw high money growth as a key driver of the rises in share prices in the booms. Moreover, he interpreted the different monetary experiences after the crashes as crucial in explaining events. The USA had recovered well after 2001, whereas it endured the Great Depression between 1929 and 1933. Meanwhile, Japan's gross domestic product had gone sideways in the 1990s, just like its money supply. In Friedman's view, here was a “natural experiment”, with “the controlled conditions of the experimenter's laboratory” (2005, p. 145). To quote from the concluding paragraph of Friedman's final article, “[W]hat happens to the quantity of money has a determinative effect on what happens to national income and to stock prices” (2005, p. 150). Why would Friedman have written in these terms if, like Samuelson, he wanted to poke fun at the stock market? Friedman was a passionate supporter of free-market capitalism, of which the stock market is among the most emblematic institutions. Readers of Milton Friedman and Economic Debate in the United States, 1932–1972 must be warned that its author attributes to Friedman views that he did not hold. Nelson has originated an IS function with two interest rate terms that is nowhere in the Friedman oeuvre, and defends and promotes it. This IS function belongs to Keynesian or New Keynesian thinking; it has nothing to do with the quantity theory of money. But the fundamental issue is not to determine who believed in a particular doctrine many decades ago or who believes in it now, or even to arbitrate on Friedman's exact meaning. Instead, our most basic concern – as Friedman himself would insist – must to discover how the economy does in fact behave. Are interest rates or money-quantity variables more successful in macroeconomic explanation? One reason – almost certainly the main reason – that Friedman and Schwartz disliked appeals to interest rates in macroeconomic testing was simple, that interest rates did not work in explaining fluctuations in national income growth. To quote from a representative Schwartz paper, written on the tenth anniversary of the 1959 report of the UK's Radcliffe Committee, “The correlations between the level or rates of change in interest rates, on the one hand, and rates of change in nominal income, prices and output, on the other, are considerably worse than those between rates of change in the quantity of money and these magnitudes” (Schwartz, 1969, p. 175). Since Nelson's second volume is dedicated to Schwartz, who is described as a “continuing inspiration”, he presumably would not question this finding about the periods that she, with Friedman, examined before 1969. But Nelson's predilection for interest rates in macroeconomic explanation – and his proposal of a Friedmanite IS function – might be justified if the American economy had behaved differently since then. The necessary exercise is – for the USA in the 50 years 1970–2019 inclusive – to regress the change in GDP on the change in the quantity of money and appropriate interest rate terms. The Appendix summarises the results, using quarterly data for the annual rates of change, with the quantity of money being the broadly defined M3 concept, and the two interest rates being the levels of the Fed funds rate and the 10-year Treasury bond yield. The exercise does not pretend to be refined econometrics: its purpose is the unambitious one of comparing the relative statistical significance of interest rate and money-quantity variables in explaining the rates of change of the USA's real and nominal gross domestic product. Anyhow the results are clear-cut. Over the 50 years the average rates of increase in money and nominal GDP were 7.3 per cent and 6.3 per cent respectively. The long-run relationship between changes in the quantity of money and nominal GDP is exactly in line with the prognosis made by Friedman in his 1959 Millar lectures at Fordham University in New York. (Yes, exactly! In his 1959 lectures Friedman said that his favoured policy rule was for broadly defined money to grow at a roughly constant rate of between 3 and 5 per cent a year, with this intended to deliver price stability after an allowance were made for “a secular decrease in velocity” of 1 per cent a year; Friedman, 1992, p. 91.) Meanwhile the short-run relationship between changes in money and GDP was not glaringly inconsistent with theoretical preconceptions. All the regression coefficients on the money terms in six estimated equations were positive (and so correctly signed), and they met the usual significance test.1212 The t statistic on the regression coefficient was above two. Table 1 shows the relationship between changes in money and nominal GDP, and reports some of its key features. However, it must be conceded that the short-run relationship was poor. Coefficients of determination of over 0.8 were obtained in two equations, but only by suppressing the intercept terms. But – if the short-run relationship between changes in money and GDP was poor – that between the levels of interest rates and the changes in GDP was dreadful. Twelve coefficients on the interest rate terms were calculated in six equations. Of these the majority (eight!) were positive and therefore incorrectly signed. (The higher interest rates were, the faster was the rate of growth of GDP.) Of the four that had the correct sign, two can be dismissed as not meeting the usual significance test, while the two that did meet that test did so only just.1313 The t statistics on the regression coefficients were minus 2.50 and minus 2.80. See Appendix for details. The conclusion has to be that Schwartz's 1969 generalisation held up in the following 50 years. The facts prompt two questions: why make such a fuss about interest rates when considering the forces that determine GDP? And is the IS function any use at all? Nelson's answer might be that central banks – including the Federal Reserve – nowadays organise policy mainly by setting a short-term