Some Monetary Facts

Questions & answers

Q1. What motivates the paper, and what question is it trying to answer?

The paper opens from the Federal Reserve’s statutory mandate – since the Employment Act of 1946 and the Full Employment and Balanced Growth Act of 1978 – “to maintain long run growth of the monetary and credit aggregates commensurate with the economy’s long run potential,” and asks directly: “Does [the Fed] have the ability, through its monetary policy, to control these economic variables?” (p. 2, quoting FR Board 1990). The authors state their purpose is “to improve upon past attempts to determine what the facts are” about the links between money growth, inflation, and output growth (p. 2), on the view that “the facts about those links can help determine how well the Fed can do its job.”

Q2. Why does the paper study cross-country correlations rather than a single country’s time series?

The authors deliberately study a cross-section of countries, rather than one country’s time series, because “if we were to study a single country, the correlations we obtained could be an artifact of the particular policy rule or rules being followed there” (p. 3): a central bank following a constant-money-growth rule would show no correlation between money growth and inflation in its own time series even if money fully determines inflation, while a central bank following a feedback rule from inflation to money growth would show a spuriously perfect correlation. “By using a large cross section of countries,” the authors reason, “we hope our correlations will be free of policy rule influences” (p. 3), since this independence from any one country’s rule is what lets the correlations “lend support to causal hypotheses that yield predictions consistent with” them, even though “correlations are not direct evidence of causality” (p. 3).

Q3. What data and sample does the study use?

The study covers 110 countries with at least 10 years of data, using the CD-ROM version of the IMF’s International Financial Statistics for 1960-1990, calculating for each country the geometric average annual growth rate of real GDP, consumer prices, and three definitions of money – M0 (currency plus bank reserves, IFS line 14), M1 (“money easily used in transactions,” line 34), and M2 (M1 plus line 35) (pp. 3-4). Two subsamples are used as “a crude test of robustness”: 21 OECD countries and 14 Latin American countries, chosen because each subsample is “more homogeneous…in terms of available technology, education, and level of development of financial (and other) institutions” than the full sample (p. 4, footnote 1 lists both country lists in full).

Q4. What is the first fact – the money growth/inflation relationship – and how strong is it?

“There is a high (almost unity) correlation between the rate of growth of the money supply and the rate of inflation,” holding across all three definitions of money and across the full sample and both subsamples; for the full 110-country sample the correlation coefficient between money growth and inflation is 0.925 or higher for every money definition, with the broader M1 and M2 aggregates (0.958 and 0.950) correlating slightly more closely with inflation than the narrower M0 (0.925) (Table 1, pp. 3-4). The OECD and Latin American subsamples confirm the pattern (correlations always above 0.89, again weaker for M0 than for the broader aggregates), and plotting average M2 growth against average inflation for all 110 countries (Chart 1) shows the relationship is close to linear with a slope near unity – consistent with the quantity-equation identity m+v=p+y (eq. 2) holding with velocity and output growth roughly constant across the long run – though the fact that the 45-degree line through the grand means does not pass through the origin implies “a central bank cannot generate a particular long-run rate of inflation by choosing an equal long-run growth rate for the money supply” (pp. 4-5).

Q5. What is the second fact – the money growth/real output relationship – and why does the OECD subsample differ from the rest?

“There is no correlation between the growth rates of money and real output. This holds across all definitions of money, but not for a subsample of OECD countries, where the correlation seems to be positive” (Table 3, p. 6): for the full 110-country sample all three money-definition correlations with real output growth are negative but smaller in magnitude than -0.05, not statistically different from zero given a standard deviation of 0.097; the Latin American subsample correlations (-0.17 to -0.25) are likewise insignificant. The 21-country OECD subsample, however, shows correlations between 0.51 and 0.71 (highest for M0), and a regression of real output growth on money growth yields a slope coefficient of about 0.1 for all three definitions, meaning “increases in money growth are associated with increases in real output growth about one-tenth as large” (pp. 7-8); the authors caution this need not be causal, since it “may reflect not a causal relationship, but rather a similarity in central bank policy” across OECD countries, for example a shared practice of letting money growth respond to real output growth through feedback rules (p. 8).

Q6. What is the third fact – the inflation/real-output relationship – and what role does the Nicaragua outlier play?

“There is no correlation between inflation and real output growth. This finding holds across the full sample and both subsamples” (Table 5, p. 9), but this conclusion requires handling an outlier: the raw full-sample correlation is -0.243 (significant given a 0.097 standard deviation), driven by “a definite outlier” – Nicaragua, which had real output growth of -12% and inflation of 52% during a period of civil war – and once Nicaragua is dropped the correlation for the remaining 109 countries falls to -0.101, no longer significantly different from zero (p. 9). The OECD subsample correlation is positive (0.390) but not significant at the 0.1 level, and the Latin American subsample correlation (-0.342, after excluding the outlier) is likewise insignificant at that level, so the authors conclude the three results “are consistent” with no robust relationship, even though the full-sample and subsample point estimates look superficially different (p. 9).

Q7. How do these findings compare with earlier studies, and where do they depart from the literature?

For money growth and inflation, the paper’s near-unity correlation matches earlier work by Vogel (1974), Lucas (1980), Dwyer and Hafer (1988), Barro (1990), Pakko (1994), Poole (1994), and Rolnick and Weber (1994) – the last of whom find the money-inflation correlation is close to unity “for fiat money regimes, but much lower, 0.61 or less, for commodity money regimes” (Table 2, pp. 5-6) – while apparent counterevidence from Sargent (1982) and Smith (1985), who document episodes where money growth and inflation diverge sharply, is “not inconsistent” with the paper’s results once one notes that the data points Sargent and Smith discuss fall below, rather than off, the paper’s own 45-degree line in Chart 1 (p. 6). For money growth and real output, prior studies disagree (Table 4, p. 10): Kormendi and Meguire (1985) find a negative correlation, Geweke (1986) finds money “superneutral” (no correlation, inconsistent with this paper’s OECD-subsample finding), Dwyer and Hafer (1988) find a statistically insignificant negative correlation, and Poirier (1991) finds mixed results country by country – a divergence the authors attribute to “the sensitivity of the results to the subsample chosen” (p. 8). For inflation and real output, the paper’s null finding departs from several studies reporting negative correlations (Fischer 1983, 1991; Kormendi and Meguire 1985; Altig and Bryan 1993; Ericsson, Irons and Tryon 1993; Barro 1995; Table 6, p. 10), a divergence the authors link to Levine and Renelt’s (1992) finding that the inflation-output relationship “is not robust to the inclusion of additional variables” in growth regressions (p. 10).

Q8. What do the authors conclude about what central banks can and cannot do?

“To the extent that we can interpret the long-run relationships we get here as causal relationships,” the authors draw two policy conclusions (pp. 10-11): first, since the money growth/inflation correlation is close to one, “we can adjust long-run inflation by adjusting the growth rate of money,” but this “does not mean that we can hit specific inflation targets,” because doing so “requires accurately predicting the growth rates of real output and velocity, something that has not been done well” – a point the paper says “should cause concern about the type of inflation targets that have been adopted by some countries recently.” Second, since money growth and real output growth are uncorrelated in the full sample, “monetary policy has no long-run effects on real output,” though “this does not rule out the possibility that it might have short-run effects,” on which “this study…is silent”; the authors add that if the long-run effect of money on output really is zero, “any short-run successes in reducing downturns can only come about at the expense of reducing upturns” (p. 11).