Monetary and Fiscal Actions: A Test of Their Relative Importance in Economic Stabilization

The paper tests three specific propositions comparing the relative influence of fiscal versus monetary actions on total spending (GNP): (I) fiscal actions have a larger effect, (II) a more predictable (reliable) effect, and (III) a faster effect — each operationalized through a common reduced-form regression of quarter-to-quarter changes in GNP on quarter-to-quarter changes in the money stock (M) or the monetary base (B) and changes in high-employment government expenditures (E) and receipts (R), using an Almon-lag distributed-lag structure over four quarters. The authors are explicit that the exercise does not test the causal theories (Keynesian, portfolio, or modern quantity theory) underlying these propositions, only whether the empirically implied relationships hold using frequently-used indicators of fiscal and monetary action; data are seasonally adjusted quarterly observations from 1952:Q1 to 1968:Q2, with two alternative differencing conventions (first differences and Kareken-Solow “central differences”) used as a robustness check.

All estimated regression coefficients for changes in money or the monetary base have the theoretically expected positive sign and are statistically significant in most lag quarters, while the coefficients for high-employment expenditures and receipts frequently have the wrong sign or are statistically insignificant; the regressions explain 53-73% of the variance in GNP changes (R² of .53 to .73). The sum of the money/base coefficients across the four lag quarters is statistically significant in every specification (e.g., 5.74 to 6.59 for money, 16.01-16.41 for the base), while the summed expenditure coefficients are small and not significant, and the summed receipts coefficients never differ significantly from zero.

Because raw regression coefficients on money, the base, and fiscal variables are not directly comparable (different units, stocks versus flows), the authors convert them to standardized “beta coefficients” and partial coefficients of determination; Proposition I implies fiscal beta coefficients and partial R² should exceed those for money/the base, but the opposite is found — beta coefficients for money exceed those for expenditures in the contemporaneous and following two quarters, and the summed partial coefficients of determination for money and the base (.38 to .53) dwarf those for expenditures, which are “virtually zero.” A separate simulation illustrates the same point in dollar terms: a $1 billion permanent increase in government spending financed by borrowing or taxes raises GNP by only $170 million after four quarters, versus $5.8 billion for an equal-sized permanent increase in the money stock with no change in the budget position — leading the authors to conclude Proposition I “is not confirmed by the evidence.”

In a scenario suggested by Milton Friedman, government spending is raised by $1 billion for four quarters and then returned to its original level, financed entirely by a $250 million increase in the money stock each quarter; the simulation finds GNP rises to a permanent level $5.8 billion higher than its starting point, a result the authors state “results entirely from monetary expansion,” not from the fiscal action itself. This is presented as a clean illustration that the estimated fiscal-spending coefficient is small on its own — the large GNP effect observed when spending is money-financed is attributed by the authors’ framework to the accompanying monetary expansion, not to the spending increase per se.

Proposition II is tested via the t-values of the regression coefficients (the ratio of a coefficient to its standard error, indicating estimation reliability); the authors report that money and monetary-base t-values exceed those for high-employment expenditures in every lag quarter except one, and the summed t-values for money and the base are large while the summed t-value for expenditures is not statistically different from zero, so “the proposition is not confirmed.”

Proposition III implies fiscal beta coefficients should exceed monetary ones in the quarter of a change and the quarter immediately after, with fiscal’s main GNP impact occurring sooner; instead, money’s beta coefficients are large and roughly even across all four quarters, the base’s peak effect occurs in the first two quarters after a change, and the largest expenditure beta coefficient does not appear until the third quarter after a change — the opposite ordering from what Proposition III predicts, so “the expected regression results implied by Proposition III were not found.”

Because none of the three original propositions are confirmed, the authors state the results are instead consistent with the reverse propositions — that monetary actions have a larger, more predictable, and faster effect on economic activity than fiscal actions — while explicitly cautioning, per their own discussion of scientific hypothesis testing, that “confirmed” does not mean “proven true,” only that the alternative propositions remain acceptable working propositions “until evidence is presented proving one or more of them false.” They further qualify that the results are estimated over 1952:Q1-1968:Q2 and implicitly assume the same general economic environment continues to hold for near-term policy use, and report a Chow test across two roughly equal subsamples finding no evidence of a structural break in the estimated relationships.

The authors argue their results imply “the advisability of greater reliance being placed on monetary actions than on fiscal actions” — a “marked departure from most present procedures” — and that monetary authorities should not wait for fiscal actions to be adopted given monetary policy’s comparatively fast, strong, and reliable effects; they also argue the money stock is a useful summary indicator of the overall stance of stabilization policy because it reflects both discretionary Federal Reserve actions and the joint Treasury-Federal Reserve financing of new government debt. Using their estimated equation, they project GNP growth from 1968:Q3 through 1969:Q4 under four alternative constant money-growth-rate assumptions (2%, 4%, 6%, 8% annually); the fastest money-growth path implies continued rapid GNP expansion while the slowest implies a “noninflationary growth rate” reached around 1969:Q3, conditional on assumed government-spending growth and explicitly excluding the influence of “the Vietnam war, strikes, agricultural situations, civil disorders, or any of the many other noncontrollable exogenous forces.”

The appendix formalizes the regression as a reduced-form equation, ΔY = α₁ΔE + α₂ΔR + α₃ΔM + α₄ΔZ (Z summarizing all other, unmeasured influences on spending), and clarifies that each estimated coefficient (e.g., α₁ on expenditures) embodies both a direct effect and any indirect effect operating through Z, so the results cannot cleanly separate direct from indirect channels of fiscal and monetary influence. As indirect evidence that the equation’s large, statistically significant constant term is not itself driven by fiscal or monetary variables working through Z, the authors report that the constant term is stable across two roughly equal subsamples (via the Chow test) even though the relative variability of the fiscal and monetary variables differed between the subsamples — including a subsample (1953:Q1-1960:Q1) in which money and base variability were smaller than expenditure and receipts variability, yet the money/base response remained larger, which the authors read as strengthening rather than weakening their main conclusions.