Hester claims bias

In 1964, Donald D. Hester criticized^[7] the F/M paper for "bias" against a "Keynesian" outcome. For that purpose, Hester argued that government deficits are endogenously determined, and not exogenously, and thus no single-equation approach could properly capture government spending and deficits, while the same principle applies for short-run private investment. Also, Hester emphasized that the actual data should have been empirically tested in first-differential form so as to extricate the trends of both explanatory variables, and thus demonstrate only the endogenously generated economic growth. Hester stated that, when he tried "improved" data and empirical methods, “the autonomous expenditure theory outperformed the quantity theory [of money],” i.e. Keynesian economics win over monetarist economics.^[7]

Friedman/Meiselman respond

In a paper published in 1964,^[8] Friedman and Meiselman conceded that Hester’s suggestion of using first differences was correct and that it is a better method for their single-equation approach. But they insisted that their interpretations of income and autonomous expenditures are relevant, rejecting Hester’s misgivings. They claimed that Hester’s use of correlation coefficients with his newly defined autonomous expenditures constituted an "unsound argument,"and summarized as follows:

We remain of the opinion that there is a striking division among students of economic affairs about the role of money in determining the course of economic events. One view is that the quantity of money matters little; the other, that it is a key factor in understanding, and even more, controlling economic change. Our paper tried to present some evidence relevant to deciding between these views. The kind of evidence we gave is not the only kind that is relevant and may not be the most important or significant. And, of course, much other evidence is available from other work by us and by many others. This other evidence needs to be added to and brought to bear on the main issue that divides economists into two groups. Hester does not quarrel with the relevance of our evidence but with the particular form of the income-expenditure theory we use. [Hester's] criticism of our procedure rests primarily on a misunderstanding of the theoretical basis of our approach. He offers neither theoretical argument nor empirical evidence in support of his alternative formulation. Hence his criticism is largely beside the point. That is unfortunate. We badly need work on these problems that will clarify the issues involved. We can ill afford to waste the energy, interest, and ability that Hester displays in his paper on frivolous quibbling.^[8]

Ando and Modigliani: both policies affect outcome

Albert Ando and Franco Modigliani, in a paper published in 1965,^[9] disputed the findings presented in the 1963 Friedman/Meiselman work. Ando and Modigliani claimed that

[The Friedman/Meiselman 1963 work] has shortcomings in procedures that if repaired change the result, but, moreover, the single-equation approach coupled with the equally single, independent variable approach and the corresponding correlations cannot shed light on macro-policy.^[9]

They argued that the consumption function^{[note 4]} was not correctly specified within the F/M use of autonomous expenditures and claimed that the variable that Friedman and Meiselman had derived was actually saving and not autonomous expenditures. They also observed that the data used in the 1963 paper would need to be modified by including corporate retained earnings, transfer payments made by the government to foreigners, and “wage accruals over disbursement.”^[9]

Ando and Modigliani objected to the use of an ordinary, least squares equation because of the induced influence on the independent variable by the dependent variable^{[note 5]} and offered their own model, which ostensibly removed the independent part from the induced part.

Ando and Modigliani criticized the Friedman/Meiselman paper for omitting to determine exogenous and endogenous components to monetary policy in the same manner as economists do with fiscal policy. Instead, Ando and Modigliani, rather than using a standard money-supply variable, introduced M*, which is meant to represent what the money stock would be if high powered money were "fully utilized", thus introducing a "high-usage variable."^{[note 6]} The purpose was to show that money is not exogenously determined: people can choose to hold money in different amounts and levels of liquidity as situations warrant, while lenders don't need to lend out all of their excess reserves if they so desire - which constitutes a standard Keynesian concept. Moreover, Ando and Modigliani found that the error variance in predicting output^{[note 7]} was "much higher" when using money than any of the fiscal variables and labeled the F & M respective results "spurious."

They concluded that Friedman and Meiselman’s results were biased in favor of monetary policy, and that, if both policy variables were to be given a balanced approach, the end result would be that both policies would have real and statistically significant effects on the economy. Indeed, in the opening statement of their paper, they state that the "number of basic shortcomings in [the Friedman/Meiselman] procedure...make the results of their elaborate battery of tests essentially worthless."^[9] Economists Michael DePrano and Thomas Mayer published^[10] a critique of the F & M paper that was generally in line with the criticisms leveled by A & M.^[11]

FM respond to AM

In 1965, Friedman and Mieselman responded^[11] to the criticism leveled at their 1963 paper, particularly by Ando and Modigliani. They claimed that although one could indeed object to their autonomous-expenditure variable, any of the alternatives that had been put forward by others were equally objectionable. Additionally, Friedman and Meiselman defended their use of their consumption function and explained why it is, in their view, the right method to use. And they pointed out that, unlike them, Ando and Modigliani used nominal data rather than real data and, therefore, the empirical results put forward by A & M could not be correctly comparable to their own.

Friedman and Mieselman agreed in theory with A & M that the term M* is a valid means to determine the exogenous versus the endogenous nature of the policy variable but still disagreed with the actual A & M methodology to determine M*. As to the consumption variance, they maintained that "of the total variance of consumption for the 25 years, 88 per cent is accounted for by the differences between the means for the two subperiods."^[11]

Finally, they claimed complete lack of bias in their research and the empirical processes and claimed that even if they had built a model that seemed to favor monetary policy over fiscal policy, that was because the theory comes out that way. They concluded as follows:

None of the calculations made by our critics for supposedly the same purpose is correct because they omit some components of income for the income-expenditure calculations, set the two theories different tasks, or use lengthy periods combining two different sub-periods. We have made some of the correct calculations for one of the alternative concepts of autonomous expenditures (Ando and Modigliani’s). Though less clear-cut, the results are in the same direction as those from our original calculations. Hence, we are left with no reason to change our earlier conclusion that “so far as these data go [and, we may now add, those adduced by Ando and Modigliani, DePrano and Mayer, and Hester]^{[note 8]} the widespread belief that the investment multiplier is stabler than the monetary velocity is an invalid generalization from the experience of three or four years. It holds for neither later nor earlier years”.^[11]

The St. Louis equation

In 1968, Federal Reserve Bank of St. Louis economists Leonall C. Andersen and Jerry L. Jordan published a study^[12] that fully supported the Friedman and Meiselman single-equation approach but expanded it in response to the criticisms of the 1963 paper. They offered their own economic output model, in which all variables are in first-differential form as denoted by ∆, as follows:

${\displaystyle \Delta Y_{t}=\alpha +\sum _{i=0}^{4}m_{i}\Delta M_{t-1}+\sum _{i=0}^{4}e_{i}\Delta E_{t-1}+\sum _{i=0}^{4}z_{i}\Delta Z_{t-1}}$ (2)

where α is a constant; Y is nominal domestic spending; M represents monetary policy defined by the monetary base; E represents variously high-employment expenditures, high-employment receipts, or high-employment surplus; and Z represents a catch-all variable defined as “a variable summarizing all other forces that influence total spending.” Those forces include weather, international trade, preferences, technology, resources, infrastructure, war, etc. In their math, they used an Almon lag technique with 4th-degree polynomials and a 4-period time lag.^{[note 9]} They concluded that, just as Friedman and Meiselman had found, monetary policy seemed to affect whatever measure was used for spending; but fiscal policy did not.^[12]

De Leeuw and Kalchbrenner: endogenous vs. exogenous

In 1969, Frank DeLeeuw and J. Kalchbrenner, also St. Louis Fed economists, published an article^[13] that criticized "severely" the Andersen/Jordan study and modeling. They argued that exogenous fiscal policy cannot be properly measured by using any of the fiscal policy definitions presented by their colleagues, nor can any single-equation approach bring into relief the particular influences of such a policy variable. There exist no ways, they claimed, to separate the endogenous- from the exogenous-policy effects because those effects are effectively lost in the complex workings of an entire economy. They pointed out, in particular, that the tax and monetary-base variables are impossibly entangled with the endogeneity-exogeneity problem and claimed that the Andersen/Jordan method leaves out the influences introduced by inflation.

The main argument by De Leeuw and Kalchbrenne was that causality cannot be demonstrated by the single-equation approach, and the direction of causation is impossible to establish, i.e. GNP could be driving is fiscal spending rather than the other way around. After making a "clear improvement"^[13] on the Andersen/Jordan model (using high employment receipts adjusted for inflation as the fiscal variable and two different versions of the monetary base), and re-running the "St. Louis equation" on the basis of data from 1952 to 1968, they proclaimed that, according to their findings, fiscal expenditures were statistically significant, positively correlated^{[note 10]} to changes in GNP in the long run - as was also true for changes in monetary policy.

Other arguments

In 1971, William L. Silber^[14] posited that the researchers were altering their equations to fit into whatever their ideological worldviews were theorizing, which was the reason he gave his paper a "highly political" title. He questioned the validity of the overall methodology behind the "St. Louis equation" approach, after running it in various time periods that were deemed to have the same underlying structural form, and finding that some periods appeared to show fiscal policy as quite significant while others did not.

In 1971, Edward Gramlich reviewed^[15] the “radio debate" up to that point and compared the multiplier and elasticity estimates for monetary and fiscal policies among the several different models and the non-single equation models. As he stated, all the models, except the Ando/Modigliani ones, showed monetary policy to have a multiplier above 1, and in every case to be larger than the fiscal multiplier. His study, which included a model "improving" the St Louis equation, supported the view that monetary policy is strongly correlated with spending but also found that fiscal policy is correlated as well.

In 1972, Stephen M. Goldfeld, Alan S. Blinder, John Kareken and William Poole, in their study,^[16] criticized the Andersen/Jordan approach as econometrically unsound. They argued that, without a reaction function, one cannot determine the nature of the “exogenous” from the "endogenous”, and that, if the rules or automatic stabilizers are done to counter-cyclical perfection, then the correlations do not show up with the statistical significance we would expect. Their paper concludes that the single-equation approach used to empirically determine the comparative efficiency of monetary and fiscal policies is "without merit."^[16]

In 1973, William Poole and Elinda B. F. Kornblith found^[17] that all the models tended to "underpredict," and attempted to provide hypotheses for that result. Their conclusion was that the “decision [about which models were correct or supported monetary or fiscal policies] must still be rated a draw.”^[17] In 1975, J. W. Elliot conducted an empirical analysis and pointed out the difficulty of comparing the regression coefficients as “multipliers” since their corresponding variables are money, which is a stock, and fiscal spending, which is a flow. He concluded that, irrespective of technique, his results supported the Andersen/Jordan results.

Ando and Modigliani return

In 1974, at a conference held at Brown University, Ando and Modigliani presented a paper where they recreated an analysis of a simulated economy using the Andersen/Jordan method, which, they concluded, was biased in favor of monetary policy. Their work was published in 1976.^[18]

Keith Carlson vs. Benjamin Friedman

In 1977, Benjamin Friedman found that, using the Andersen/Jordan model but extending the data set out to the 2nd quarter of 1976, fiscal policy was now statistically significant in the determination of expenditures -although serious heteroscedasticity problems had appeared. He also found that, if he used data starting at the 1st quarter of 1960, the results were even more favorable to discretionary fiscal policy, reiterating the existence in the Andersen/Jordan model of an inherent coefficient bias. Ultimately, he concluded that the St Louis equation methodology was "not salvageable."^[19]

In response, Keith Carlson, in 1978, made an empirical modification to the original Andersen/Jordan model, whereby instead of using a first-difference approach, he posited that a rate-of-change approach eliminated the heteroscedasticity problems discovered by B. Friedman. Carlson’s model is as follows:

${\displaystyle {\dot {Y}}_{t}=\alpha +\sum _{i=0}^{4}m_{i}{\dot {M}}_{t-1}+\sum _{i=0}^{4}e_{i}\ {\dot {E}}_{t-1}}$ (3)

where the variables are the same as in the Andersen/Jordan model but the dots over the terms denote growth rates for the respective variables. Carlson determined that his model once again supported the original conclusion of significant monetary-policy effects and insignificant fiscal-policy effects.