Response to a Skeptic

Questions & answers

Q1. How does Prescott frame his response to Summers’s critique at the outset?

Prescott opens by locating Summers’s attack within normal scientific practice: “New findings in science are always subject to skepticism and challenge. This is an important part of the scientific process. Only if new results successfully withstand the attacks do they become part of accepted scientific wisdom” (p. 28). He frames Summers as attacking the specific finding “that business cycles are precisely what economic theory predicts given the best measures of people’s willingness and ability to substitute consumption and leisure, both between and within time periods” (p. 28), and argues that by “challenging the existing quality of measurement and not providing measurement inconsistent with existing theory, Summers has conceded the point that theory is ahead of business cycle measurement” (p. 28) — the title claim of Prescott’s companion paper in the same issue.

Q2. How does Prescott respond to Summers’s “Where are the prices?” objection?

Prescott calls the question “puzzling,” responding that the real business cycle mechanism is “competitive equilibrium,” which “ha[s] relative prices,” and that the relevant prices he and Kydland examine are the real wage (“the price of people’s time”) and the rental price of capital, whose behavior in their models “conforms with that observed” (p. 28). On the objection that some unemployed workers would gladly trade places with employed ones — which Summers’s price-free critique implicitly invokes — Prescott cites Rogerson (1984) as having proven wrong Solow’s (1986, p. S34) prediction that resolving this would require abandoning price-mediated market clearing: because of nonconvexities in the mapping from time allocated to market activity to labor services produced (e.g., commuting time), competitive equilibrium with employment lotteries accounts for both “the observation that the principal margin of adjustment in aggregate hours is the number of people employed rather than the number of hours worked per person” and for “so-called involuntary unemployment” (p. 28).

Q3. How does Prescott respond to Summers’s objection that there is no independent evidence of technology shocks?

Prescott reframes the empirical target: “our finding is not that infrequent large shocks produce fluctuations; it is, rather, that small shocks do, every period,” noting that “at least since Slutzky (1927), some stable low-order linear stochastic difference equations have been known to generate cycles” from small shocks rather than a few large ones, and that “the equilibrium allocation for the calibrated neoclassical growth model with persistent shocks to technology turns out to be just such a process” (p. 28-29). He separately clarifies the scope of his claims: he does not argue the Depression was the equilibrium response to technology shocks, nor that disruptions in the payment and credit system have no effect; his claim is only that “technology shocks account for more than half the fluctuations in the postwar period, with a best point estimate near 75 percent,” which “does not imply that public finance disturbances, random changes in the terms of trade, and shocks to the technology of exchange had no effect in that period” (p. 29).

Q4. How does Prescott defend the model’s real-interest-rate and time-discount parameterization against Summers’s challenge?

Prescott argues the relevant return for the neoclassical growth model “is not the return on T-bills. It is the return on tangible capital, such things as houses, factories, machines, inventories, automobiles, and roads,” and that this return, computed from the U.S. National Income and Product Accounts as capital income net of adjusted capital consumption allowance divided by the business-sector capital stock, is “approximately 4 percent, about the average real return for the model economies” for the postwar years (p. 29-30) — directly countering Summers’s claim that the relevant historical real rate was about 1 percent.

Q5. How does Prescott defend the leisure-share (time-allocation) parameter, and what role does Rogerson’s (1984) indivisible-labor theory play?

Prescott defends the leisure-share parameter using Ghez and Becker’s (1975) reported U.S. annual productive time endowments (5,096 hours for males, derived to 4,836 hours for females) and average male employment hours (about 2,000 hours), yielding an average working-age-population employment-time fraction “about 0.30,” which with commuting time added is “close to those for our models” of 0.30-0.31 (p. 30). He traces the history of this parameter across his and Kydland’s work: with simple time-additive preferences the implied productivity-hours relationship was initially inconsistent with the data, so Prescott “did not take seriously” theory’s prediction at that stage; adding a distributed lag of leisure (rationalized by Kydland 1983’s household-capital-stock argument, later supported by Hotz, Kydland, and Sedlacek 1985 and Eckstein and Wolpin 1986) brought “theory and measurement of the business cycle” to “roughly equal” (p. 30). He then argues an “important advance in aggregate theory has made moot the issue of whether Kydland’s and my assumed preferences for leisure are supported by micro measurement”: Rogerson’s (1984) finding that, given an important nonconvexity in the time-to-labor-services mapping, “the aggregate elasticity of labor supply to temporary changes in the real wage is large independent of individuals’ willingness to intertemporally substitute leisure” — reconciling the “disparate micro and macro labor findings” (small at the micro level per Ashenfelter 1984, large at the macro level per Eichenbaum, Hansen, and Singleton) once Hansen (1985) incorporated this nonconvexity into the growth model (p. 30-31).

Q6. How does Prescott respond to Summers’s argument that measured productivity is procyclical because of cyclical labor hoarding?

Prescott argues the Fay and Medoff (1985) survey Summers cites “actually has little if anything to say about cyclical movements,” because the survey asked plant managers how many extra blue-collar workers they had in their “recent downturn” rather than separately at the trough and peak quarters of the most recent business cycle — a distinction Prescott says is needed to establish how hoarding “varies over the cycle,” and he adds the underlying questions were “simply too ambiguous for most respondents to interpret them the same way” (p. 31). He further argues the theoretical premise is backwards: because workers are “less likely to be hired by another firm” in bad times, “a firm with low output may be less reluctant to lay off workers… in good times,” which “suggests that labor hoarding associated with firm-specific output variations should be procyclical,” implying labor’s quasi-fixed nature (Oi 1962) “makes measured productivity less, not more, cyclically volatile than productivity really is” (p. 31).

Q7. What does Prescott argue about measuring the labor input in “effective units,” and what does this imply?

Prescott argues standard aggregate-hours measures treat “an experienced MBA from one of the best business schools” the same as “a high school dropout,” despite their hours commanding very different market prices, which is “evidence that they are not the same commodity”; the correct aggregation, following the neoclassical growth model, is in “effective units of labor” weighted by relative productivity (p. 31-32). Citing Kydland’s (1984, p. 179) finding that a 1-percentage-point change in the aggregate unemployment rate for adult males is associated with a 1.24 percent change in equally weighted hours but only a 0.65 percent change in effective-unit hours — because the more-educated, higher-paid workers have much less employment variability — Prescott concludes this is “strong evidence that if the labor input were measured correctly, the measure of productivity would vary more” (p. 32), the opposite direction from Summers’s hoarding-based claim that measured productivity is artificially too procyclical.

Q8. How does Prescott characterize the underlying disagreement, and how does the paper conclude?

Prescott argues Summers “cannot be attacking the use of competitive theory and the neoclassical growth environment in general,” since Summers himself “uses this standard model to predict the effects of alternative tax policies on aggregate economic behavior,” and that Summers offers no criteria for when the model’s implications should be taken seriously and when not; Prescott’s “guess is that the reason for skepticism is not the methods used, but rather the unexpected nature of the findings” (p. 32). He closes by agreeing with Summers on a narrower point: “We agree that labor input is not that precisely measured, so neither is technological uncertainty… we agree that theory is ahead of business cycle measurement” (p. 32) — restating the thesis of his companion paper as the one point of genuine, acknowledged common ground.