The Welfare Costs of Moderate Inflations

Questions & answers

Q1. How does this model differ from the authors’ 1989 paper, and why do those differences matter for studying a transition to zero inflation?

The model adds two features absent from Cooley and Hansen (1989): distorting taxes on labor income (τh) and capital income (τk), with all revenue – from seigniorage and these taxes together – rebated to households as lump-sum transfers, and the economy has no uncertainty (agents have perfect foresight) rather than being driven by stochastic technology shocks. These additions let the paper study “the potential interaction of the inflation tax with other distorting taxes” and the revenue consequences of inflationary policy, neither of which the earlier paper addressed; as in that paper, money enters via a cash-in-advance constraint applied only to a “cash good” (here, consumption), with investment and leisure as untaxed “credit goods” in the terminology of Lucas and Stokey (1983, 1987).

Q2. How is the key preference parameter α (the relative weight on cash versus credit goods) calibrated, and how robust is it?

The paper calibrates α – the share parameter on the cash good in the period utility function – two ways: first via a money-demand regression of consumption velocity on the three-month Treasury bill rate using 1970-86 data, which implies α≈0.88 (though the paper flags this regression’s sensitivity to sample period and a possible spurious trend); second via Federal Reserve survey data on how households actually paid for transactions in 1984 and 1986, which found that cash, checks, and money orders together (“cash goods” transactions) accounted for 84% of transactions, implying α=0.84 by the model’s mapping between that percentage and the preference parameter. Because the two independent calibration approaches “are close,” the paper adopts α=0.84 as its baseline but also reports results under a much lower α=0.5 (a robustness check suggested by discussants Benabou and Wright) to show how sensitive the quantitative results are to the weight placed on the cash good.

Q3. What does the paper find about how the presence of labor and capital income taxes changes the welfare cost and revenue potential of the inflation tax?

“The presence of other distorting taxes approximately doubles the estimated welfare cost of the inflation tax and decreases the revenue potential of the tax”: for example, at the calibrated α=0.84, the welfare cost of a steady-state 10% annual inflation is 0.963% of GNP when labor and capital taxes are also present, versus only 0.375% of GNP when inflation is the only distortion in the economy (Table 1). The paper summarizes this pattern as the presence of other distorting taxes “nearly doubl[ing]” welfare costs while reducing inflation-tax revenue (relative to GNP) by roughly one-third compared with the no-other-taxes case, because the capital and labor taxes have already pushed the economy’s allocation away from its efficient point before the inflation tax is layered on.

Q4. What are the welfare and revenue consequences of permanently reducing moderate inflation to zero, without replacing the lost revenue?

Permanently eliminating a 5%, 10%, or 20% annual inflation rate (with labor and capital taxes held at their baseline rates, and the resulting revenue shortfall absorbed through lower lump-sum transfers) raises welfare by 0.34%, 0.66%, and 1.28% of GNP respectively (Table 2) – described in the paper’s own summary of results as “increases welfare by only 0.33 percent of GNP” for the 5%-to-zero case – while costing annuity-value revenue losses of 2.39%, 4.45%, and 7.87% of the relevant base respectively. The paper characterizes these as “very sizeable revenue losses obtained for a small improvement in welfare,” noting that “in the current budgetary climate even 2 percent would be important.”

Q5. What happens if the government instead has to replace the lost inflation-tax revenue with higher labor or capital taxes, and why does capital taxation fail for larger revenue losses?

When the revenue lost from ending inflation must be replaced by raising the labor income tax, welfare costs relative to the original inflationary policy rise rather than fall – by 0.425% of GNP for the 5%-inflation case, 1.018% for 10%, and 2.422% for 20% (Table 3) – and replacing the revenue with the capital income tax is worse still, raising welfare cost by 3.231% of GNP in the 5%-inflation case, while “replacing larger revenue losses with the capital tax is simply not possible” at all, because the capital tax base shrinks too much as the tax rate rises to sustain the needed revenue. The paper reads this as showing that labor and capital income taxes are themselves more distorting, at the margin, than a moderate inflation tax, so “replacing the inflation tax with taxes that are more distorting” leaves the economy strictly worse off than tolerating the original moderate inflation.

Q6. Why does a temporary (rather than permanent) reduction in inflation make the economy worse off, according to the model?

When a reduction in inflation is known by agents to be only temporary (lasting one, two, or five years before reverting to the original 10% inflation), welfare and the present value of revenue both decline relative to maintaining the original policy, because households respond to the temporarily untaxed “credit good” (investment) by sharply increasing investment during the tax holiday and correspondingly cutting consumption, then substituting back once the inflation tax returns – this “intertemporal substitution” leaves consumption permanently lower in the long run and the present value of revenue “marginally lower” as well. The paper’s conclusion from this exercise is that “unless there is a commitment technology available that would bind the government to policies that keep the inflation rate at zero,” temporary disinflation policies will not achieve their intended welfare-improving purpose and can backfire.

Q7. What does the paper conclude overall about the policy proposal to require the Federal Reserve to target zero inflation?

The paper’s bottom-line conclusion is that “the costs of lowering inflation to zero may well exceed the benefits” in this model economy, particularly when there is no credible commitment mechanism to sustain the policy indefinitely, a result the authors present while explicitly flagging that “this simple model abstracts from many of the details of the economy and the fiscal system that may have an important bearing on the consequences of policies to reduce inflation.” The paper separately notes, as a methodological benchmark, that the welfare cost of the pre-existing labor-and-capital-tax distortion alone (absent any inflation) is estimated at 16.843% of GNP, which it describes as “within the range of, but at the low end of, estimates that have appeared elsewhere in the literature” – context for judging how large the additional, inflation-specific distortions discussed above are relative to the tax system’s other distortions.