The Role of Automatic Stabilizers in the U.S. Business Cycle

The authors note that despite being “large” (the CBO estimated automatic stabilizers accounted for $386 billion of the $1089 billion U.S. deficit in 2012) and widely praised by policy institutions, “very little work has been done on automatic stabilization… in the last 20 years” (quoting Blanchard 2006), with most prior studies either using microsimulations or measuring the correlation between fiscal aggregates and the business cycle (Introduction, pp. 1-2). Their approach is explicitly different: “whereas this work focusses on measuring the presence of stabilizers, our goal is instead to judge their effect on the business cycle” (p. 4) – i.e., they build a structural model in which stabilizers can be counterfactually switched off to measure their causal effect on aggregate volatility, rather than inferring their strength from co-movement statistics.

The model is designed to include: “(i) intertemporal substitution, so that marginal incentives matter, (ii) nominal rigidities, so that aggregate demand plays a role in fluctuations, (iii) liquidity constraints and unemployment, so that Ricardian equivalence does not hold… and (iv) incomplete insurance markets and precautionary savings, so that social insurance affects the response to aggregate shocks” (Sec. 2, p. 5). Methodologically, “we merge the standard incomplete-markets model surveyed in Heathcote, Storesletten, and Violante (2009) with the standard sticky-price model of business cycles in Woodford (2003),” building on Reiter (2009) to numerically solve for the ergodic distribution of aggregates when the wealth distribution is itself a state variable and prices are sticky (p. 4) – a methodological contribution the authors flag as useful beyond this specific application.

The stabilization coefficient is S = V’/V − 1, where V is the ergodic variance of an aggregate at the calibrated parameters and V’ is its variance in a counterfactual with some stabilizers removed (Sec. 4, p. 29). The authors prove (Proposition 2) that in a degenerate representative-agent economy with flexible prices, proportional taxes, constant employment probability, fixed capital, and no fixed costs, S is exactly zero even though the traditional Pechman (1973) “built-in flexibility” measure – the ratio of tax changes to before-tax income changes – would register a large stabilizer in this same economy. This formally demonstrates that “the estimates of the size of the stabilizer following the Pechman… approach would be large in this economy” even where the stabilizer has literally zero effect on output volatility (p. 29), motivating why the paper measures actual variance reduction rather than built-in flexibility.

Cutting proportional tax rates by 10% “lowers the variance of the business cycle by a negligible amount,” and “removing the stabilizer[s], actually leads to a slightly more stable economy” in the full model (Sec. 4.1, Table VI, p. 30). The intuition is that “because the tax rate is the same in good and bad times, it does not induce any intertemporal substitution of hours worked, nor does it change the share of disposable income available in booms versus recessions” (p. 31); cutting the intercept (level) of the personal income tax produces a similar result (Table VII), with the additional finding that lower tax levels are associated with “slightly less volatile business cycles,” a mechanism the paper traces to progressive taxation’s interaction with monetary policy (Q6).

Reducing unemployment and poverty benefits by 80% (0.6% of GDP) “raises output volatility by 6% and raises the variance of hours worked by as much as 9%” in the full model (Sec. 4.2, Table VIII, p. 32). The mechanism is not the disposable-income channel: “transfers have no direct effect on the labor supply of recipients as they do not have a job in the first place,” but they are “funded by higher taxes on the patient households, who raise their hours worked in response to the reduction in their wealth,” which stabilizes aggregate hours (redistribution channel); separately, transfers reduce the precautionary savings households would otherwise accumulate against unemployment risk, so removing them raises volatility further via the social-insurance channel. The paper confirms this decomposition with the representative-agent and hand-to-mouth special cases: in the representative-agent economy, where both channels are shut off, “the transfer experiment has a negligible effect on all variables” (p. 32).

Table IX shows progressive taxes have “a modest effect on the volatility of output or hours,” in contrast to transfers, even though both are redistributive (Sec. 4.3, p. 34). The paper attributes this to two facts: first, “the level of progressivity in the current U.S. tax system is modest in the sense that the marginal tax rate function is relatively flat above median income,” so marginal rates barely change between booms and recessions; second, “the redistribution channel is significantly weaker than with transfers, because it is less targeted” – income-tax redistribution mostly moves resources among already-employed households whose labor-supply responses partly offset the stabilizing effect, unlike transfers to non-working needy households. The paper also documents an interaction with monetary policy: progressive taxation lowers and dampens the after-tax real interest rate’s response to inflation, which “lead[s] to lower real rates after positive monetary policy shocks, and thus more volatile responses of output and hours” (p. 35).

Combining all the experiments (flattening the income tax, cutting proportional taxes 10%, cutting transfers 80%), Table X shows “the stabilizers have had a marginal effect on the volatility of the U.S. business cycle in output or hours,” while “removing the stabilizers would significantly raise the variance of aggregate consumption” because government purchases would no longer be as countercyclical (Sec. 4.4, p. 35). In fact, “the current mix of stabilizers actually increase[s] the volatility of aggregate output and hours” in the model. Section 4.5 shows this depends heavily on how deficits are financed: if distortionary tax rates (rather than lump-sum taxes or government purchases) adjust to pay down debt after a recession, “raising tax rates in a recession to pay for the debt contracts economic activity,” making the stabilizers destabilizing under that financing rule, while removing government purchases’ cyclicality (holding spending fixed) eliminates the stabilizers’ effect on consumption volatility entirely (Table XI, pp. 36-37).

The paper’s key insight (Sec. 5.1) is that with flexible prices or a near-optimal monetary rule (from Schmitt-Grohé and Uribe 2007), the impact of removing all stabilizers “is very similar to the flexible-price case,” confirming that “with an effective monetary policy, there is little room left for the automatic stabilizers to work through aggregate demand” (Table XII, p. 38). Only under a distinctly accommodative Taylor rule (weak response to inflation) do the stabilizers substantially reduce output and hours volatility, because “demand shocks will lead to larger fluctuations in activity as they are not offset by monetary policy” (p. 39). This establishes that fiscal stabilizers and monetary policy are largely substitutes for demand management in this model – fiscal stabilizers matter most exactly when monetary policy is unable or unwilling to do the job itself.

Simulating a combined preference, unemployment-risk, and job-finding shock calibrated to produce a 4% cumulative employment decline (matching the U.S. Great Recession), the paper finds that without stabilizers “the impact of the shock is now substantially larger”: the ZLB binds for an additional period, and consumption and output fall by an additional 2 and 1 percentage points respectively, while impatient households’ consumption falls by an additional 5 percentage points, for a total drop of nearly 15% (Sec. 5.2, p. 40). This confirms that “the automatic stabilizers are more effective in mitigating the extent of the contraction during a zero lower bound episode,” consistent with the broader ZLB fiscal-policy literature (e.g. Woodford 2011, Christiano, Eichenbaum and Rebelo 2011) that the paper cites.

Reducing all stabilizers makes every impatient household worse off (an equally-weighted average welfare loss of 0.151 in consumption-equivalent units) while raising patient households’ welfare by 0.136, reflecting the stabilizers’ “large redistributive component” (Sec. 6.1, Table XIII, p. 42). A utilitarian social welfare function that weights groups by population share shows cutting the stabilizers lowers average welfare by 0.080 consumption-equivalent units. Crucially, when the authors isolate the welfare cost specifically attributable to business-cycle fluctuations (by comparing the cost of individual recessions with and without stabilizers, Table XIV), they find business-cycle costs are small in either case (consistent with Lucas 1987) except at the ZLB – so “the large welfare benefits of the stabilizers are due to redistribution and social insurance, not business-cycle fluctuations” (Sec. 6.2, p. 44).

The conclusion frames the paper as adjudicating a long debate – Milton Friedman (1948) championed automatic stabilizers over discretionary policy, while Solow (2005) argued for more research focus on them – and states: “we found that reducing the scope of all the stabilizers would have had little impact on the volatility of the U.S. business cycle in the last decades. This depends on monetary policy having responded aggressively to inflation and being close to optimal” (Sec. 7, p. 45). The authors flag that their model treats employment transitions as exogenous, ignoring the disincentive effects of transfers on job search effort or reservation wages (citing Young 2004 and Hagedorn, Karahan, Manovskii and Mitman 2013); they report that in an extension with endogenous search effort, “stabilizers were even less effective, and more consistently destabilizing” (p. 45), suggesting their baseline estimates of the stabilizers’ net benefit, where positive, may if anything be generous once labor-supply disincentives are fully modeled.