JOB MARKET PAPER
"The Macroeconomic Effects of Tax Policy Surprises" PDF SSRN
This paper exploits newly digitized Federal Reserve staff forecasts of tax receipts from the Greenbook to construct an instrumental variable for tax policy surprise shocks. The instrument isolates tax rate forecast revisions around narratively identified legislated tax changes. I find that expansionary tax policy shocks have permanent positive effects on real GDP and personal consumption, and a larger but more transitory, hump-shaped effect on total investment. Although the shock acts as a textbook aggregate demand shock that tightens the labor market, it triggers a temporary deflation: firms meet higher demand by using existing workers and installed capital more intensively, which lowers unit production costs. These results suggest that tax-based stimulus can boost output without adding to short-run inflationary pressure. Unlike pure narrative approaches (Romer and Romer, 2009; Mertens and Ravn, 2012), which rely on the quarterly dating of legislated tax changes, the near-monthly frequency of Greenbook forecasts provides finer control for tax foresight; moreover, the new instrument preserves its statistical power beyond the 1940s and allows for a much more efficient estimation of the macroeconomic effects of tax policy shocks.
"Estimating the Macroeconomic Effects of Government Spending Shocks Using Survey of Professional Forecasters" with V. Palagi Working Paper
Proxying agents expectations on future fiscal changes by Survey of Professional Forecasters (SPF) predictions is a common identification strategy for studying the causal effects of government spending shocks and controlling for fiscal foresight (Ramey (2011) and Auerbach and Gorodnichenko (2012b)). This paper provides theoretical and empirical evidences that SPF raw forecast error of federal government spending mixes multiple macroeconomic shocks and is predictable by past economic conditions. We orthogonalize the raw forecast error with respect to current beliefs about future economic conditions and to other macroeconomic structural shocks, in the spirit of Miranda-Agrippino et al. (2026), and use the resulting instrument to identify government spending shocks in a large BVAR-IV. The cleaned instrument substantially changes the composition of the response implied by the raw forecast error: an increase in government spending crowds in non-residential investment and non-durable goods consumption, while crowding out residential investment and durable goods consumption. Once aggregated into GDP, these opposite movements partly offset each other, so the differences across instruments are less visible at the multiplier level: the impact multiplier is around 2, only slightly above the one obtained from the raw forecast error, while the 5-year cumulative discounted multiplier converges to 0 under the novel instrument, against 1.5 under the raw forecast error specification.
It depends. According to conventional wisdom a good-quality measure of potential output should react to supply shocks and its response gradually converge to the path of actual output. In contrast, policy shocks shouldn’t induce any variations on potential, as affecting actual GDP only transitory. Unlike, this paper shows that potential’s dynamic responses differ greatly according to the estimation method. Some popular approaches, namely Hodrick and Prescott (1997) and the quadratic trend, don’t react to any kind of shocks as estimate a low-variability trend. Many others respond similarly to every shocks (16 quarters moving averages, Blanchard and Quah (1988), Hamilton (2018), and the Beveridge and Nelson (1981) decomposition) and their IRFs can hardly be considered consistent with those simulated by New Keynesian models. I found that when potential output is estimated by a Baxter and King (1999) band-pass filter extracting frequencies between 3 and 28 quarters, its responses are in line with my testable assumption. Such results are robust across economic shocks identification schemes.
Although its poor end-of-sample performance and the resulting real-time unreliability, Hodrick and Prescott (1997) filter is the most popular detrending method. Here I exploit forecasts by Survey of Professional Forecasters (SPF) to construct a real-time reliable output gap estimate based on a forecast augmented HP filter. I show that the new filter has a sharp better forecasting properties for output growth, unemployment rate and inflation by a standard Phillips curve relation. It is much more correlated with Policy Institutions output gaps and also slightly outperforms one-sided Hamilton (2018) filtered output gaps in these comparisons. An analytical formulation of the forecast augmented HP filter is provided. The good real-time properties of forecast augmentation are extended also to Butterworth et al. (1930), Christiano and Fitzgerald (2003) band-pass filters and Canada’s country case.
DORMANT PROJECTS
"Not all Oil Price shocks are alike: Testing alternative Narrative Instruments"