Accuracy Targets when Forecasts Influence Outcomes
[Slides] Abstract. In the business setting, economic forecasts often influence predicted outcomes. This poses a challenge to evaluating forecast accuracy, and makes it difficult to relate accuracy to concrete business value. Moreover, feedback loops in forecasting imply that the predictability of future outcomes is implicitly driven by company decision-making. We propose a simple stylized model to address these issues in the context of company Sales Quotas. We show that an effective quota mechanism implies that forecast error variance must be bounded away from zero, and that companies can increase the predictability of future earnings by reducing size concentration across customers, and by reducing correlation in performance across individual Sales Reps.
Risk in Network Economies
Winner of Walter Heller Memorial Prize
[PDF] Abstract. Economic models with input-output networks assume that firm or sector growth is driven by a combination of trade partners' growth and idiosyncratic shocks. This assumption generates unrealistic restrictions on network weights. Allowing for correlated shocks exposes units to additional risk that captures their ability to substitute away from supply and demand shocks. I provide evidence that substitutability between trade partners is related to technological and product dispersion that is not captured by standard firm and industry definitions. I propose a production-based asset pricing model in which supply chain substitutability depends on product/technology dispersion and shock correlation driven by shared suppliers and customers. The model predicts that assets positively exposed to upstream and downstream shocks are useful hedges and earn lower average risk premia than less exposed peers. This is confirmed by estimated return spreads of -11.4% and -4.2% and a negative association with aggregate growth.
Why Does GDP Move Before Government Spending? It’s all in the Measurement
with Edoardo Briganti, Reject & Resubmit at American Economic Review
[PDF, Online Appendix] Abstract. We find that the early impact of defense news shocks on GDP is due to a rise in business inventories, as contractors ramp up production for new defense contracts. These contracts do not affect government spending (G) until payment-on-delivery, which occurs 2-3 quarters later. Novel data on defense procurement obligations reveals that contract awards Granger-cause shocks to G identified via Cholesky decomposition, but not defense news shocks. We show that Cholesky shocks to G miss early changes in inventories, and thus result in lower multiplier estimates relative to the narrative method.
Deep Learning and Long-Run Risk
with Tjeerd de Vries
[Draft coming soon] Abstract. In dynamic asset pricing, stochastic discount factor (SDF) processes summarize the relationship between equilibrium asset prices and underlying economic conditions. SDFs can be factorized into permanent and transitory components, where the permanent component captures pricing at long payoff horizons. Hansen and Scheinkman (2009) show that the permanent-transitory decomposition can be cast as the unique solution to a Perron-Frobenius eigenvalue problem, for which analytic solutions are only available for a limited array of examples. Moreover, standard numerical approaches are not equipped to handle this general class of problems due to the curse of dimensionality and lack of well-developed boundary conditions or parametric restrictions. We develop a novel algorithm for solving this class of eigenvalue problems in a very general class of asset pricing models without boundary conditions or parametric assumptions on the eigenfunctions. We demonstrate the accuracy of the algorithm in the context of several workhorse structural asset-pricing models, and argue that our approach applies to models which feature a very general class of Levy processes in arbitrarily high dimensions.
Breaking Down the U.S. Employment Multiplier Using Micro-Level Data
with Edoardo Briganti, Holt Dwyer, and Ricardo Duque Gabriel
[PDF] Abstract. We use restricted data from the Quarterly Census of Employment and Wages (QCEW) to link the universe of U.S. establishments with the universe of contractors in the Federal Procurement Data System (FPDS). Leveraging detailed institutional knowledge of federal acquisitions, we construct a new set of unanticipated contracts and examine their effects on employment growth. We find positive, significant, and persistent effects on firms with fewer than 150 employees. Using loan data from the Federal Reserve (Y14-Q), we show that small firms expand their credit and face lower interest rates after winning unanticipated contracts. At the regional level, we estimate a cost-per-job of $57,000 per year using unanticipated contracts—an order of magnitude lower than previous estimates based on all defense contracts. Lastly, by leveraging restricted QCEW data, we decompose the employment multiplier into a direct effect on contractors and an indirect multiplier effect on non-contractors, finding a 55-45% split, respectively.
Tax Multipliers in the US: a Regional Perspective
with Edoardo Briganti and Carlos Goes
[Slides, Draft coming soon] Abstract. Policy-induced tax changes have very different incidence profiles. For instance, the tax reforms pursued by President Bush (2002) and President Obama (2013) were very different: the former was a large tax cut for both low-income and high-income tax units, while the latter was a small tax cut for the low-income households and a large tax hike for high-income earners. Presumably, such events could have very different macroeconomic implications. Individuals at different points of the wealth and income distributions have different marginal propensities to consume, which could substantially impact fiscal multipliers. Standard macroeconometric only allows us to estimate average multipliers over time. In this project, we estimate the event-specific causal effect of different federal personal income tax reforms on local economic activity, exploiting event-county-level variation in tax incidence induced by all the major federal personal income tax reforms in the United States since 2000. We use adjusted gross income brackets at the county level to produce a novel data set of local taxable income distributions. We then combine the county fiscal income distributions and tax policy variation at the federal level to construct county tax shocks, using a shift-share approach to estimate local multipliers of local tax liability on different measures of local economic activity, in particular employment, consumption, and retail GDP.
