![Sample our Social Sciences journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/110822/TOC-Subject-Sample-2021-Social-Sciences.jpg"})
ABSTRACT
Donald Trump was the President of the United States from January 2017 to January 2021. During that time, except for the period since spring 2020 when the COVID-19 pandemic took its toll on economic activity, the US economy has been doing very well according to key indicators like the unemployment rate and GDP growth. Does Trump deserve credit for the booming economy? To address this question, we develop a counterfactual scenario for how the US economy would have evolved without Trump – we let a matching algorithm determine which combination of other economies best resembles the pre-election path of the US economy. We then compare the performance of the US economy during Trump’s Presidency to this synthetic “doppelganger”. There is little evidence for a Trump effect.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 For instance, Trump (Citation2018) states: “We’ve got the greatest economy, maybe, ever – maybe in history. We have the greatest economy we’ve ever had … If we didn’t win, this economy would be a wreck.”
2 V is a weighting matrix assigning different relevance to the characteristics in x1 and X0. Although the matching approach is valid for any choice of V, it affects the weighted mean squared error of the estimator (see the discussion in Abadie, Diamond, and Hainmueller (Citation2010), 496). Following Abadie and Gardeazabal (Citation2003) and Abadie, Diamond, and Hainmueller (Citation2010), we choose a diagonal V matrix such that the mean squared prediction error of the outcome variable (and the covariates) is minimized for the pre-election period. Including the covariates in the optimization differs from Kaul et al. (Citation2018) who have raised concerns about including all pre-intervention outcomes together with covariates when using the SCM. The size of V is determined by the number of observations to be matched. These include I time series of length T and the K covariate averages.
3 In fact, the nature of the estimation procedure makes it difficult to combine trending variables (like real GDP) with cyclical indicators (like unemployment). This is because the minimization procedure is then dominated by the more volatile cyclical indicators. While in principle it is possible to define V, the weighting matrix in (1), such that it would assign less weight to unemployment relative to real GDP, this adjustment would necessarily be arbitrary.
4 In order to best match on “predictors” of growth, we take averages of covariates one year prior to the election, that is, 2015Q3 to 2016Q2.
5 As data for 2020Q4 is not fully available at the time of the OECD Economic Outlook publication in December 2020, we rely on nowcasts included in the same publication.
6 In the appendix we show results for an alternative procedure where we estimate four separate doppelgangers for each of the four variables. The benefit of this exercise is that we obtain a better fit pre treatment, since each doppelganger needs to match only T + K = 92 observations. Importantly, also this alternative procedure gives rise to a very similar picture.
7 The extent to which economic factors affected the election result remains debated (Mutz Citation2018). From a macroeconomic perspective, the overall situation at the time of the election appeared relatively benign in the sense that the US economy was not in a recession or in a phase of high unemployment.
8 The importance of oil price shocks in accounting for the business cycle, notably the relative importance of supply and demand shock remains controversial to date (Kilian and Zhou Citation2020). For the US, for instance, Kilian and Vigfusson (Citation2017) find that oil price shocks are less important for real GDP “than widely believed”.
9 Still economies are subject to structural change. For instance, as a far as energy is concerned the US turned from net importer to net exporter in 2019. Still, we believe that these structural changes are taking places at lower frequencies and are thus unlikely to drive our results in a quantitatively important way.
10 More formally, we follow Hahn and Shi (Citation2017) and test whether the post-election doppelganger gap and all the pre-election doppelganger gaps of the same length can be considered to come from the same distribution. We cannot reject this hypothesis.
Additional information
Notes on contributors
Benjamin Born
Benjamin Born is an Associate Professor at the Department of Economics at the Frankfurt School of Finance & Management and a CEPR research affiliate. His research interests are mainly in macroeconomics, with a focus on fiscal and monetary policy, uncertainty, and business cycles.
Gernot J. Müller
Gernot J. Müller is Professor of Economics at the University of Tübingen and a CEPR Research Fellow. His research focuses on monetary and fiscal policy, in particular in an open economy context.
Moritz Schularick
Moritz Schularick is Professor of Economics at the University of Bonn, an elected member of the Academy of Sciences of Berlin-Brandenburg, and a Research Fellow of the Centre for Economic Policy Research. His research interests include macroeconomics, international finance, and economic history. Working at the intersection of economic history and macroeconomics.
Petr Sedláček
Petr Sedláček is a Professor of Economics at the University of Oxford and a CEPR Research Affiliate. His research mainly deals with macroeconomic questions, with a focus on business cycles, firm and worker heterogeneity.