Scott Baker, Aniket Baksy, Nicholas Bloom, Steven Davis, Jonathan Rodden22 December 2020
Elections are a potential source of economic uncertainty because they sometimes bring large shocks to policy and to the investment environment. This is especially so for elections that take place in a politically polarised context.
Election-related uncertainties may, or may not, resolve quickly after an election. Legal challenges, recounts, lingering uncertainties about party control of legislatures and parliaments, and questions about key appointments in a new administration can draw out the uncertainties for weeks and months. The ongoing uncertainty about whether Democrats or Republicans will achieve party control in the US Senate in 2021 is a case in point.
Rising polarisation and greater policy differences among political actors, parties, and coalitions create greater scope for elections to become important sources of economic uncertainty. In the US and several other democracies, voters have come to see party platforms as much further apart today than in the past and have grown more hostile in their evaluations of the parties they oppose (Iyengar et al. 2019, Boxell et al. 2020).
Partly because of these developments, policy regimes have become more dependent on which political groups control the government. Elections have therefore become more consequential for the direction of policy and a larger source of economic uncertainty. Changes in policy directions – and uncertainty about those changes – affect investment, hiring, and spending decisions by firms and households.1
Recent examples abound of uncertainty-tinged and highly consequential elections. They include Australia in 2013 (Tony Abbott victory), India in 2014 (Narendra Modi), the US in 2016 (Donald Trump) and 2020 (Joe Biden), Brazil in 2018 (Jair Bolsonaro), and the UK in 2019 (Boris Johnson). In each of these elections, competing candidates offered starkly different policy proposals, and leadership changes led to marked shifts in economic policies (or is likely to do so in the case of Joe Biden). These election outcomes were genuine surprises or decided by slim margins.
Motivated by these observations, in a new paper (Baker et al. (2020) we study how national election cycles in 23 countries influence economic policy uncertainty, as measured by the newspaper-based indices of Baker et al. (2016). Average economic policy uncertainty values are 13% higher in the month of and the month before an election than in other months during the same national election cycle. This finding resonates strongly with evidence based on equity option markets in Kelly et al. (2016).
When we look more closely at the US, we find that the election-related rise in economic policy uncertainty is driven by elections that are close ex ante (as measured by poll-based predictions) and that take place among a polarised electorate.
Measuring economic policy uncertainty and political polarisation
We quantify economic policy uncertainty by the share of articles in leading newspapers that discusses matters related to uncertainty and economic policy. We obtain these indexes for 23 large economies, with data extending back to 1900 for the US and the UK and often into the 1980s or 1990s for other countries.2
To construct a database of national elections for each of our 23 countries, we draw on the Manifesto Project (Volkens et al. 2020) and the Constituency-Level Elections Archive (Kollman et al. 2019), adding more recent elections manually. We focus on elections that determine the national leader of the executive branch of government.
We classify an election as ‘close’ using polling data, which arguably better reflects electoral uncertainty in the run-up to an election than the actual results. We use data from Jennings and Wlezien (2018), who construct a daily measure of the average expected vote share for each party in an election. We average across all polls within a country-month to obtain a monthly expected vote share for each party. We classify an election as ‘close’ if the average absolute difference in expected vote shares over the three months preceding an election is smaller than 5%.
When studying American elections from 1952–2016 in more depth, we measure polarisation using individual survey data from the American National Election Study. We use a measure of affective polarisation, calculated as the (demographic-weighted) mean difference in affect felt by each respondent toward either party.3 Figure 1 plots time series for US economic policy uncertainty and political polarisation, showing a strong increasing trend in both.
Figure 1 Polarisation in US presidential elections and economic policy uncertainty
Notes: Time series of the level of policy uncertainty and a measure of affective polarisation in the US for elections, 1952–2016. Policy uncertainty (the left axis) is measured using the level of the economic policy uncertainty normalised so that the mean level between 1985 and 2009 is 100, and then averaged across months within each year. For 2020, we only include data until February. Affective polarisation is measured by the mean absolute difference in affect between parties averaged over all respondents and weighted by demographic weights (right axis). Ideological polarisation is an analogous measure, calculated as the mean absolute difference in ideological positions on a Liberal – Conservative scale, between parties averaged over all respondents and weighted by demographic weights (right axis). Policy uncertainty data from policyuncertainty.com, and our polarisation data is created using the American National Election Study.
Economic policy uncertainty around national elections
Our economic policy uncertainty measures are correlated over time and across countries, but they exhibit substantial within-country time-series variation. Different election schedules and cycles may drive some of this country-specific variation in national economic policy uncertainty.
We fit distributed lag regressions that relate monthly national economic-policy-uncertainty data to the timing of national elections while controlling for common global shocks, country fixed effects, and country-specific trends.
Figure 2 shows our estimates for coefficients on the distributed lag model. Our baseline specification shows that, on average, economic policy uncertainty rises by 13.2% in the month of a national election relative to economic policy uncertainty for the same country and election cycle outside a 10-month window around the election itself. Our results are robust to restricting the data to a balanced sample of countries and to excluding any single country.
Figure 2 Economic policy uncertainty and elections
Notes: Coefficients on dummies for ten periods prior to and after an election in our baseline distributed lag specification. The solid line reflects the behaviour of log(economic policy uncertainty) in a period surrounding an election relative to periods outside a 15-month window of an election. An observation is a country-month-year, and the dependent variable is log(economic policy uncertainty). The regression is run on the global sample of countries listed in Table 1, Baker et al. 2020. The shaded region depicts the 95% confidence intervals implied by standard errors clustered at the country level.
In the US, we see somewhat larger average effects of national elections, with policy uncertainty rising by 18.3% on average in November of presidential election years conditional on our controls. Additional investigation reveals that this effect varies greatly across elections with the closeness of polls in the run-up to the election and the extent of polarisation in the electorate.
Elections that are not close – in the sense that pre-election polls strongly favour one candidate – do not produce much economic policy uncertainty. Moreover, when political polarisation is low, even close elections produce modest uncertainty. In contrast, elections that are both close and polarised see an increase in economic policy uncertainty of roughly double the average election-induced impact in the US.
Policy uncertainty and political polarisation have trended upward for decades in the US (Baker et al. 2014). The Republican and Democratic parties are viewed by voters and investors as offering increasingly divergent platforms. Partly as a result, the policy stakes around US national elections have increased, and elections have become bigger sources of economic uncertainty. Although the particulars differ, many other democratic societies have also experienced political upheavals and rising political polarisation in recent years.
Some degree of election-related uncertainty is unavoidable, even desirable, in a democratic polity. Still, our findings suggest at least two reasons for concern. First, in the context of highly polarised electorates and divergent policy preferences, there are more possibilities for elections to create economic uncertainty. In this regard, the separation of governmental powers and checks and balances in a constitutional system requires broad agreement to achieve major policy changes (e.g. Persson et al. 1997). By slowing the pace of policy change and reducing the scope for large back-and-forth policy reversals, these institutional features moderate election-related uncertainties in a polarised environment.
Second, while some election-related uncertainty is inevitable, weaknesses in and doubts about the integrity of the election process can greatly magnify the uncertainties created by close elections. The US presidential elections in 2000 (Bush vs Gore), 2016 (Clinton vs Trump) and 2020 (Biden vs Trump) all raised concerns about the robustness of the election process and, among some, the legitimacy of the outcomes. It is easy to imagine how a closer electoral college outcome in 2020 could have produced a crisis of legitimacy and tremendous uncertainties, with negative consequences for the US and global economies. In this light, we see recent experience as a wake-up call about the need to strengthen the robustness and integrity of the election process.4
Baker, S R, N Bloom and S Davis (2016), “Measuring economic policy uncertainty”, The Quarterly Journal of Economics 131(4): 1593–636.
Baker, S R, N Bloom, B Canes-Wrone, S J Davis and J A Rodden (2014), “Why has US policy uncertainty risen since 1960?”, American Economic Review Papers and Proceedings 104(5): 56–60.
Baker, S R, A Baksy, N Bloom, S J Davis and J A Rodden (2020), “Elections, political polarization, and economic uncertainty”, NBER Working Paper 27961.
Bipartisan Policy Center, Commission on Political Reform (2014), “Governing in a polarized America: A bipartisan blueprint to strengthen our democracy”.
Boxell, L, M Gentzkow and Jesse M Shapiro (2020), “Cross-country trends in affective polarization”, Stanford J Institute for Economic Policy Research, working paper.
Canes-Wrone, B, and J K Park (2012a), “Electoral business cycles in OECD countries”, American Political Science Review 106(1): 103–22.
Canes-Wrone, B, and J K Park (2012b), “Elections, uncertainty and irreversible investment”, British Journal of Political Science 44(1): 83–106.
Hasen, R (2020), Election meltdown: Dirty tricks, distrust, and the threat to American democracy, New Haven CT: Yale University Press.
Iyengar, S, Y Lellkes, M Levendusky, N Malhotra and S Westwood (2019), “The origins and consequences of affective polarization in the US”, Annual Review of Political Science 22(7): 1–18.
Jennings, W, and C Wlezien (2018), “Election polling errors across time and space”, Nature Human Behavior 2: 276–83.
Kelly, B, L Pastor and P Veronesi (2016). “The price of political uncertainty: Theory and evidence from the option market,” Journal of Finance 71(5): 2417–80.
Kollman, K, A Hicken, D Caramani, D Backer and D Lublin (2019), “Constituency-level elections archive [data file and codebook]”, Ann Arbor MI: Center for Political Studies, University of Michigan, 17 June.
Persson, T, G Roland and G Tabellini (1997), “Separation of powers and political accountability”, Quarterly Journal of Economics 112(4): 1163–202.
Presidential Commission on Election Administration (2014), “The American voting experience: Report and recommendations of the Presidential Commission on Election Administration”.
Volkens, A, T Burst, W Krause, P Lehmann, M Theres, N Merz, S Regel, B Weßel, and L Zehnter (2020), “The Manifesto Data Collection. Manifesto Project (MRG / CMP / MARPOR)”, Berlin: Berlin Science Center for Social Research (WZB).
1 For instance, see Canes-Wrone and Park (2012a, 2012b) for evidence that firms and households delay relatively irreversible investments in the face of upcoming elections.
3 This polarisation measure is closely correlated with other measures, including a measure of the average perceived distance between the parties on a liberal – conservative scale. However, these alternative measures are available for a shorter time period.
4 For proposals to strengthen the election process and reform the electoral system, see Bipartisan Policy Center, Commission on Political Reform (2014), Presidential Commission on Election Administration (2014), and Hasen (2020).