Fear thy neighbor: Spillovers from economic policy uncertainty

Because the theory and evidence are about how agents behave, not about the distribution itself – and the paper says so explicitly rather than letting the negative connotation pass unexamined (Section 1, p. 2). “While EPU has commonly a negative connotation, its sole meaning refers to uncertainty surrounding future policy, which does not necessarily imply a higher probability of implementing a policy that yields a worse outcome. Rather, a higher EPU just reflects that future shocks have a wider probability distribution (or even an unknown one), which includes shocks that would lead to better outcomes. Given these considerations, a higher EPU is not necessarily bad news, but the existing evidence points to the fact that economic agents dislike EPU and tend to react by re-weighting probabilities toward unfavorable events, in essence confounding certain for good and uncertain for bad and taking a ‘wait-and-see’ approach.” The rest of the paper’s contractionary results therefore rest on this behavioural premise, not on uncertainty being intrinsically adverse.

No – two families of theories predict contraction and two predict expansion, but the models built specifically around policy uncertainty all predict contraction (Section 1, pp. 2-3). On the contractionary side: the “real option” theory (Bernanke 1983; Brennan and Schwartz 1985; Dixit and Pindyck 1994; McDonald and Siegel 1986) implies that when irreversibility is high the option value of waiting rises with uncertainty, so agents are “reluctant to invest, hire, or purchase durable goods during periods of elevated uncertainty”; and the “risk aversion” theory (Arellano et al. 2010; Christiano et al. 2014; Gilchrist et al. 2014) implies higher risk premia and borrowing costs. On the expansionary side: the “growth option” theory (Paddock et al. 1988; Bar-Ilan and Strange 1996) has uncertainty raising the potential gains from a project, and the “Oi-Hartman-Abel” theory (Oi 1961; Hartman 1972; Abel 1983) has firms’ flexibility to vary production volumes under output-price uncertainty producing expansion. The narrower EPU literature, however, breaks the tie in one direction: Rodrik (1991), Hassett and Metcalf (1999), Born and Pfeifer (2014) and Fernandez-Villaverde et al. (2015) together imply “that EPU acts as a tax on macroeconomic aggregates, inducing adverse effects on economic activity” (p. 3).

Three things: decomposing the effect into spillovers and domestic effects, identifying the three largest origins, and mapping regional heterogeneity (Section 1, pp. 3-4). The authors place their work in “a decisively slimmer body of recent empirical studies that looks at whether uncertainty shocks spill over to other economies,” most of which focus on US-origin shocks: Mumtaz and Theodoridis (2015) on US GDP-volatility spillovers to the UK; Kamber et al. (2013) on a VIX-proxied US uncertainty shock across a six-economy world sample; Carriere-Swallow and Cespedes (2013) on VIX shocks and investment and consumption in 40 countries. The closest antecedents are IMF (2013), where US and European EPU shocks suppress activity in 43 economies, and Colombo (2016) on US EPU spillovers to the euro area. The contribution as stated: “First, we disentangle the impact of EPU on the growth rate of output, private consumption, and private investment into spillovers and domestic effects… Second, we zoom in on the impact of EPU shocks originating in the US, Europe, and China… Third, we explore regional patterns of EPU spillovers.” A deliberate scope choice separates the paper from the global-uncertainty literature (Mumtaz and Theodoridis 2017; Berger et al. 2016): “While these studies rely on broad measures of uncertainty that reflect volatility in economic or financial variables, our study centers on uncertainty arising specifically from economic policy.”

By the newspaper-based EPU index of Baker, Bloom and Davis (2016), available monthly for 19 countries and averaged to quarterly – with measurement error the authors name themselves (Section 2, pp. 4-5). The index “measures economic policy-related uncertainty based on newspaper frequency counts of keywords associated with EPU,” with search terms adapted per country “to account for particularities of the country and its language” and specified in the newspapers’ native language; it “captures uncertainty regarding who will make the economic policy decision, when, and what the effect would be as a result of the action” (fn. 4, p. 4). Coverage is unbalanced: 1985-2016 for the longest series, at least 2003-16 for the shortest, and “for the majority of the countries in the dataset, the series start in the first half of the 1990s.” A footnote flags the limits candidly: “Despite its widespread use in the literature, the EPU index remains potentially subject to measurement error. This may arise, for example, from unequal media coverage across journals and countries, subjective interpretation of the facts, and different writing styles” (fn. 5, p. 4). For comparability across countries the paper uses the purely newspaper-based US index rather than the composite one Baker et al. also construct. The macro aggregates – real GDP, private consumption and private investment growth – come from the IMF World Economic Outlook supplemented by OECD and national accounts, with an annual dataset built alongside because most control variables exist only annually.

Negative and significant relationships with all three aggregates, correlations between -0.2 and -0.3, and a strong tendency for countries to experience uncertainty episodes simultaneously (Section 2, pp. 5-6). The scatter plots give “a correlation coefficient ranging from -0.2 for growth in private investment to -0.3 for growth in GDP and private consumption,” more negative still if observations with EPU above 250 are dropped (fn. 7, p. 5). The country-by-year heatmap shows heterogeneous uncertainty up to 2002, a uniformly low period between 2003 and 2007, a rise from the global financial crisis with persistent cross-country differences “possibly reflecting different degrees of credibility of the commitments to take certain policy actions,” another group affected by the 2011 European debt crisis, and a uniform increase again in 2016-17 – “prima facie evidence that countries often tend to experience episodes of low and high EPU at the same time,” which is precisely what makes separating common from idiosyncratic components necessary. Regionally, US EPU correlates negatively with real GDP growth in Asia and Pacific, Europe and the Western Hemisphere but not in Africa or the Middle East and Central Asia; for consumption and investment the negative correlation survives only in Europe and the Western Hemisphere.

By allowing complete cross-country heterogeneity and decomposing each structural shock into a cross-sectional common part and a country-specific residual, following Pedroni (2013) (Section 3.1.1, pp. 7-10). The model is a bivariate heterogeneous panel structural VAR in demeaned EPU and one activity variable, with country-specific lag lengths and an unbalanced panel. Each structural shock is written as a country-specific loading on a common shock (the cross-sectional average) plus an idiosyncratic residual; loadings are recovered from sample correlations between each country’s structural residuals and the cross-section-average residuals, which under the normalization used are the OLS estimates. Common and idiosyncratic impulse responses follow from combining the composite responses with the loading matrices, and idiosyncratic shocks are rescaled “such that all impulse responses can be interpreted as responses to unit shocks.” The definition of the object of interest is precise: “The common component refers to the effect of EPU that originates in any other country in the sample and spills over to the domestic economy. The idiosyncratic component, on the other hand, refers to the effect of domestic EPU on the domestic economy” (p. 7).

That uncertainty is predetermined with respect to the macroeconomic aggregates – justified by a data-release argument, and tested by reversing it (Section 3.1.1, p. 9; Section 4.1, p. 20). Following Bloom (2009) and Carriere-Swallow and Cespedes (2013), the recursive restriction orders uncertainty first. The justification is informational: “newspapers – and therefore economic agents – come to know about current macroeconomic aggregates only when data is released, which is in the following quarter. At the same time, any discussion in the news that generates uncertainty about economic policy can affect the behavior of economic agents contemporaneously.” The paper does not hide the counterargument: a slow-to-fast ordering would put EPU after the aggregates, and “while we test the robustness of the results to this alternative ordering, we deem more convincing the arguments supporting the baseline identification” (fn. 12, p. 9). The test confirms the sign and shape: under the inverted ordering “the response to an EPU shock remains negative and significant and presents a similar shape,” though “the size of the negative effect at the through is smaller under the alternative ordering, but quickly becomes of similar magnitude.”

Small per unit but significant and long-lived, with investment responding about three times as strongly as output (Section 3.1.2, pp. 10-11). At the median, and reading off a ten-unit increase in the index: real GDP growth falls 0.1 percentage point at its trough one quarter after the shock, with the response “statistically significant up to three years after the shock”; private consumption growth falls 0.1 point two quarters after, significant up to two years; and private investment growth falls 0.3 points two quarters after, with significance lost “during the fourth year after the shock.” The paper is explicit that heterogeneity across countries is large and that it grows as the sample shrinks: “The decline in real GDP growth in the third quarter after the shock is more than three times as large for the 25th percentile than it is for the 75th percentile” (19 countries); consumption shows more dispersion still on 15 countries, “possibly reflecting different shares of purchases of durable goods across countries,” with the effect small for about a quarter of the sample; and for investment, on only 8 countries, “the effect for the 25th percentile is about six times larger than for the 75th percentile.”

That the common (foreign-origin) shock accounts for roughly two-thirds of the median decline in GDP growth at every horizon, and dominates for consumption and investment as well (Section 3.1.2, p. 11). “Figure 6 shows that a common uncertainty shock explains roughly two-thirds of the decline in GDP growth across all horizons. Common shocks are also more important than idiosyncratic shocks for private consumption and private investment growth. Hence, spillovers of policy uncertainty do impact economic activity, and to a larger extent than domestic shocks.” The authors note this sits alongside prior findings rather than overturning them: spillovers from general US uncertainty have been found “large and similar in magnitude to domestic effects” (Colombo 2016; Kamber et al. 2013; Mumtaz and Theodoridis 2015), and the global-uncertainty literature finds global uncertainty more important than country-specific uncertainty (Berger et al. 2016). Note what the decomposition is and is not: “common” means originating in any other sample country, not specifically in a large economy – the large-economy question is the local projections exercise.

For flexibility – separate regressions per horizon, a large control set, and nonlinear regional specifications – with exogeneity imposed by dropping the origin economies from the sample (Section 3.2.1, pp. 11-13). Local projections “generate estimates that are less vulnerable to misspecification of the data generating process because the impulse response is estimated separately for each horizon,” allow “controlling for a relatively large set of variables, which would be impractical in a regular VAR setting,” and “can easily accommodate non-linear specifications.” No year dummies are included “because the variable of interest, EPU-j, is common across all countries.” Because US, European and Chinese uncertainty are highly collinear, the three are estimated separately, following IMF (2013). Exogeneity comes from exclusion: “we compute the response to a shock to EPU for the sample that excludes the countries in which the shock originates” – so the US response is dropped from the US-shock sample, and France, Germany, Italy, Spain and the UK from the European one (European EPU being the aggregate of those five). A footnote addresses the obvious alternative: using each origin’s EPU orthogonalized against the others “would greatly reduce the variation in the series related to idiosyncratic events occurring during the same year,” listing 2011 (euro crisis, US debt ceiling, China leadership transition), 2015 (European immigration crisis, Chinese equity sell-off) and 2016 (Brexit, US election) as overlapping episodes; when the orthogonal components are used anyway, “the results for shocks to EPU in the US and China are broadly consistent, and … the results for shocks to European EPU are mixed” (fn. 17, pp. 12-13).

US shocks are the largest, European and Chinese shocks similar to each other and smaller, all lasting about two to three years (Section 3.2.2, pp. 13-14). For the United States, a ten-unit rise in the index is associated with roughly a 0.2 percentage point decline in real GDP growth across 64 other economies, largest three quarters after the shock and dying out during the third year; private consumption (37 economies) and private investment (13 economies) trough four quarters after the shock, with investment the most responsive at “roughly 0.6pp at the trough.” The authors supply a concrete scale: the quarterly average US index rose 53 units between 2008Q2 and 2008Q3, and “an increase in the US EPU of this magnitude is associated with a decline of 1.1pp in GDP growth, on average, in other economies.” For Europe and China, the negative effect on foreign GDP growth “lasts for about two years and displays the trough four quarters after the shock,” with consumption and investment troughing at three to four quarters, and “spillovers from Europe are similar to spillovers from China, and are smaller than the spillover effects from the US.”

Asia and Pacific, Europe and the Western Hemisphere – with the latter two hit hardest, and Africa and the Middle East and Central Asia largely unaffected (Section 3.2.2, pp. 14-15). Using annual data for broader regional coverage and a regional-dummy interaction, “spillovers from policy uncertainty in the US are significant for Asia and Pacific, Europe, and the Western Hemisphere. Compared to the rest of the world, the spillover effects from US policy uncertainty are larger and more statistically significant for Europe and the Western Hemisphere.” The authors’ reading is deliberately modest: “These findings make sense intuitively since they support the idea that spillovers occur between economies that are more closely integrated.” Two asymmetries are worth carrying. European uncertainty is narrower in reach than US uncertainty: “an increase in the European EPU index does not, however, affect consumption and investment growth outside Europe.” Chinese uncertainty is broader than that: it produces negative responses in GDP, consumption and investment in Europe and the Western Hemisphere, while the Asia and Pacific GDP response is “only marginally significant” – attributed both to “the presence of many small economies in the region, in which economic activity is strongly affected by natural disasters, climate change, and tourism” and to the trade channel, since “while China does trade with the other countries in the region, the US and Europe are its main trading partners.”

The paper tests both, and the results hold with one named exception (Section 3.2.2, p. 15; Section 4, pp. 20-26). Against omitted variables, the annual local projections add controls chosen from the growth and consumption/saving literatures – log real per capita GDP, terms-of-trade growth, the real oil price, the old-age dependency ratio, the urban population share, and lagged inflation, real appreciation, the real interest rate, private credit flow growth and real government purchases, consumption and investment growth – with potentially endogenous controls lagged a year; the results are “consistent both in terms of timing and magnitude with the results reported without the controls.” The authors explain the weak fit rather than ignoring it: “The low R2 can be explained by the heterogeneity of countries included in our sample. When we restrict the sample to advanced economies, the R2 increases and the coefficients of interest remain similar” (fn. 21, p. 15). Against rival concepts of uncertainty, both the panel VAR and the local projections are re-run adding, one at a time, GDP forecast dispersion (general uncertainty), stock market volatility (financial uncertainty), and consumer and business confidence – the last motivated by Redl (2017), “who cautions against conflating uncertainty and confidence.” The median panel-VAR responses “remain negative and significant for all aggregates,” with the investment trough halved when general uncertainty is controlled for. In the local projections the exception is named: “The decline in private investment in response to a European EPU shock is the only response that ceases to be significant.”

Sub-samples, alternative shock definitions, lag structures, annual re-estimation on a far larger country set, and a system-GMM treatment of reverse causality (Section 4, pp. 20-26). Splitting at the global financial crisis leaves results “remarkably similar to the ones with the unrestricted sample, even though the confidence intervals are comparatively larger.” Redefining the EPU shock as a dummy for the HP-filtered cyclical component exceeding its mean by more than 1.65 standard deviations – the IMF (2013) definition – or as the percent deviation from trend “does not affect our results,” and nor does adding lags. Re-estimating the panel VAR on annual data for countries with at least 20 years of observations gives a trough one year after the shock with “somewhat larger” magnitude and the same spillover-dominates-domestic decomposition. Annual local projections expand coverage dramatically – “the annual data allows a larger sample of 185 countries” – and give spillovers “comparable in magnitude” lasting two to three years. Finally, replacing the lagged-regressor fix for reverse causality with a two-step system GMM estimator (Arellano and Bond 1991; Blundell and Bond 1998, with a collapsed instrument matrix and the Windmeijer correction) leaves “the trough in year one … of the same magnitude and statistically significant,” though “the significance levels of the responses at other horizons change for some of the local projections.”

In associational language throughout, with the two-thirds spillover share as the headline and a communication-and-buffers policy message (Section 5, pp. 27-28). “In line with the literature, we find that EPU shocks are associated with declines in the real growth of output, private consumption, and private investment. Spillovers are the key driver of the negative impact of uncertainty on domestic macroeconomic aggregates, accounting for about two-thirds of the negative effect. Furthermore, surges in EPU in the US, Europe, and China depress economic activity in other countries up to about two years following the shock. This effect is mostly felt in Europe and the Western Hemisphere.” On policy, the prescriptions are split between origin and recipient: “policy decisions – where possible – should be clearly and timely communicated,” and “credible commitments to policy implementation may be helpful in minimizing the risk of economic agents adopting a ‘wait-and-see’ approach,” while on the receiving end “sound policy frameworks, sufficient policy buffers, and strong fundamentals in recipient economies can help mitigate the effects of uncertainty shocks.” The closing qualification is explicit that no single prescription follows: “More specific policy responses, though, may vary across countries depending on the available policy mix, the strength of the exposure, and their openness.”