Does US partisan conflict matter for the Euro area? Chak Hung Jack Cheng∗

William B. Hankins†

Ching-Wai (Jeremy) Chiu‡

October 8, 2015

Abstract This paper highlights the international transmission of political uncertainty originated from a US partisan conflict shock, a newly identified shock that transmits a type of uncertainty beyond the economic policy uncertainty spillovers identified by Colombo (2013). Using the recently developed US Partisan Conflict Index (USPC) developed by Azzimonti (2014), we find that a one standard deviation USPC shock leads to a 0.2 percent decline in European industrial production. We also show that, compared with US policy uncertainty shocks, a shock to US partisan conflict creates deeper and more persistent spill-over effects to the Euro area. Keywords: Partisan conflict, Economic policy uncertainty, U.S.–Euro area spillovers, Bayesian VAR JEL Classification: E32, F42, D72

1

Introduction

As the world is more financially and economically integrated, domestic shocks from one country are increasingly likely to be transmitted across borders. Colombo (2013) showed that heightened economic policy uncertainty in the United States could have a detrimental impact on the European economy. Over the past decade, political conflict in the US has become much more severe. At the same time, evidence shows that Europeans believe that the US and US politics matter in global affairs.1 Whether US political uncertainty, which can stem from a partisan conflict shock, as opposed to US policy uncertainty shocks, which can be triggered by factors that are completely unrelated to political conflict, spill over to the rest of the world becomes an important question for economists and political scientists alike. As Azzimonti (2014) notes, an intensification of partisan conflict creates uncertainty about which policies politicians will choose. This is distinct from uncertainty ∗ Corresponding author. Assistant Professor of Economics, George Dean Johnson, Jr. College of Business and Economics, University of South Carolina Upstate, 160 East St. John Street, Spartanburg, SC 29306, USA. Email: [email protected]. Phone: 864-503-5510. Fax: 864-503-5583. † Assistant Professor of Economics, George Dean Johnson, Jr. College of Business and Economics, University of South Carolina Upstate, 160 East St. John Street, Spartanburg, SC 29306, USA. Email: [email protected]. Phone: 864-503-5580. Fax: 864-503-5583. ‡ Bank of England, Threadneedle Street, London, United Kingdom EC2R 8AH. Email: [email protected]. 1 From 2003 – 2006, in every European Commission’s Eurobarometer Survey, at least 70% of respondents believed that the US played either a positive or negative role regarding growth of the world economy. Additionally, a 2009 survey conducted by the Wall Street Journal found that most Europeans felt that American political influence over the period 2004 – 2009 had been negative (http://www.wsj.com/articles/SB124534162608828017).

1

over the effects of existing government policies, which is generated by economic policy uncertainty shocks. There is a literature investigating the domestic impact of political shocks. Azzimonti (2015) proposes that a partisan conflict shock can cause a reduction in private investment by raising uncertainty about government policy. Azzimonti and Talbert (2014) theorized that more frequent political turnover and a higher degree of political polarization will lead to more economic policy uncertainty. Using a news-based measure of economic policy uncertainty from Brogaard and Detzel (2015) for nineteen developed and developing countries, they show a positive correlation between own-country political polarization and economic policy uncertainty. This paper hypothesizes further that more intense partisan conflict will have an economic impact not only in the US but also in the Euro area. As European economic agents have become more aware of the influence that US politics can have on the global economy, political gridlock in the US may prompt Europeans to postpone consumption because of a precautionary saving motive as well to defer investment due to the increase of the option-value of waiting (Bloom (2009)). Moreover, the US has a close trading relationship with the Euro region. As output and employment decline in the US after a political shock, European economies can suffer due to a decreased demand for their exports. We investigate the question of interest by building on Azzimonti (2015). Using Azzimonti’s novel Partisan Conflict Index (USPC), and after controlling for economic policy uncertainty and macroeconomic indicators, we show that a one standard deviation shock to USPC is associated with a 0.2 percent decline in European industrial production and the negative effect lasts for up to eighteen months. Furthermore, we find that the effect of a partisan conflict shock on the Euro area is deeper and more persistent than policy uncertainty shocks identified by Colombo (2013). These results thus extend a growing literature showing how political disagreements can have real economic consequences both domestically and internationally. The remainder of the paper is organized as follows. Section 2 focuses on the data and the empirical model. Section 3 discusses the results and Section 4 concludes.

2

The Empirical Model

The effects of a partisan conflict shock are estimated through the following vector autoregression (VAR) model: B(L)yt = d + t where yt represents the endogenous variables, d is a constant term and t are the reduced-form residuals, fulfilling E(t ) = 0 and E(t 0t ) = Σ. B(L) is given by = I + B1 L + B2 L2 + ... + BN LN , where N is the lag length of the VAR model. The following variables are included in the VAR model: the US partisan conflict index (U SP Ct ), the US consumer price index (U SCP It ), the US industrial production index (U SIPt ), the federal funds rate (U SF F Rt ), the US economic policy uncertainty index (U SEP Ut ), the consumer price index for the Euro area (EAHCP It ), the European industrial production index (EAIPt ), the three-month interest rate for the Euro area (EARt ) and the European economic policy uncertainty index (EAEP Ut ). We follow Colombo (2013) and employ monthly data from 1999M1 to 2008M6 in our estimation. Our sample’s starting date is based on the creation of the Euro area and the end date is chosen to 2

Figure 1: US Partisan Conflict Index and Economic Policy Uncertainty: 1999 – 2008

Note: The left axis measures the USPC index and the right axis measures the USEPU index.

avoid possible non-linearities due to the financial crisis. U SP Ct refers to the news-based Partisan Conflict Index (USPC) developed by Azzimonti (2014).2 We proxy for U SEP Ut and EAEP Ut using the news-based components of the economic policy uncertainty (EPU) indexes developed by Baker, Bloom, and Davis (2015). Figure 1 displays the USPC Index and the news-based USEPU index and shows that uncertainty over economic policy can be high during periods of political calm and vice versa.3 Azzimonti(2014,2015) suggests that during periods of high partisan conflict, where meaningful policy is unlikely to be passed, investors and consumers will expect the policy status quo to persist, making economic policy uncertainty low. Furthermore, she shows that the USPC index is mainly driven by political factors whereas the EPU index can be affected by financial shocks and monetary policy. All variables, except for the federal funds rate and the European interest rate, are expressed in log values. The industrial production indexes and consumer price indexes for both the US and the Euro area are detrended with a linear time trend.4 In the baseline specification, we assume the following ordering:

yt = [U SP Ct , U SCP It , U SIPt , U SF F Rt , U SEP Ut , EAHCP It , EAIPt , EARt , EAEP Ut ] The model is estimated with Bayesian methods using four lags. To recover the structural shocks from the residuals, we follow a standard Cholesky decomposition approach. Following Colombo (2013), we order the US block before the European block and order policy uncertainty last within each block. We also assume that USPC does not respond to lagged values of other US and European variables by imposing tight zero-mean priors to shrink the coefficients of the relevant parameters to zero. The rationale is that increased political conflict likely originates during election periods or is due to increased political polarization.5 2

This is available at https://www.philadelphiafed.org/research-and-data/real-time-center/ partisan-conflict-index/. 3 The unconditional correlation between the two indexes is -0.17. 4 Data sources: Federal Reserve Bank of St. Louis, European Central Bank, and the Economic Policy Uncertainty website, http://www.policyuncertainty.com/. 5 As robustness checks, we also consider: (i) relaxing the exogeneity assumption on the USPC index (ii) different

3

3

Results and Discussion

The estimated impulse responses to a one standard deviation positive shock to USPC (a rise of 15%) are shown in Figure 2. We display the median response functions along with the 68% error bands as recommended by Sims and Zha (1999).6 The model predicts that a positive shock to USPC exerts a significant negative impact on US industrial production. US industrial production reaches its lowest values of around -0.2 percent after ten months. This result is consistent with the findings in Azzimonti (2014). An increase in partisan conflict is a signal to investors that the government is less willing or able to enact policies that reduce the risk of costly rare events. Thus, they will reduce investment. The model also predicts a significant fall in the US interest rate. Interestingly, the underlying dynamics are more nuanced. There is a short-lived reduction in economic policy uncertainty after the positive shock. This result is seemingly consistent with Azzimonti’s hypothesis that relatively high levels of partisan conflict can be associated with low levels of economic policy uncertainty, since investors and consumers expect little change in economic policy in the short term. This probably explains a small and temporary growth in industrial production and prices within the first six months of the shock. More importantly, our VAR model predicts a negative and significant response of European industrial production after a shock to US partisan conflict. Industrial production in the Euro area reaches its maximum contraction of -0.2 percent after ten months and the negative effect does not wear out until after eighteen months. Owing to a large VAR system estimated with a relatively short data sample, the estimation uncertainty is not small. However, the median response of European industrial production is estimated precisely enough to conclude that the response differs from zero at the recommended 68% level. Figure 3 displays the impulse responses to a one standard deviation shock (a rise in 20%) to the USEPU index. These results are largely consistent with Colombo (2013). Two interesting observations are in order. First, a shock to US partisan conflict creates a deeper fall in European output. Second, the impact of a US policy uncertainty shock on both the US and Euro area are relatively short-lived, and tends to occur within six months of the shocks. In contrast, the significant negative impact of a US partisan conflict shock does not set in until after half a year, and last for about ten months. Table 1 displays the results from the forecast-error-variance (FEV) decomposition. We can see that a US partisan conflict shock plays a non-trivial role in explaining the fluctuations in European industrial production, especially at longer horizons: the shock explains sixteen percent of FEV at a horizon of eighteen months and thirteen percent at a horizon of twenty-four months. A US partisan conflict shock is also important for explaining the variations in European consumer prices and policy rates. It accounts for approximately fifteen percent and seventeen percent, respectively, of FEV in European consumer prices and the three-month interest rate at the twentyfour month horizon. Table 1 also shows that, in general, a US policy uncertainty shock is more relevant than its European counterpart. More importantly, we find that the contribution of a US partisan conflict shock on the variations of European aggregates is larger than those from US and European policy lag length specifications; (iii) introducing additional variables to the VAR, i.e. S&P 500 Index, Euro Stoxx 50 Index, nominal exchange rate, and consumer sentiment index; and (iv) different orderings within each country-specific block. None of these change the results significantly. These results are available in the appendix. 6 Sims and Zha (1999) suggest using 68% interval bands to provide a more precise estimate of the true coverage probability.

4

Figure 2: Impulse responses to a one standard deviation positive shock to US partisan conflict. Note: The error bands correspond to 68 percent intervals. USPC

USCPI

USIP

5 0

percent

10

0

0.1

percent

percent

15 0 −0.1 10

20 months

30

40

0

10

USFFR

0

10

20 months

30

40

30

40

30

40

EAHCPI

−0.01 20 months

30

40

0 −5 0

EAIP

10

20 months

0

pp

5

0

40

10

20 months

30

−10 0

40

10

20 months

20 months

EAEPU

−5

−0.2

10

EAR

−3

x 10

0.2

30

0 −0.05 0

percent

10

0.05

percent

percent

pp

0

percent

40

5

0.01

0

30

0 −0.2

USEPU

0.02

0

20 months

0.2

30

40

5 0 −5 0

10

20 months

Figure 3: Impulse responses to a one standard deviation positive shock to US policy uncertainty. Note: The error bands correspond to 68 percent intervals. USPC

USCPI

0 −1 10

20 months

30

percent

percent

percent

1

0

0 −0.1

40

0

10

30

40

0

10

20 months

30

40

30

40

30

40

EAHCPI

−0.01 10

20 months

30

40

percent

percent

10 0 0

EAIP

10

20 months

pp 20 months

30

40

0

0

10

10

20 months

5

20 months

EAEPU

−5 10

−0.1 0

40

5

0

−0.05

EAR

−3

x 10

0.2

30

0

percent

pp

20 months

USEPU

0

percent

0

20

0.01

−0.2 0

0.2

−0.2

USFFR

0

USIP

0.1

2

30

40

10 0 −10 0

10

20 months

uncertainty shocks at long horizons. Table 1: Forecast error variance decomposition of European variables due to USPC, USEPU and EAEPU shocks Horizon (Months) 2 6 12 18 24 36

Consumer Prices USPC USEPU EAEPU 0 1 0 12 4 4 18 12 5 15 13 4 15 13 4 18 13 3

Industrial Production USPC USEPU EAEPU 0 0 2 1 3 5 10 4 5 16 8 4 13 9 3 15 8 5

USPC 2 1 5 15 17 13

Policy Rate USEPU EAEPU 13 0 15 2 10 2 9 3 11 2 10 2

Note: Unit in percent

4

Conclusion

In this paper, we estimate a Bayesian VAR model using a novel political indicator developed by Azzimonti (2014). The results show that even after controlling for US and European economic indicators and policy uncertainty, a positive shock to US partisan conflict exerts a significant negative effect on European industrial production.

Acknowledgments The views expressed in the paper are those of the authors and do not necessarily reflect the views of the Bank of England. The authors would like to thank Pierre-Daniel Sarte (Editor) and an anonymous referee for helpful comments. Cheng and Hankins would like to thank the University of South Carolina Upstate for providing research support. Any remaining errors are the sole responsibility of the authors.

References Azzimonti, Marina (2015). “Partisan conflict and private investment”. NBER Working Paper No. 21273. Azzimonti, Marina (2014). “Partisan conflict”. Working Paper No. 14-19 Federal Reserve Bank of Philadelphia. Azzimonti, Marina and Matthew Talbert (2014). “Polarized business cycles”. Journal of Monetary Economics 67, pp. 47–61. Baker, Scott R, Nicholas Bloom, and Steven J Davis (2015). “Measuring economic policy uncertainty”. Stanford University and University of Chicago Booth School of Business. Bloom, Nicholas (2009). “The impact of uncertainty shocks”. Econometrica 77.3, pp. 623–685. Brogaard, Jonathan and Andrew Detzel (2015). “The asset-pricing implications of government economic policy uncertainty”. Management Science 61.1, pp. 3–18. Colombo, Valentina (2013). “Economic policy uncertainty in the US: Does it matter for the Euro area?” Economics Letters 121.1, pp. 39–42.

6

Sims, Christopher A and Tao Zha (1999). “Error bands for impulse responses”. Econometrica 67.5, pp. 1113–1155.

7