US Political Cycles and Internations Stock Returnswp.lancs.ac.uk/fofi2020/files/2020/04/FoFI-2020-054-Hsuan-Fu.pdf · Hsuan . is with the Department of Finance, Insurance and Real

US Political Cycles and International Stock Returns

Pasquale Della Corte∗ Hsuan Fu†

Imperial College London Laval University

January 11, 2020

PLEASE DO NOT CITE WITHOUT AUTHORS’ PERMISSION

Abstract

This paper demonstrates the US political cycle is important to understand the risk premia of the stock and currency markets worldwide. We frst document that the stock excess returns are higher and the USD is more appreciated when the US president is a Democrat rather than a Republican. Then, we argue that the US trade policy generates the cross-country spillovers to the non-US markets. From the networks constructed by trade and cross-border capital fows, we also fnd that the peripheral countries have higher stock returns when the US president is a Democrat. Lastly, we fnd that the return di˙erences are largely mitigated after controlling for the trade tari˙. It seems that the US presidential puzzle which is also observed in the international fnancial markets can be rationalized by the level of trade protectionism.

Keywords: Political Cycles, International Stock Market, Currency Risk Premia, Trade, Tari˙, Network Centrality

JEL Codes: D72, F13, G15, G40, P16

∗Pasquale is with the Department of Finance in Business School, Imperial College London. Contact: [email protected]

†Hsuan is with the Department of Finance, Insurance and Real Estate, Faculty of Business Administration, Laval University. Contact: [email protected]

1

1 Introduction

International stock markets perform much better when the president of the US is a Democrat rather

than a Republican. Among the 23 countries contained in MSCI World Index, the average excess

return denominated in USD is 7.4% when the US president is a Democrat and -0.4% when the presi-

dent is a Republican. This 7.8% di˙erence in excess stock returns (henceforth, the return di˙erence)

is both signifcantly and economically large. Astoundingly, the average return di˙erence denomi-

nated in local currencies is as large as 11% per year. It remains an ongoing debate for researchers to

understand the reasons causing such di˙erence whilst it could be of interest to practitioners around

the world to build a trading strategy based on US presidential cycle.

While there were some papers documenting and explaining how the US presidential cycle matters

for macroeconomic outlook, only a few papers studied the presidential “puzzle” over stock market

performance. The return di˙erence of the US stock market was documented by Santa-Clara and

Valkanov (2003) while Pastor and Veronesi (2019) attempted to rationalize it with the US tax

policy and time-varying risk aversion. It is worth noting that the literature found a much smaller

di˙erence in the US stock market in the subsample between 1950s and 1980s. Coincidently the

international trade liberalization began to thrive during the same subsample. In addition, Figure 5

shows the global market began to have cross-border capital fows only since 1980s. We hence argue

that international factors may be one critical reason causing the return di˙erence in the non-US

countries.

In this paper, we aim to analyze the relationship between the US presidential cycle and the

return di˙erences in the international stock and currency markets. We frst analyze if the non-US

political cycles can also generate sizable return di˙erences. The results are ambiguous. Generally,

the return di˙erences by non-US political cycles are statistically signifcant but the signs are incon-

sistent cross-sectionally and across subsample periods. Therefore, it is diÿcult to reach a conclusion

that the conventional bipartisan hypothesis plays any role in our analysis. Only the US political

cycle is found to generate consistent and signifcant return di˙erences in the international stock and

foreign exchange returns. The US political cycle also generates di˙erence in other risk-based mea-

sures, including Sharpe ratio, Value at Risk, maximal drawdown, etc. Under Democrat presidential

terms, the risk-adjusted stock returns are found to be persistently higher than under Republican

2

presidential terms.

To observe the data from the perspectives of both local and global investors, we investigate

the stock returns denominated both in USD and in local currencies. We fnd the return di˙erence

denominated in local currencies to be higher than in USD but the subsample exercises did not show

any signifcant result until 1980s. In the most recent subsample, we fnally discover a statistically

signifcant USD appreciation under the Democratic rather than under the Republican presidents.

We argue that the US political cycle can indeed spillover to the non-US stock and currency market

potentially via the international trade.

Before digging further into the trade channel, we conduct a few analyses to rule out wealth e˙ect

as a possible explanation for the return di˙erences in the international stock and market. If wealth

e˙ect exists, the investors feel richer when the US stock return rises under Democrat presidents.

Therefore, they increase the long position in both domestic and foreign the stock markets. That

would be a channel through which the US political cycle propagates to the international stock

markets. We did not fnd very strong evidence of the wealth e˙ect. The return di˙erence shrinks

only slightly and remains statistically signifcant after we control for wealth e˙ect. We argue that

there should be other factors which are more important than the wealth e˙ect to understand the

international return di˙erences.

Next, we would like to investigate the possible explanations to rationalize the cause of the return

di˙erence. Trade policy is a natural candidate in the international context. Given our focus on the

trade policy, the US president stands in a very special state compared with other presidents or

prime ministers in non-US countries. There are some evidence reporting that the US president can

easily bypass the congress to sign a tari˙ order.1 Therefore, it also justifes our sole focus on the

presidential cycle rather on congressional characteristics.

We consider the international trade relations as a trade network because an increase in the trade

volume of one country might stimulate its import indirectly from some other countries. We observe

the trade network e˙ect in the data. If the trade integration is one critical factor driving the return

di˙erences, we expect to fnd stronger e˙ect in the peripheral countries, who are more likely to

beneft (su˙er) from a pro-(anti-)trade US government. Since the capital fows are usually highly 1Anecdotal evidence: President Trump and Bush impose tari˙ on metal imports between 10% and 30% in March

2018 and 2002, respectively. The evidence from academic literature is reviewed in Section 2.

3

correlated with the trade fows, for the robustness, we also investigate a similar network with the

data of mutual fund fows. Under Democratic presidents, the peripheral countries are found to have

higher returns, implying that the fund mangers seem to be more diversifed in their portfolios. To

provide more direct evidence on the trade channel, we further analyze the infuence of trade tari˙ on

the international stock and currency markets. We demonstrate that the return di˙erence is largely

mitigated after the inclusion of tari˙-related variables in our analysis. Therefore, we argue that

the level of Protectionism, commonly measured by tari˙, is essential to understand the US political

cycle in the international stock and currency markets.

In short, our contribution can be summarized in three dimensions. First, we establish the

connection between the US political cycle and international stock and foreign exchange markets.

We demonstrate that it is the US political cycle, not any other country’s, that generates the return

di˙erences. Secondly, we rule out the wealth e˙ect as a potential hypothesis to explain why the US

return di˙erence propagates to the international capital markets. In addition, we also rule out the

hypothesis that the return di˙erences can be anticipated given the pre-election economic conditions.

Thirdly, we propose trade policy as a plausible explanation for the sizable return di˙erences in the

international markets. We show that the trade centrality amplifes the return di˙erences in the

international stock markets and we also demonstrate that the return di˙erences drop dramatically

after controlling for the tari˙ variations.

The remaining of the paper is organized as follows. We review the related literature in Section 2.

Section 3 describes our data resources. Section 4 presents the main fndings to document the

spillovers of US political cycles to international stock markets. Section 5 rules out the market

expectation as an explanation to our main fnding of the return di˙erence. Then, we propose a

trade protectionism explanation in Section 6 for the return di˙erence. Section 7 concludes.

2 Related Literature

The return di˙erence is well known in the US stock market. Santa-Clara and Valkanov (2003) ruled

out some potential explanations and documented this phenomenon as “presidential puzzle.” They

did not fnd signifcant relations between stock returns and congressional variables while similar fnd-

ings were reported by Blinder and Watson (2016) over economic growth and congressional variables.

4

The model in Pastor and Veronesi (2019) focused on the imposed tax policy under the Democratic

and Republican presidents. They rationalized the return di˙erence in the US stock market by

providing an explanation of fscal policy and time-varying risk premia. The infuence of political

uncertainty on the risk premia was well-explored with the US data. Pástor and Veronesi (2013)

proposed a general equilibrium model to rationalize the price dynamics responding to political news.

Kelly, Pástor, and Veronesi (2016) used option data to verify the link between political uncertainty

and risk premia.

Our paper is also related to the bipartisan models (e.g., Hibbs, 1977) and the political real

business cycle (e.g., Nordhaus, 1975), in which Democrats prioritize growth over infation and

unemployment while Republicans favor the opposite. Alesina and Roubini (1992) investigated 18

OECD economies to document the long-run bipartisan di˙erences in infation and the temporary

bipartisan di˙erences in output and unemployment. On the contrary, Blinder and Watson (2016)

documented the bipartisan di˙erence in the US economic growth. Hence, it remains inconclusive

whether the bipartisan hypothesis is an important factor to understand the international stock and

currency market.

Lohmann and O’Halloran (1994) showed empirical evidence of lower tari˙ under Democratic

president. They also showed similar fnding under unifed government where the President is in

the same party as the House and Senate majority. Recently, Fajgelbaum, Goldberg, Kennedy,

and Khandelwal (2019) and Fetzer and Schwarz (2019) investigated the economic losses due to the

trade war raised by President Trump’s administration via a specifc dimension of the retaliation

tari˙ enacted by the US trade partners. Our paper complements theirs by documenting that tari˙

rate is an important factor to explain also the fnancial returns in a longer sample. On the other

hand, Liu and Shaliastovich (2017) argued the relations between the policy approval and currency

risk premium. They showed higher rate of policy approval predicts higher economic growth and

lower currency risk premium. Our paper echoes theirs by relating the US trade policy and the

international asset prices. We further propose the US trade policy as a potential explanation for

the return di˙erence in the international stock markets.

Moreover, the infuence of political uncertainty on foreign exchange market was discussed in a

few papers. Bachman (1992) o˙ered an information-based explanation, showing that the forward

exchange premium is mitigated after general elections in a number of industrial countries. On the

5

international dimension, our paper is related to but di˙erent from the work by Brogaard, Dai, Ngo,

and Zhang (2019). They studied the impact of political uncertainty on the international asset prices

but the focus was to show the negative infuences from the pre-election uncertainty rather than to

understand the policy choice throughout the presidential terms. Our goal is beyond this type of

event study which treats election as an exogenous shock. We aim to explain the political-economic

fuctuations in the risk premia of the stock and foreign exchange markets. With a more general

setting, Lobo and Tufte (1998) specifed a GARCH model to analyze if the US bipartisan e˙ect

can explain the exchange rate volatilities. Although they found the US political cycles important,

whether exchange rate volatilities increase or decrease under Democratic presidents remained in-

conclusive within a small sample of only 4 countries. In order to document some stylized facts, we

include in this paper a comprehensive set of 23 countries and study a fairly long sample period from

1929 to 2018. That would allow us to have suÿcient power to identify the long-standing infuences

of political cycles on the risk premia.

The trade centrality measure we used is similar to the eigenvector centrality in network literature

(e.g., Bonacich and Lloyd, 2001; Jackson, 2010). Richmond (2016) used trade network centrality

to rationalize the economic source for the deviation from Uncovered Interest Parity. Instead, our

tests based on the network centrality further validate the role of trade integration in explaining

the return di˙erence. Rossi, Blake, Timmermann, Tonks, and Wermers (2018) used additional

centrality measures to test the fund performance. We also include these measures in our analysis

for the sake of robustness.

3 Data

Following the MSCI World Index, we include 23 countries in our sample, which covers approximately

85% of the free foat-adjusted market capitalization. We collect our data mainly from the following

4 sources.

3.1 Financial Variables

First, stock market indices in both local currency and USD are from the Global Financial Data

(GFD). The data is at monthly frequency. The number of observation varies across countries.

6

Generally, we include the 90-year period from 1929 to 2017 whenever the data is available. We use

3-month Treasury yields as the risk-free rate, which is also from the Global Financial Data. As

for the risk-free rate, we use the US Government 90-day T-bills yield from the secondary market

when calculating the excess returns in the US Dollars. We also calculate the excess returns in local

currencies where the risk-free rate is the 3-month Treasury bill yields in each countries. The results

by using USD are similar to local currencies.

The Net Foreign Asset from the GFD is at annual frequency, whose availability varies across

countries, but is generally available after 1980.

The fund fow data is used to construct the global network and is from the EPFR Database.

Specifcally, we are using the country fow data which aggregates the fund-level data to country-

level data. The monthly data is available only from 1995. It is much shorter than the whole sample

period but its granularity is an advantage to improve the power of our regression analysis.

3.2 Political Variables

We hand-collect the time-series data of the presidential election and political parties information in

G7 countries. To categorize a party to be leftist or rightist, we compare the party ideology between

the election winner and the loser, which has largest number of popular votes or seats in parliament.

If the winner party is to the left of the loser party, it is categorized as leftist even though it might be

centrist to central-right in the international standard, e.g., newly-elected French president Macron

and his centrist party En Marche.

The World Bank Database of Political Institutions (DPI) by Scartascini, Cruz, and Keefer (2018)

has an abundant collection of electoral variables across 180 countries. The earliest observation is

from 1975 and the latest release is updated to 2017. The data is coded at annual frequency.

3.3 Trade Variables

The bilateral trade data in USD is from the International Monetary Fund (IMF) Direction of Trade

Statistics. It is available after 1980 and is at both monthly and annual frequency. We use it to

calculate the trade network centrality and later on to construct one of the tari˙ measure.

In addition, the tari˙ data comes from two sources. The frst measure is the Most Favored Nation

(MFN) tari˙ originally from the World Trade Organization (WTO). I collect the tari˙ series by the

7

World Bank (WB) which is calculated as the weighted-average of tari˙ rate across product within a

country. The data is only available from 1988. The MFN tari˙ might miss some information on the

duty free items and hence the political economy literature proposed a modifed tari˙ rate to fx this

concern. The second measure is originally from IMF Government Finance Statistics Yearbook but,

similar to the frst measure, I collect the annual data from the WB database. This tari˙ measure

requires three elements: Customs and other import duties, Tax revenues, and Total imports. The

advantage of this measure is its long time-series. The earliest possible observation is from 1972.

Nevertheless, there are many countries that do not have one of the three elements, e.g. Belgium,

Hong Kong, etc. It is impossible to construct a representative tari˙ measure based on the IMF

data. Therefore, we make use of these two tari˙ variables to construct a third tari˙ measure. To

avoid any discontinuity in the time-series within each country, we rescale the Average Tari˙ to have

the same range of variations as the MFN Tari˙ and use the scaled Average Tari˙ to fll the data

whenever MFN is unavailable.

3.4 Other Control Variables

There are 4 sub-groups of control variables. First, the forecast data, including the term spread,

the dividend-price ratio and the relative short rate (RREL), is from the Global Financial Data.

The term spread is the logarithm di˙erence between 10-year and 90-day government bond yields

and RREL is calculated as the current month’s T-bill rate minus a moving average of the past 12

months’ rate.

Second, the wealth data is proxied by the Household Net Worth provided by OECD, from which

we include two variables in our analysis: the percentage growth of the Household Net Worth and the

USD-denominated Financial Net Worth of the Household. In addition, the Federal Reserve System

also has Household Net Worth data but is available only for the US households. The OECD data

is at annual frequency and the Fed data is at quarterly frequency.

Thirdly, the market capitalization data is from the Global Financial Data. Its frequency is

monthly and is widely available in the most recent 30 years. Lastly, the economic variables like

the GDP denoted in USD is from the IMF World Economic Outlook and is available since 1960 at

annual frequency.

8

4 Main Findings

In this section, we will present the main fndings with graphs and regression analyses. We fnd

a very strong cross-sectional persistence of higher stock returns under the Democrat presidents.

Furthermore, we also fnd lower currency risk premia, i.e. more appreciated USD, during the same

period.

Figure 1 shows the return di˙erence across 23 countries during the last 30 years. The cross-

section average of the return di˙erence is more than 10% per annum. When the White House is

controlled by the Democrats, the average stock market excess returns and risk-adjusted returns are

persistently higher than the Republican presidents.

We further examine the stock market performance in a 90-year sample from 1929 to 2017.

Following Santa-Clara and Valkanov (2003), we then divide our sample into three subsamples of

equal lengths to check the if our fndings are consistent overtime.

Table 1 shows the summary statistics of both stock and currency returns before considering

the bipartisan e˙ect. Usually, the currency risk premium is positive correlated with stock returns

denoted in local currencies but this was not the case in the value-weighted return during the middle

subsample. We argue that the liberalization of the capital control in the 70s could contribute some

disruptions to the asset markets. Specifcally, when the Pound Sterlings depreciates against with

Euros, the expected cash fows of the British frms become larger as they are denoted in Euros.

The Panel A of Table 2 repeats the exercise in Table 1 and shows the value-weighted statistics

by political party. It is worth noting that the correlation between currency risk premium and stock

return are always positive under the Democrat but often negative under the Republican presidents.

In Figure 2 we show the risk-adjusted performance of international stock markets by calculating

their Sharpe ratios. In several countries, including United Kingdom, France and Netherlands among

others, the stock returns under Republican presidents are on average negative. On the contrary, the

stock returns are positive under Democratic presidents across all 23 countries in our sample. We

also look at di˙erent in some risk-based measures2 and found similar evidences of less left-tailed 2We also look into disastrous risk and downside risk measures such as Value at Risk (VaR), Expected Shortfall (ES),

and Maximum Drawdown (Maxdd) to performance risk-based analysis on the stock market returns under di˙erent terms of the US political cycles. Figure 3 shows that the international stock markets have higher VaR and ES (smaller negative returns). Figure 4 shows lower volatilities and smaller Maximum drawdown under Democratic presidents rather than Republican presidents. In other words, there are more fnancial crises happening under Republican

9

events that support higher stock returns under Democratic presidents among the majority of the

23 countries.

To control the serial correlation of the equity returns and country fxed e˙ect, we test the the

null hypothesis H0 : β1 = β2 with the following baseline regression:3

yt = β1 × DEMt−1 + β2 × REPt−1 + εt (1)

Aligned with the fnding in Santa-Clara and Valkanov (2003) and Pastor and Veronesi (2019),

Tables 3 and 4 present the fndings of weaker bipartisan e˙ect in the middle subsample from the year

1957 to 1986. Moreover, the stock market performance under Democratic presidents in quite a few

countries is worse than under Republican presidents. This counterfactual fnding implies that the

bipartisan e˙ect of the US political cycles seemed to be ambiguous during the frst two subsample

periods when capital control was still in place and very few international trade took place.4

Table 5 repeats the baseline regression in Equation (1) with the dependent variable as currency

risk premia. One can observe immediately that the coeÿcient for the Democratic dummy is in-

signifcant in the case of value-weighted currency risk premia for the frst two subsamples. On the

contrary, the coeÿcients for the Democratic dummy in the most recent subsample are signifcantly

negative, regardless of cross-country weighting method. This strong evidence indicates that the

USD is more appreciated when the presidents are from the Democratic party.

The lower panel of Table 5 presents the results by country. It is worth noting that most countries

in the middle subsample have negative coeÿcients for the Democratic dummy, as found in the last

subsample, although they are insignifcant. The fnding of the stronger USD under Democratic

presidential terms could be possible only after the 70’s when the capital mobility across countries is

largely improved. Combined with the fndings from the equity returns, it is therefore more relevant

to focus our analysis solely on the most recent subsample. In the following sections, we will use

only the observations since 1987, if they are available, to run the panel regressions.

presidents relative to Democratic presidents. Although the risk-based results are less robust than simple average results, the number of exception countries has been reducing with time.

3As demonstrated by Santa-Clara and Valkanov (2003), the baseline regression specifcation (1) equals the following one y = α + β × DEM + ε. Note that I suppress the time and country indices for brevity.

4Harrison and Tang (2005) reviewed the trade liberalization history and indicated that the waves of trade liberal-ization happened to the industrial countries after the Second World War and to the emerging countries in the 70’s, which is covered in our middle subsample.

10

5 Expected or Unexpected Return Di˙erences?

In this section, we begin to explore the explanation for the return di˙erence between the political

parties in the US. First, we would like to understand what is the role of political parties in the

non-US countries. If the market expectations are correlated with the bipartisan e˙ect, we should

observe similar patterns when regressing equity returns on the dummy of the leftist party’s in each

of the non-US country. Secondly, we would like to understand to what extent the return di˙erence

in the international stock markets can be directly inferred by the growth of the US investors’ wealth.

Two di˙erent variables, market capitalization and households’ net worth, are employed to capture

the wealth e˙ect. Lastly, we would like to understand if the projections of the economic outlook

throughout the whole tenure of a presidential term can explain the return di˙erence. If stronger

economic growths under Democratic presidents are expected to generate higher equity returns, we

should observe the explanatory power by the variables that capture pre-election real business cycle.

None of these hypotheses are suÿciently supported by the empirical data, so we rule out that the

expectations of the market or the economic outlook can explain the return di˙erence.

5.1 Local Political Cycles

In the previous sections, we have shown that the US political cycles can generate the bipartisan

e˙ect in the international stock market. A natural question to ask is: Can the bipartisan e˙ect in

each country explain the cycle of their own market performance? Table 6 shows summary statistics

of local political cycles of G7 countries. In the second column, we show example names of recently-

elected leftist party. We provide the following 3 reasonings to rule out the possibilities of bipartisan

phenomenon in international stock markets generated by local political cycles.

First, we fnd no consistent pattern of bipartisan e˙ect in international stock market by using

the political cycle. Secondly, the left-wing parties in foreign countries do not win as many elections

as the US along the history. The results are quite similar after considering the length of incumbent

time. Thirdly, the last column of Table 6 documents large standard deviations for the duration of

the election cycle, which indicates that the election happens less regularly in the non-US countries.

The standard deviation for US election cycle is approximately 10 days while in other countries, it

11

ranges from 6 to 22 months.5

In the next step, we test the null hypothesis H0 : β1 = β2 in the following regression:

yt = β1 × Leftt−1 + β2 × Rightt−1 + εt, (2)

where Left is an indicator of the periods when the presidents or prime ministers of one of the G7

countries are from the left-wing party and Right is an indicator of right-wing party. We could have

expanded this exercise to all 23 countries in our sample but we would like to see if we can fnd a

consistent pattern among the global leaders, G7 countries.

Table 7 presents the results of the equity returns. The fndings are demonstrated in three di˙erent

returns. Panel A shows the results of a pooled regression, Panel B shows the single-country returns

of each non-US countries and Panel C shows the global returns weighted by country’s GDP. In

the frst two subsamples, the political cycles of Germany, Japan and United Kingdom reject the

null hypothesis. However, the sign of di˙erence is fipped. In other words, the prime ministers

from the Left-wing parties seem to generate lower equity returns than those from the Right-wing

parties. This is inconsistent with the fndings related to the US political cycle. Moreover, in the

most recent subsample, only the US political cycle can reject the null hypothesis regardless the

return defnitions. Thus, there is no suÿcient evidence showing the bipartisan e˙ect generated by

non-US countries.

Table 8 has similar presentation as the previous table and shows the results of the currency

risk premia. If the bipartisan e˙ect is the main reason for causing the return di˙erence, we should

observe the same sign for all the non-US countries and the opposite sign for the US as the currency

risk premia take USD as the based currencies. Having the experiences from the previous analyses,

we frst interpret the most recent subsamples of the single-country currency risk premia from Panel

B. The di˙erence of the currency risk premia are mostly insignifcant although they are mostly

positive. The positive di˙erence indicates more appreciated foreign currencies against with USD

when a left-wing government is in power. As we can observe from the Panels A and C, this fnding

is aligned with the US case in which negative di˙erence suggests a more appreciated USD under 5Here, endogeneity is a critical issue when the general election does not occur regularly, as discussed in Goto,

Matsumoto, and Yamasaki (2018). In fact, six of the G7 countries do not have fxed terms for their presidents or prime ministers. It complicates further the specifcation if our main focus was the political cycles of all countries. For the moment, we would leave it for the future research.

12

the Democratic presidents.

The panel C of Table 8 is presented with the equally-weighted currency risk premia. We choose

this defnition in order to speak to a growing feld that studies the dollar factor in the international

fnance literature. See Verdelhan (2018) for further reference. Similarly, we fnd that only the US

political cycle generates the impact on the performance of the international markets in a consistent

pattern across all subsamples. The results of the non-US countries are either insignifcant or having

the wrong sign. In this exercise, Italian political cycle is an exception which makes it diÿcult to

interpret. The null hypothesis is rejected so it seems to demonstrate certain level of bipartisan e˙ect.

However, the sign of the di˙erence in Panel B is inconsistent with other six countries. Nevertheless,

this exception cannot be understood as the representative phenomenon among G7 countries as Italy

is relatively small in economic size. In conclusion, these fndings disprove the bipartisan e˙ect of

the non-US political cycles as an explanation for the return di˙erence in the international stock and

foreign exchange markets.

5.2 Real Business Cycle

Following the spirit of Santa-Clara and Valkanov (2003), we examine whether the US presidential

cycles merely capture the variation in expected returns due to business cycle fuctuations in other

countries. If the political variable is a good proxy for the real business cycle, its signifcance should

vanish once we include the factors of real business cycle. We run the following long-horizon predictive

regression:

yit = α + β1 × DEMt−1 + β2 × Xit−j + γ × Country FEi + εit. (3)

Our approach is di˙erent from Santa-Clara and Valkanov (2003) in a few dimensions. First, we

regress the cumulative returns yit throughout the whole tenure of one presidential term, i.e. 4 years.

Secondly, we run a panel regression with time index t and country index i. Although our sample

period is shorter than Santa-Clara and Valkanov (2003), more countries included in our sample

allow us to obtain enough power by running a panel regression. Thirdly, the factors of real business

cycle Xit−j are lagged j = 1, 6 and 12 months prior to the election. The advantage of this design is

to make sure that we have a clean specifcation because after the elections, the real business cycle

may be confounded with the political variable due to new announcements of government policies.

13

We only use the information prior to the election to predict the performance of stock markets during

each presidential term.

We borrow from the literature three widely-used variables for the real business cycle, which

are collected and constructed for each non-US country. They are term spreads, dividend yields and

relative real risk-free rate, which were documented to have predictive power. See Lamont (1998) and

Santa-Clara and Valkanov (2003) for further references. Indeed, it is common that the predictive

power usually grows with the investment horizon. Therefore, our specifcation provides the highest

possibility for the factors of real business cycle to demonstrate its predictability.

Table 9 presents the result. The predictive regressions explain average 45% of the returns.

However, only 2% is contributed by the factors of real business cycle.6 It is aligned with the literature

that the real business cycle can explain the variation in the stock market returns. Their predictability

lasts up to one year prior to the elections. The coeÿcients of Democratic dummy variable are

signifcantly positive and large in this analysis. For instance, the regression (1) demonstrates that

under Democratic presidents, the returns of international stock markets are 55% higher than under

Republican presidents in a 4-year period. In other words, the Democratic returns are 13.8% higher

per annum. Interestingly, after controlling the real business cycle, the return di˙erence slightly

widens rather than evaporates. It is not a surprise as Santa-Clara and Valkanov (2003) had exactly

the same fnding using the US data. Here, we rule out the doubts that the US political cycle is

merely a proxy for the variation of the world-wide business cycles.

5.3 Wealth E˙ect

As the US has a large position of net external debt, the booming of the US stock market should

lead to income growth for the US households. Then such stimulation of the domestic demand would

eventually lead to more net exports and more investments in the other countries. This is the wealth

e˙ect explanation for the return di˙erence in non-US countries.

Following the literature, there are two di˙erent ways to proxy the wealth e˙ect: Lettau and

Ludvigson (2001) used the household net worth and Case, Quigley, and Shiller (2005) used the

market capitalization. Similar to the previous subsection, if the political variable is merely a good

proxy for wealth e˙ect, one should expect to see the political indicator becomes insignifcant after 6We do not include this result in this version as it does not contribute much to our main results.

14

the consideration of wealth e˙ect.

Table 10 presents the results of the panel regression (3) where we replace the control variables

Xit−1 for the wealth variables. Throughout the three panels, the coeÿcients for the household net

worths of non-US countries are insignifcant and are of di˙erent signs. It is the same results as

those wealth variables that only consider fnancial wealth. On the contrary, the coeÿcient for the

US household net worth demonstrates its signifcance in both equity and foreign exchange markets.

These results are similar to the literature in which Lettau and Ludvigson (2001) found strong

predictability of the household wealth on the stock returns.

It is also worth noting that the US data has higher frequency and the non-US data is only

available in the annual frequency. The slow-moving feature of the non-US data may be responsible

for the insignifcance of the wealth variables but these are the closest, comparable variables with the

US data used in Lettau and Ludvigson (2001). In Panel A, comparing the regression (1) with (3),

the return di˙erence drops approximately 1.2% per annum after controlling for the US household

net worth. This di˙erent is relatively small and thus ignorable as the return di˙erence we document

in this paper is more than 10% per annum. The US wealth e˙ect can merely explain fairly small

amount of the di˙erence. In Panel B, we also observe a drop in return di˙erence after controlling

the wealth e˙ect although the descent is quantitatively small.

Surprisingly, Panel C shows that the US wealth is positively correlated with the currency risk

premia. In other words, when the US households have higher net worth, the demand for foreign

currencies increases and therefore the currency risk premia increase. Consequently, by comparing

regression (11) with other regressions (12) – (14) in Panel C, we fnd the di˙erence in currency risk

premia becomes wider after introducing the wealth variables into the regressions. It seems that

the US household net worth dominates the non-US counterparts in the regressions although it does

not explain much the equity return di˙erences. On the contrary, it exacerbates the di˙erence in

currency risk premia. In short, the evidence regarding the household net worth remains inconclusive

in Table 10.

For the robustness check, we employ a di˙erent proxy for the wealth e˙ect, i.e. market capi-

talization, which is perhaps more relavant in the context of the asset returns as we only account

for the wealth in the stock markets. In addition, the advantage of using the market capitalization

lies in its frequency. We are able to have monthly observations but not collinearity concerns. The

15

market capitalization is an interaction of trading price and volume which are rarely correlated with

the equity returns.

Table 11 presents the results in three panels. Similarly to the previous table, the US market

capitalization seems to dominate the regressions over its non-US counterparts and the descent of

the return di˙erence denoted in local currencies is about 1.4% per annum. This value is slightly

higher than using household net worth but still remains quantitatively small as in Case, Quigley,

and Shiller (2005) which they did not fnd the signifcant wealth e˙ect correlated with the stock

market returns.

On the other hand, we fnd the US market capitalization helps to reduce the di˙erence in

currency risk premia at about 0.7% per annum. The value is still quantitatively small but at least

this result is more consistent with the equity returns in Panel A and B. Therefore, we could conclude

that the market capitalization is a cleaner proxy for the wealth e˙ect as it only contains the relavant

information. Only the US market capitalization is useful and it explains about 1/8 of the return

di˙erence we document in this paper. More importantly, the results of the stock returns can be

reconciled with those of the currency risk premia.

6 A Trade Policy Explanation

The subsample analysis from Table 1 to 5 implies that the bipartisan phenomenon of the US politi-

cal cycle have become more robust in the most recent decades. Nevertheless, this phenomenon still

remain non-trivial in the most recent subsample. This fnding is aligned with the proposed expla-

nation of trade channel as international trade began to thrive only in the most recent subsample.

The Net Foreign Asset aggregated across countries also shows the same pattern in Figure 5.

We construct a measure of the net export growth in two steps. First, we calculate the net export

X−Mto the US divided by total trade to the US in each quarter, i.e., NX = Note that import X+M .

(M) and export (X) solely referred to the bilateral trade between US and the non-US countries.

Secondly, we take di˙erence relative to the previous quarter, ΔNXt = NXt − NXt−1 and compare

the net export growth under Democratic presidents relative to Republican presidents in Figure 6.

We fnd that the non-US countries export more to the US under Democratic than Republican

presidents. This result is not a surprise as the Democratic party is generally more international in

16

their policy-making.

Frankel and Romer (1999) discovered that the income growth generated from the booms of

international trade is non-trivial. Thus, the real growth brought by the international trade could be

an explanation for higher stock market returns in foreign countries under Democratic rather than

Republican presidents. We will thus have a closer look into the trade channel.

6.1 Trade Networks

We introduce the network centrality to explain the cross-country variation. Following Richmond (2016),

we construct the Katz centrality measures for each country and plot the centrality time series in

Figure 7. Larger value of centrality implies the country has bilateral trade with many other coun-

tries and has better-connected partners in the global network. We refer such countries as central

countries. Alternatively, countries that have lower centrality value is labeled as peripheral countries.

We run a panel regression as below,

yit = α + β1 × DEMt−1 + β2 × Centit−1 + β3 × DEMt−1 × Centit−1 + γi × Country FEi + εit, (4)

where dependent variable yit can be the average or volatility of the risk premia in stock and currency

markets across di˙erent countries. Centrality measure is lagged one year to align with the dummy

variable DEMt−1 for Democratic presidents. Due to the data availability of the GDP and trade

data, our panel data is constructed at annual frequency. The Democratic indicator is defned as one

when the whole year is under a Democratic president. In the case of election year, the Democratic

indicator also equals to one when the incumbant president is from the Democratic Party. Otherwise,

the indicator is defned as zero. Intuitively, central countries are likely to have larger exposure to the

US political cycles and export more to the US than the peripheral countries, since US is a gigantic

economy and net importer in the world. Together with the wealth e˙ect mentioned in Frankel and

Romer (1999), we expect to fnd certain correlations between centrality rankings and stock market

return under Democratic than Republican presidents.

Table 12 presents the results. Both centrality measure and the interaction term are insignifcant

to explain the international excess returns. Their coeÿcients are indi˙erent from zero. Nevertheless,

the coeÿcient for the Democratic indicator drops after we control the centrality rankings. On

17

average, the return di˙erence narrows down by 2.5% per annum. Approximately, the trade network

help to explains a quarter of the return di˙erence. Although we are still unable to explain the

majority of the return di˙erence, a quarter is a signifcant improvement compared with the case of

controlling wealth e˙ect in Table 11.

In Panel B, the dependent variable is the volatilities of the risk premia in the stock and foreign

exchange markets. The Democratic indicator is signifcantly negative for the volatility of stock

returns but becomes positive for the currency risk premia. The drop in volatility is not signifcant

when the stock returns are denoted in USD. The increase in the currency risk premia sets o˙ the

beneft obtained in the local stock markets when the households invest globally. In summary, we

demonstrate that the bipartisan e˙ect of higher returns and lower volatilities is found in the stock

returns denoted in local currencies but is ambiguous with the USD returns.

6.2 Fund Flow Networks

We have shown the evidence from the trade variables but before presents a direct analysis of the

relationship between trade and asset returns, we would like to study the network of cross-border

capital fows of the mutual funds. As trade fow is often correlated with the capital fow, we expect

to fnd similar network results as in the previous subsection. What is di˙erent from the previous

analysis is the centrality measure. We also examine additional centrality measures such as degree

centrality and betweenness centrality, which are used in Rossi, Blake, Timmermann, Tonks, and

Wermers (2018) to study the fund performance.

Table 13 presents the results in three panels. In Panel A and B, we fnd that the centrality

measure itself cannot explain much the return di˙erence. The centrality measure is positively

correlated with the equity returns. That is to say the central countries generally have higher stock

returns than the peripheral countries although the relationship is sometime insignifcant.

Then, we interact the political variable with centrality measure and fnd the coeÿcient is sig-

nifcantly negative. This fnding is di˙erent from the trade data. Our interpretation is that under

the Democratic presidents, the peripheral countries are faced with larger return di˙erence. In other

words, the mutual fund managers allocate more funds to the peripheral countries and thus their

portfolios are more diversifed when the Democratic presidents are in power. The shift of the US

trade policy from protectionism to internationalism under the Democratic presidents provides more

18

opportunities for the peripheral countries to establish trade partnership with the US and perhaps

other countries. The results with the betweenness centrality are indeterminate.

In Panel C, we fnd the central countries usually have more appreciated currency values against

with the USD. As we knew from the earlier sections that the di˙erence of the currency risk premia

due to the political cycle goes the opposite direction of the centrality measures. Accordingly, the

interaction terms become insignifcant. The political variable remains a dominating explanation for

the di˙erence in the currency risk premia.

6.3 Trade Protectionism

To demonstrate the impact of trade policy on the international asset markets, it is essential to include

a direct, trade-related variable in our regression analysis. We follow the political economy literature,

e.g., Gardner and Kimbrough (1989) and Lohmann and O’Halloran (1994), using tari˙ data to proxy

the level of trade protectionism. The frst tari˙ variable “Average Tari˙” is measured by the custom

duties collected as a percentage of the value of total import in each country. Figure 8 shows the

cross-country observations of the Average Tari˙ and the tari˙ di˙erence between the Democrat and

Republican presidential terms during a long sample period of 1972 and 2017. At a frst glance,

we can easily observe that the Average Tari˙s are systematically higher under the Republican

presidents. The second tari˙ measure “MFN Tari˙” is the weighted-average rate of e˙ective tari˙

across all product applied to any WTO members.7 This is a tari˙ measure widely used in the

trade literature, e.g., Feenstra (1989) and Fajgelbaum, Goldberg, Kennedy, and Khandelwal (2019).

Since the Average Tari˙ has longer time-series but smaller cross-sectional observations than the

MFN Tari˙, we also construct a “Combined Tari˙” variable that consists of both tari˙ series in

order to maximize the information we can obtain from the database. We run the following panel

regression,

yit = α + β1 × DEMt−1 + β2 × Tari˙it−1 + β3 × DEMt−1 × Tari˙it−1 + γi × Fixed E˙ecti + εit, (5)

where the dependent variable is the fnancial returns and the explanatory variables contain one of

the three above-mentioned tari˙ measures. Di˙erent from the previous regression analysis, here we 7The acronym “MFN” is referred to the most favored nation who is usually a WTO member without sanction.

19

treat fxed e˙ect in two di˙erent ways. First, we control country and year fxed e˙ects and the

results are presented in Panel A; Secondly, we control a time-varying country fxed e˙ect and the

results are shown in Panel B of Tables 14 and 15.

Table 14 presents the relations between tari˙ and the international stock returns in a fxed-e˙ect

regression analysis. The dependent variable is the stock excess return in annual terms. In Panel A,

we control the country and year fxed e˙ects. The Average Tari˙ does not show signifcant results

and the adjusted R-squared is low because of missing observations. Compared with the regression

(1), about 40% of data is missing in regressions (2) and (3). In the case of the MFN Tari˙, we did

not have signifcant results in the Panel A either, but the coeÿcient signs are consistent with the

Panel B where we can show both the signifcant result and higher R-squared compared with their

baseline regressions in (1) and (8). The only di˙erence in the Panel B is that we control time-varying

country fxed e˙ect so as to account for the variations in country characteristics. After controlling

for the time-varying country fxed e˙ect in Panel B, the results are more signifcant than in Panel

A and the signs remain consistent between two panels.

The tari˙ is found to be negatively correlated with the stock returns in regressions (6) and (13),

even though the coeÿcient magnitudes of Democrats dummy remain the same. Next, we include

an interaction term of tari˙ and Democrats dummy. There are two fndings worth noting. First,

the impact of tari˙ on the stock market remains negative but signifcant now in regressions (7) and

(14). Secondly, the coeÿcients of interaction terms are positively signifcant while the Democrats

dummy becomes insignifcant. It implies that trade protectionism does play an important role in

this analysis. After controlling for the tari˙ properly, we fnd that the stock returns under the

Democrat presidents are around 7.3% to 11.5% higher than under the Republican presidents. In

this numbers, the main contributor to the return di˙erence is from the interaction term of tari˙ and

Democrat dummy.

The fndings of the stock returns denoted in USD are quantitatively similar to Table 14 so we do

not report in the main body of this paper for the sake of brevity. In the following, we are going to

show the infuence of trade protectionism on the currency market in Table 15. It is not surprising

that we fnd even stronger results when the dependent variable is currency risk premia.

In Table 15, we investigate the infuence of trade protectionism on the currency risk premia.

Similar to the previous table, the Average Tari˙ does not yield any signifcant result even in the

20

context of foreign exchange returns. This evidence provide some supports for the explanation of

missing observations. On the contrary, the MFN tari˙ generates signifcant results. If we exclude

the interaction term of tari˙ and Democrat dummy as in regressions (4), (11), among others,

the coeÿcients of tari˙ are counter-intuitively negative. However, in most cases, they are also

insignifcant. We suspect this might be due to missing variables. Therefore, we propose to focus on

the regressions that contain the interaction terms.

In regressions (7) and (14) of Table 15, we fnd that the tari˙ coeÿcient is positive and mildly

signifcant. Therefore, higher tari˙ is correlated with the USD depreciation which is usually an

unfavorable condition for the international trade. It is interesting to recall that the USD depreciation

is also correlated with the period of Republican presidential terms. On the other hand, the coeÿcient

of the interaction term is signifcantly negative, aligned with the sign of Democrat dummy in the

baseline regressions (1) and (8). As in the previous table, we also observe that the Democrat dummy

becomes insignifcant. It demonstrates again the importance of including trade protectionism to

explain the bipartisan di˙erence in currency market. From our analysis, it seems like there are three

elements clustered: higher tari˙, USD depreciation and the Republican presidential terms. We are

keen to better understand the causal relations among these elements.

7 Concluding Remark

We document the correlation between the US political cycles and the risk premia of the stock and

foreign exchange markets worldwide. When the US presidents are from the Democratic Party, the

stock market outperforms those periods under the Republican presidents. The return di˙erence

is on average more than 10% per annum among the 23 countries and it persists regardless of the

currency denomination. In the foreign exchange market, we also fnd more appreciated USD and

lower currency risk premia under the Democrat presidents then under the Republican presidents.

We rule out a couple of possible explanations for the return di˙erence in the stock and foreign

exchange markets, including the political cycles of non-US countries, the (fnancial) wealth e˙ect

and the real business cycle. Controlling these variables helps to explain solely a small fraction of

the return di˙erence. The majority of the sizable return di˙erence remains a puzzle.

Motivated by the abundant amount of anecdote evidences, we further analyze the channel of

21

trade in a network setting. We employ bilateral trade and mutual fund fows to construct global

networks. Both trade and capital fows are found to help with the explanation of return di˙erence.

Hence, we argue that the US political cycle spillovers to other countries probably via the trade

channel. We then investigate the infuence of trade protectionism on the performance of international

fnancial markets and demonstrate that the return di˙erence is mitigated after controlling for the

tari˙ variations.

Future works can extend our results in a number of ways. For instance, investigating some

country-specifc characteristics of political institution would allow us to better understand why

the foreign political cycle does not seem to generate the return di˙erence in the fnancial markets.

Considering the role of Congress in imposing a trade policy could help us understand the interaction

between political institutions, i.e., Congress and the President, and the related implications for

economic growth and asset returns.

22

References

Alesina, A., and N. Roubini, 1992, “Political cycles in OECD economies,” Review of Economic Studies, 59(4), 663–688.

Bachman, D., 1992, “The e˙ect of political risk on the forward exchange bias: the case of elections,” Journal of International Money and Finance, 11(2), 208–219.

Blinder, A. S., and M. W. Watson, 2016, “Presidents and the US economy: An econometric explo-ration,” American Economic Review, 106(4), 1015–45.

Bonacich, P., and P. Lloyd, 2001, “Eigenvector-like measures of centrality for asymmetric relations,” Social Networks, 23(3), 191–201.

Brogaard, J., L. Dai, P. T. Ngo, and B. Zhang, 2019, “Global political uncertainty and asset prices,” .

Case, K. E., J. M. Quigley, and R. J. Shiller, 2005, “Comparing wealth e˙ects: the stock market versus the housing market,” Advances in Macroeconomics, 5(1).

Fajgelbaum, P. D., P. K. Goldberg, P. J. Kennedy, and A. K. Khandelwal, 2019, “The return to protectionism,” working paper, NBER Working Paper No. 25638.

Feenstra, R. C., 1989, “Symmetric pass-through of tari˙s and exchange rates under imperfect com-petition: An empirical test,” Journal of International Economics, 27(1-2), 25–45.

Fetzer, T., and C. Schwarz, 2019, “Tari˙s and politics: evidence from Trump’s trade wars,” working paper.

Frankel, J. A., and D. H. Romer, 1999, “Does trade cause growth?,” American Economic Review, 89(3), 379–399.

Gardner, G. W., and K. P. Kimbrough, 1989, “The behavior of US tari˙ rates,” American Economic Review, 79(1), 211–218.

Goto, J., T. Matsumoto, and J. Yamasaki, 2018, “Political statement cycle,” .

Harrison, A., and H. Tang, 2005, “Trade Liberalization: Why So Much Controversy?,” Economic Growth in the 1990s: Learning from a Decade of Reform, pp. 131–56.

Hibbs, D. A., 1977, “Political parties and macroeconomic policy,” American political science review, 71(4), 1467–1487.

Jackson, M. O., 2010, Social and economic networks. Princeton University Press.

Kelly, B., L. Pástor, and P. Veronesi, 2016, “The price of political uncertainty: Theory and evidence from the option market,” Journal of Finance, 71(5), 2417–2480.

Lamont, O., 1998, “Earnings and expected returns,” Journal of Finance, 53(5), 1563–1587.

Lettau, M., and S. Ludvigson, 2001, “Consumption, aggregate wealth, and expected stock returns,” Journal of Finance, 56(3), 815–849.

Liu, Y., and I. Shaliastovich, 2017, “Government Policy Approval and Exchange Rates,” .

23

Lobo, B. J., and D. Tufte, 1998, “Exchange rate volatility: Does politics matter?,” Journal of Macroeconomics, 20(2), 351–365.

Lohmann, S., and S. O’Halloran, 1994, “Divided government and U.S. trade policy: theory and evidence,” International Organization, 48(4), 595–632.

Nordhaus, W. D., 1975, “The political business cycle,” Review of Economic Studies, 42(2), 169–190.

Pástor, L., and P. Veronesi, 2013, “Political uncertainty and risk premia,” Journal of Financial Economics, 110(3), 520–545.

Pastor, L., and P. Veronesi, 2019, “Political cycles and stock returns,” .

Richmond, R. J., 2016, “Trade network centrality and currency risk premia,” working paper, Avail-able at SSRN: https://ssrn.com/abstract=2607735.

Rossi, A. G., D. Blake, A. Timmermann, I. Tonks, and R. Wermers, 2018, “Network centrality and delegated investment performance,” Journal of Financial Economics, 128(1), 183–206.

Santa-Clara, P., and R. Valkanov, 2003, “The presidential puzzle: Political cycles and the stock market,” The Journal of Finance, 58(5), 1841–1872.

Scartascini, C., C. Cruz, and P. Keefer, 2018, “The Database of Political Institutions 2017 (DPI2017),” working paper.

Verdelhan, A., 2018, “The share of systematic variation in bilateral exchange rates,” The Journal of Finance, 73(1), 375–418.

24

AU

SA

UT

BE

LC

AN

HK

GD

NK

FIN

FR

AD

EU

IRL

ISR

ITA

JPN

NLD

NZ

LN

OR

PR

TS

GP

ES

PS

WE

CH

EG

BR

US

AS

imple

Weighted

Pooled

-10

-5

0

5

10

15

20

25

30

% p

er a

nnum

Stock Return Difference (DEM-REP) in Local Currencies

USD Stock ReturnsCurrency Excess Return

Figure 1. Di˙erence of excess returns under Democratic and Republican presidents The sample period is from 1929 to 2017 and the excess returns are denoted in US Dollars. Whole sample statistics is presented in black dashed line. The three subsample are of equal length. The statistics of each subsample is plotted in blue, red, and green bars, respectively.

25

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00

1.25

US

A

GB

R

JP

N

FR

A

GE

R

AU

S

ES

P

SW

E

NL

D

CA

N

AU

T

BE

L

DN

K

FIN

IRL

ITA

NZ

L

NO

R

PR

T

SG

P

CH

E

HK

G

ISR

B. Sharpe Ratio of Equity Return in Local Currency

Diff Dem Rep

Figure 2. Sharpe ratio di˙erential between under Democratic and Republican presidents The sample period is the same as Figure 1 and we only report the whole sample results. The blue dashed line is the Sharpe ratio under Democratic presidents and the red solid line is under Republican presidents.

26

0 10 20-0.1

0

0.1

0.2

0.3

Mea

n

-0.2

-0.1

0

0.1

0.2

Whole sample

0 10 20-0.1

0

0.1

0.2

0.3

-0.1

0

0.1

0.2

0.3

Sub I(1929:01-1958:03)

0 10 20-0.2

-0.1

0

0.1

0.2

0.3

-1

-0.5

0

0.5

1

Sub II(1958:04-1987:09)

0 10 20-0.1

0

0.1

0.2

0.3

0.4

-0.1

0

0.1

0.2

0.3

Sub III(1987:1-2017:08)

0 10 200

0.1

0.2

0.3

VaR

-1

-0.8

-0.6

-0.4

0 10 20-0.2

-0.1

0

0.1

0.2

0.3

-1

-0.8

-0.6

-0.4

-0.2

0 10 20-0.1

0

0.1

0.2

0.3

0.4

-1

-0.8

-0.6

-0.4

-0.2

0 10 20-0.1

0

0.1

0.2

0.3

-0.9

-0.8

-0.7

-0.6

-0.5

-0.4

0 10 20-0.1

0

0.1

0.2

0.3

ES

-1

-0.9

-0.8

-0.7

-0.6

0 10 20-0.2

-0.1

0

0.1

0.2

0.3

-1

-0.8

-0.6

-0.4

-0.2

0 10 20-0.2

-0.1

0

0.1

0.2

0.3

-1

-0.8

-0.6

-0.4

-0.2

0 10 200

0.1

0.2

0.3

-1

-0.9

-0.8

-0.7

-0.6

Figure 3. Risk-based di˙erence of stock returns under Democratic and Republican presidents The sample period is from 1929 to 2017 and the three subsamples are of equal length. The frst column is the whole sample results and the second to the forth columns are the following three sub-sample periods: 1929:01–1958:03, 1958:04–1987:09, and 1987:10–2017:08. The upper row presents the average di˙erential which is the same to Figure 1. The middle row presents the Value at Risk (VaR) at 5% which are essentially the 5% value from the distribution under Democratic and Repub-lican presidents. The bottom row presents the Expected Shortfall (ES) which are the average of all the values below 5% threshold from the distribution under Democratic and Republican presidents.

27

0 10 20-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

Std

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

Whole sample

0 10 20-0.1

-0.05

0

0.05

0.1

0.15

0.2

0.25

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

Sub I(1929:01-1958:03)

0 10 20-0.15

-0.1

-0.05

0

0.05

0

0.1

0.2

0.3

0.4

0.5

0.6

Sub II(1958:04-1987:09)

0 10 20-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

Sub III(1987:1-2017:08)

0 10 20-0.6

-0.4

-0.2

0

0.2

0.4

0.6

Max

dd

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 10 20-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0

0.2

0.4

0.6

0.8

1

0 10 20-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0

0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0 10 20-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Figure 4. Di˙erence of stock market volatilities under Democratic and Republican presidents The sample period is from 1929 to 2017 and the three subsamples are of equal length. The frst column is the whole sample results and the second to the forth columns are the following three sub-sample periods: 1929:01–1958:03, 1958:04–1987:09, and 1987:10–2017:08. The upper row presents the di˙erence in standard deviation while the lower row reports the di˙erence in maximum draw-down, which is a downside risk measure.

28

1960 1970 1980 1990 2000 2010-1

-0.5

0

0.5

1

1.5

US

D (

Tril

lions

)

Agg. Change of Net Foreign Position

1960 1970 1980 1990 2000 20100

0.5

1

1.5

2

US

D (

Tril

lions

)

Agg. Change of Net Foreign Position (abs)

Figure 5. Net foreign assets The fgure shows time series of the net foreign asset aggregated across 23 countries. The upper subfgure considers the sign of the position where positive position in one country might be canceled by negative position in another country. The bottom subfgure reports the position of net foreign assets. The time-series sums up the absolute values of net foreign assets across countries.

29

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

GB

R

AU

S

AU

T

BE

L

CA

N

CH

E

DN

K

ES

P

FIN

FR

A

GE

R

HK

G

IRL

ISR

ITA

JP

N

NLD

NO

R

NZ

L

PR

T

SW

E

SG

P

%

Growth of Net Export

Diff Dem Rep

Figure 6. Growth of Net Export The net export is solely referred to the bilateral trade between the US and any foreign country. The sample period is from 1980 to 2016. The net export is the export minus import to the US standardized by the sum of export and import to the US. The growth of net export is the simple di˙erence relative to the previous quarter.

30

0 20 40

5101520

Australia

0 20 40

5101520

Austria

0 20 40

5101520

Belgium

0 20 40

5101520

Canada

0 20 40

5101520

HongKong

0 20 40

5101520

Denmark

0 20 40

5101520

Finland

0 20 40

5101520

France

0 20 40

5101520

Germany

0 20 40

5101520

Ireland

0 20 40

5101520

Israel

0 20 40

5101520

Italy

0 20 40

5101520

Japan

0 20 40

5101520

Natherlands

0 20 40

5101520

NewZealand

0 20 40

5101520

Norway

0 20 40

5101520

Portugal

0 20 40

5101520

Singapore

0 20 40

5101520

Spain

0 20 40

5101520

Sweden

0 20 40

5101520

Switzerland

0 20 40

5101520

UnitedKingdom

0 20 40

5101520

UnitedStates

Figure 7. Trade network centrality ranking This fgure presents centrality ranking which is constructed by the bilateral trade data. The sample period is from 1980 to 2016. The bilateral data is at annual frequency. The network centrality in each year is rescaled between 1 and 23. Central countries have high centrality and peripheral countries have low centrality.

31

AU

S

AU

T

BE

L

CA

N

CH

E

DE

U

DN

K

ES

P

FIN

FR

A

GB

R

HK

G

IRL

ISR

ITA

JPN

NLD

NO

R

NZ

L

PR

T

SG

P

SW

E

US

A

-7

-6

-5

-4

-3

-2

-1

0

1

0

2

4

6

8

10

12

14

16

Duties/Tax Revenue (%)Sample period: 1972-2017

Figure 8. Average Tari˙ This fgure presents the Average Tari˙ which is calculated by Custom Duties as a percentage of Total Import within each country. The sample period is from 1972 to 2017. The data is at annual frequency. The blue solid (red dashed) line represents the Average Tari˙ under the Democrat (Republican) presidents. The values are referred to the right axis. The yellow bars represent the di˙erence between Democrat and Republican presidents. The values are referred to the left axis.

32

1927:02-1956:12 1957:01-1986:12 1987:01-2018:07

Series Returns Mean Std AR Mean Std AR Mean Std AR

VWR

EWR

LOC USD FXR Corr(L,F)

LOC USD FXR Corr(L,F)

6.53 7.61 1.08 5.58

7.36 7.46 0.10 1.34

9.35 10.15 3.51

17.47 17.62 2.06

0.31 0.24 0.01

0.15 0.14 0.00

3.14 5.86 2.72 -1.84

3.30 4.01 0.71 3.27

9.46 9.71 2.38

10.43 10.62 1.65

0.27 0.29 0.30

0.15 0.17

-0.02

3.48 4.41 0.94 5.42

3.46 3.55 0.08 2.71

15.07 16.90 6.86

14.46 15.40 4.94

0.21 0.16 0.06

0.16 0.11 0.05

Table 1. Summary Statistics: Unconditional Average Returns The table reports the sample average (Mean), the standard deviation (Std), and the autoregressive coeÿcient (AR) of equity excess returns denoted in local currency (LOC) or US Dollars (USD), and currency excess return (FXR). Panel A presents the unconditional returns. We use value-weighted return (VWR) and equal-weighted return (EWR) to aggregate the average returns of 23 countries. In the ‘Return’ column, we also report the correlation between FXR and LOC. All numbers are presented in percentage points and all returns are computed in logarithmic form and expressed in annualized terms.

33

Party

DD-RD

DD

RD

Panel A: Average Returns Conditional on Political Party of US Presidents

1927:02-1956:12 1957:01-1986:12 1987:01-2018:07

Returns Mean Std AR Mean Std AR Mean Std

LOC 8.42 -6.53 -0.10 1.17 -2.69 0.00 11.39 -2.10 USD 7.89 -6.26 -0.10 0.22 -2.94 0.00 7.87 -1.11 FXR -0.51 2.45 0.01 -0.95 -1.89 0.07 -3.52 -0.18 Corr(L,F) 12.68 6.23 20.41

LOC 10.15 15.03 0.15 4.00 8.75 0.24 9.08 13.31 USD 10.08 15.29 0.15 4.14 8.77 0.25 7.43 14.82 FXR -0.07 2.51 0.98 0.14 0.23 0.97 -1.65 4.83 Corr(L,F) 2.40 9.74 14.93

LOC 1.73 21.56 0.24 2.83 11.44 0.24 -2.30 15.41 USD 2.19 21.55 0.24 3.92 11.71 0.25 -0.44 15.94 FXR 0.44 0.06 0.97 1.09 2.12 0.89 1.87 5.01 Corr(L,F) -10.28 3.51 -5.48

AR

-0.03 -0.02 0.00

0.10 0.01 0.08

0.12 0.03 0.08

Months

T

Panel B: Number of Observations

1927:02-1956:12 1957:01-1986:12

Total DD RD Total DD

359 240 119 360 144

RD

216

1987:01-2018:07

Total DD

379 192

RD

187

Table 2. Summary Statistics of Average Returns by Presidential Cycle The table reports the average returns during Republican (RD) and Democratic (DD) presidential terms. All rates are presented in annualized percentage points. Panel A reports that the summary statistics is the same as Table 1. The ‘DD-RD’ row reports the di˙erence taken between the DD and RD statistics. Panel B reports the number of months T in each subsample and during RD and DD presidential terms.

34

1927:02-1956:12 1957:01-1986:12 1987:01-2018:07

DD RD Di˙ DD RD Di˙ DD RD Di˙

Pooled 9.64 2.05 7.59 *** 3.30 2.74 0.55 9.60 -2.71 12.31 *** (0.45) (0.62) (0.76) (0.38) (0.51) (0.63) (0.27) (0.39) (0.47)

EWR 9.51 0.52 8.98 * 2.97 3.25 -0.28 9.60 -2.81 12.41 ** (0.55) (1.51) (1.61) (0.68) (0.97) (1.19) (1.00) (1.44) (1.76)

VWR 10.15 1.73 8.42 4.00 2.83 1.17 9.08 -2.30 11.39 ** (0.96) (2.12) (2.33) (0.75) (0.92) (1.18) (0.90) (1.31) (1.59)

AUS 8.16 5.51 2.65 12.41 4.04 8.37 6.59 -1.09 7.68 (0.73) (1.49) (1.66) (1.22) (1.45) (1.89) (0.90) (1.44) (1.70)

AUT -6.73 3.69 -10.42 ** 2.66 3.47 -0.81 (0.78) (1.39) (1.60) (1.45) (2.23) (2.66)

BEL 13.52 11.40 2.13 -1.73 3.25 -4.99 10.13 -2.00 12.13 ** (2.94) (1.24) (3.19) (0.82) (1.08) (1.35) (1.07) (1.57) (1.91)

CAN 9.30 14.34 -5.04 10.39 -2.77 13.16 ** 8.80 -1.37 10.17 ** (1.07) (1.69) (2.00) (1.09) (1.21) (1.63) (1.08) (1.23) (1.64)

HKG 32.50 9.10 23.41 9.19 7.10 2.09 (4.26) (3.75) (5.68) (1.92) (1.99) (2.77)

DNK -9.32 1.51 -10.83 14.47 -0.23 14.70 ** (2.17) (1.69) (2.75) (1.18) (1.54) (1.94)

FIN 10.00 4.25 5.75 -2.70 10.51 -13.21 *** 20.61 -7.80 28.41 *** (1.82) (1.64) (2.45) (0.79) (1.22) (1.45) (1.96) (2.14) (2.90)

FRA 9.54 3.94 5.60 -1.51 0.76 -2.27 7.95 -6.17 14.12 ** (1.86) (1.89) (2.65) (1.24) (1.51) (1.95) (1.24) (1.58) (2.01)

DEU 12.65 -9.05 21.70 ** -4.46 7.54 -12.00 ** 8.89 -7.00 15.90 ** (1.61) (2.26) (2.78) (1.18) (1.27) (1.73) (1.28) (1.77) (2.19)

IRL 13.66 -5.82 19.48 ** (1.36) (1.79) (2.25)

ISR 2.79 6.45 -3.66 (1.49) (1.86) (2.38)

ITA 17.38 -0.98 18.36 ** 1.44 2.60 -1.17 8.48 -9.02 17.50 ** (2.37) (1.95) (3.07) (1.76) (1.81) (2.52) (1.57) (1.49) (2.17)

JPN 10.33 -2.19 12.53 0.86 9.18 -8.32 * 2.95 -4.48 7.43 (2.07) (2.02) (2.89) (1.32) (1.12) (1.73) (1.47) (1.70) (2.25)

NLD 4.80 -2.79 7.60 0.09 0.69 -0.60 10.94 -7.30 18.23 *** (1.29) (2.19) (2.55) (1.16) (1.26) (1.71) (1.11) (1.68) (2.01)

NZL 6.70 -8.22 14.92 ** (0.98) (1.72) (1.98)

NOR 12.22 2.90 9.32 11.58 -1.89 13.46 * (4.61) (2.04) (5.04) (1.32) (2.30) (2.65)

PRT 6.35 -10.18 16.53 ** (1.58) (1.80) (2.40)

SGP 19.20 2.66 16.55 6.61 2.87 3.73 (2.49) (2.89) (3.82) (1.65) (2.03) (2.61)

ESP 5.20 23.72 -18.52 *** -0.37 4.98 -5.34 10.13 -2.82 12.95 * (1.42) (1.41) (2.01) (1.21) (1.32) (1.79) (1.59) (1.58) (2.24)

SWE 7.94 -0.83 8.77 * 1.72 11.00 -9.28 * 17.78 -5.64 23.42 *** (0.79) (1.70) (1.87) (1.02) (1.32) (1.67) (1.36) (1.93) (2.36)

CHE 4.90 3.29 1.60 11.85 -1.77 13.61 ** (1.51) (1.29) (1.99) (1.10) (1.43) (1.80)

GBR 6.73 1.85 4.88 9.46 3.90 5.57 9.36 -1.57 10.93 ** (0.91) (1.27) (1.56) (1.40) (1.66) (2.17) (0.85) (1.33) (1.58)

USA 11.19 2.46 8.73 5.84 1.74 4.11 12.30 1.34 10.96 ** (1.42) (2.90) (3.23) (1.06) (1.03) (1.48) (0.91) (1.22) (1.52)

Table 3. Local-currency Return Di˙erence by Country The table reports mean excess returns during Democratic (DD) and Republican (RD) presiden-tial terms. The upper panel presents the estimate of equal-weighted, value-weighted, and pooled portfolios while the lower panel presents it by country. The standard errors below the coeÿcients are obtained from the Newey-West heteroskedasticity and serial-correlation robust statistics. The asterisk indicates signifcance tests on the null hypothesis that stock returns are the same during DD and RD presidential terms.

35

1927:02-1956:12 1957:01-1986:12 1987:01-2018:07

DD RD Di˙ DD RD Di˙ DD RD Di˙

Pooled 10.43 3.62 6.81 *** 4.63 7.83 -3.20 * 7.72 1.08 6.64 *** (0.49) (0.57) (0.75) (0.38) (0.45) (0.59) (0.30) (0.40) (0.50)

EWR 10.41 1.97 8.45 * 3.66 7.33 -3.67 7.72 1.02 6.69 (0.62) (1.50) (1.62) (0.67) (1.06) (1.25) (1.13) (1.53) (1.90)

VWR 10.08 2.19 7.89 4.14 3.92 0.22 7.43 -0.44 7.87 * (0.97) (2.12) (2.33) (0.75) (0.99) (1.24) (0.97) (1.32) (1.64)

AUS 9.07 7.10 1.98 12.90 4.27 8.63 7.46 3.76 3.70 (0.74) (1.26) (1.46) (1.22) (1.52) (1.95) (1.42) (1.88) (2.36)

AUT -7.08 3.88 -10.96 ** -0.29 7.01 -7.30 (0.71) (1.43) (1.60) (1.62) (2.46) (2.94)

BEL 16.72 13.70 3.02 -0.19 4.96 -5.14 7.37 1.03 6.34 (2.96) (1.25) (3.22) (0.84) (1.10) (1.38) (1.24) (1.77) (2.16)

CAN 9.50 14.42 -4.92 11.26 -1.86 13.12 ** 6.88 2.22 4.66 (1.07) (1.70) (2.01) (1.09) (1.20) (1.63) (1.40) (1.51) (2.06)

HKG 29.61 8.04 21.57 9.56 6.56 2.99 (4.31) (3.75) (5.71) (1.91) (1.99) (2.76)

DNK -1.72 10.13 -11.85 11.22 6.24 4.98 (2.10) (1.61) (2.65) (1.19) (1.70) (2.07)

FIN 14.24 8.14 6.10 0.31 13.74 -13.43 *** 18.24 -2.62 20.86 ** (1.82) (1.64) (2.45) (0.77) (1.21) (1.44) (2.00) (2.24) (3.00)

FRA 11.20 4.28 6.91 -0.99 4.63 -5.62 4.43 -1.17 5.60 (1.86) (1.89) (2.65) (1.25) (1.62) (2.04) (1.38) (1.60) (2.11)

DEU 3.34 -6.43 9.78 -5.91 8.21 -14.12 ** 4.68 -2.77 7.45 (2.79) (2.47) (3.73) (1.18) (1.43) (1.85) (1.38) (1.78) (2.25)

IRL 9.98 -1.12 11.10 * (1.39) (1.90) (2.36)

ISR 6.63 4.35 2.29 14.75 38.53 -23.78 ** 4.65 12.04 -7.39 (2.60) (2.97) (3.94) (2.49) (2.90) (3.82) (1.69) (1.98) (2.61)

ITA 21.37 1.75 19.62 ** 3.25 7.47 -4.21 4.97 -4.09 9.07 (2.37) (1.91) (3.04) (1.77) (1.93) (2.62) (1.77) (1.61) (2.40)

JPN 13.48 0.50 12.98 1.15 11.13 -9.97 * 1.42 -4.25 5.67 (2.08) (2.04) (2.91) (1.31) (1.24) (1.81) (1.57) (1.67) (2.29)

NLD 5.68 -2.90 8.57 -0.67 1.41 -2.07 6.82 -2.70 9.53 * (1.30) (2.18) (2.54) (1.16) (1.32) (1.76) (1.16) (1.72) (2.08)

NZL 8.69 -2.16 10.85 * (1.47) (1.82) (2.34)

NOR 10.40 4.16 6.23 8.38 3.59 4.79 (4.58) (2.08) (5.03) (1.61) (2.44) (2.92)

PRT 5.59 -4.70 10.28 (1.77) (2.04) (2.70)

SGP 16.85 0.72 16.13 5.25 1.20 4.05 (2.49) (2.92) (3.84) (1.65) (2.02) (2.61)

ESP 8.54 25.79 -17.25 ** 2.76 7.29 -4.53 8.26 2.26 6.01 (1.43) (1.41) (2.00) (1.12) (1.30) (1.72) (1.77) (1.73) (2.47)

SWE 9.64 -0.62 10.26 * 2.68 13.93 -11.25 ** 13.72 -0.75 14.47 ** (0.79) (1.71) (1.88) (1.02) (1.41) (1.74) (1.51) (2.02) (2.52)

CHE 0.97 2.05 -1.07 10.07 0.74 9.32 * (1.51) (1.42) (2.07) (1.15) (1.30) (1.74)

GBR 6.99 2.95 4.04 11.45 6.15 5.30 7.82 2.61 5.22 (0.92) (1.26) (1.56) (1.42) (1.70) (2.21) (1.04) (1.42) (1.76)

USA 11.19 2.46 8.73 5.84 1.74 4.11 12.30 1.34 10.96 ** (1.42) (2.90) (3.23) (1.06) (1.03) (1.48) (0.91) (1.22) (1.52)

Table 4. USD Return Di˙erence by Country The table reports mean excess returns denoted in USD during Democratic (DD) and Republican (RD) presidential terms. The upper panel presents the estimate of equal-weighted, value-weighted, and pooled portfolios while the lower panel presents it by country. The standard errors below the coeÿcients are obtained from the Newey-West heteroskedasticity and serial-correlation robust statistics. The asterisk indicates signifcance tests on the null hypothesis that stock returns are the same during DD and RD presidential terms.

36

1927:02-1956:12 1957:01-1986:12 1987:01-2018:07

DD RD Di˙ DD RD Di˙ DD RD Di˙

Pooled 0.93 1.58 -0.65 0.72 4.77 -4.05 * -1.97 3.97 -5.94 *** (0.32) (0.05) (0.32) (0.10) (0.76) (0.77) (0.15) (0.15) (0.21)

EWR 0.91 1.44 -0.53 0.69 4.08 -3.39 *** -1.88 3.83 -5.71 ** (0.28) (0.03) (0.28) (0.06) (0.33) (0.34) (0.49) (0.55) (0.74)

VWR -0.07 0.44 -0.51 0.14 1.09 -0.95 -1.65 1.87 -3.52 ** (0.16) (0.01) (0.16) (0.04) (0.21) (0.21) (0.34) (0.40) (0.52)

AUS 0.92 0.79 0.13 0.49 0.23 0.26 0.87 4.85 -3.98 (0.13) (0.31) (0.33) (0.08) (0.37) (0.38) (0.81) (0.86) (1.18)

AUT -0.36 0.19 -0.55 -2.95 3.54 -6.49 ** (0.37) (0.17) (0.41) (0.70) (0.62) (0.93)

BEL 3.20 2.31 0.89 *** 1.55 1.70 -0.15 -2.77 3.02 -5.79 ** (0.06) (0.04) (0.07) (0.10) (0.08) (0.13) (0.70) (0.61) (0.93)

CAN 0.20 0.08 0.12 0.86 0.91 -0.04 -1.92 3.59 -5.51 ** (0.09) (0.02) (0.09) (0.05) (0.10) (0.11) (0.49) (0.53) (0.73)

HKG -2.89 -1.06 -1.84 ** 0.37 -0.54 0.91 *** (0.20) (0.12) (0.23) (0.04) (0.06) (0.07)

DNK 7.60 8.62 -1.02 -3.25 6.47 -9.72 ** (0.25) (0.68) (0.73) (0.69) (0.86) (1.11)

FIN 4.23 3.89 0.35 3.00 3.22 -0.22 -2.37 5.18 -7.55 ** (0.07) (0.10) (0.13) (0.13) (0.20) (0.24) (0.72) (0.62) (0.95)

FRA 1.66 0.35 1.31 ** 0.52 3.87 -3.35 ** -3.52 5.00 -8.52 ** (0.09) (0.13) (0.16) (0.11) (0.54) (0.55) (0.68) (0.84) (1.08)

DEU -9.31 2.61 -11.92 -1.45 0.67 -2.12 -4.21 4.24 -8.44 ** (3.33) (0.04) (3.33) (0.11) (0.58) (0.59) (0.69) (0.87) (1.11)

IRL -3.68 4.71 -8.38 ** (0.67) (0.89) (1.11)

ISR 1.86 5.59 -3.73 * (0.50) (0.48) (0.70)

ITA 3.99 2.73 1.26 * 1.82 4.86 -3.05 -3.51 4.93 -8.43 ** (0.18) (0.16) (0.24) (0.39) (0.60) (0.72) (0.69) (0.86) (1.10)

JPN 3.15 2.70 0.45 0.29 1.94 -1.65 -1.53 0.23 -1.76 (0.16) (0.45) (0.48) (0.40) (0.65) (0.76) (0.83) (0.76) (1.12)

NLD 0.87 -0.10 0.98 *** -0.76 0.71 -1.47 -4.12 4.59 -8.71 ** (0.07) (0.07) (0.09) (0.04) (0.43) (0.43) (0.69) (0.87) (1.11)

NZL 1.98 6.06 -4.08 (0.82) (0.83) (1.17)

NOR -1.82 1.27 -3.09 ** -3.20 5.48 -8.67 ** (0.17) (0.49) (0.51) (0.70) (0.88) (1.13)

PRT -0.76 5.48 -6.25 ** (0.70) (0.66) (0.96)

SGP -2.35 -1.94 -0.42 -1.35 -1.67 0.32 (0.15) (0.11) (0.19) (0.27) (0.16) (0.31)

ESP 3.34 2.07 1.27 *** 3.13 2.32 0.81 -1.87 5.08 -6.95 ** (0.07) (0.05) (0.08) (0.46) (0.27) (0.54) (0.69) (0.62) (0.93)

SWE 1.69 0.21 1.48 *** 0.96 2.93 -1.97 -4.06 4.89 -8.95 ** (0.06) (0.05) (0.08) (0.15) (0.53) (0.55) (0.82) (0.84) (1.17)

CHE -3.92 -1.25 -2.68 -1.78 2.51 -4.29 (0.40) (0.90) (0.98) (0.74) (0.87) (1.14)

GBR 0.27 1.10 -0.83 *** 1.98 2.25 -0.27 -1.54 4.18 -5.71 * (0.06) (0.07) (0.09) (0.08) (0.47) (0.47) (0.60) (0.81) (1.01)

Table 5. Di˙erence of Currency Risk Premia by Country The table reports mean excess returns on currencies for each country during Democratic (DD) and Republican (RD) presidential terms. The upper panel presents the estimate of equal-weighted, value-weighted, and pooled portfolios while the lower panel presents it by country. The standard errors below the coeÿcients are obtained from the Newey-West heteroskedasticity and serial-correlation robust statistics. The asterisk indicates signifcance tests on the null hypothesis that stock returns are the same during DD and RD presidential terms.

37

G7 Leftist Party Election cycle

Country Party Winners Incumbent Average Std Name counts period (year) (month)

USA Democrat 52.0% 53.9% 4.0 0.1 CAN Grits 63.0% 65.6% 3.4 16.9 DEU SPD 16.7% 15.6% 3.8 5.6 FRA PS 29.4% 31.0% 5.8 21.9 GBR Labour 43.5% 35.9% 4.1 21.6 ITA Olive Tree 15.8% 16.7% 4.0 14.8 JPN Minshuto 18.8% 20.2% 2.8 12.8

Table 6. Summary statistics of local political cycle The sample period is from 1929 to 2017 in order to aligned with Figure 1. In this Table we only report the G7 country. Since there is no clear pattern, we argue it is not necessary to extend our sample to all 23 countries. The election under dictators in Germany (DEU) and Italy (ITA) are excluded from the above statistics. Other country acronyms refer to Canda (CAN), France (FRA), United Kingdom (GBR), and Japan (Japan). The party acronyms refer to Social Democratic Party of Germany (SPD), Parti socialiste (PS), Parti libéral du Canada (Grits).

38

1927:02-1956:12 1957:01-1986:12 1987:01-2018:07

Election Left Right Di˙ Left Right Di˙ Left Right Di˙

Panel A: Pooled estimate

USA

GBR

FRA

DEU

JPN

CAN

ITA

9.64 (0.45) 1.43 (0.87) 10.31 (0.76) -21.84 (1.02) 1.53 (0.62) 9.07 (0.39)

2.05 (0.62) 9.54 (0.36) 6.13 (0.42) 11.06 (0.38) 9.95 (0.45) -2.80 (0.97)

7.59 (0.76) -8.10 (0.94) 4.18 (0.87) -32.90 (1.09) -8.42 (0.77) 11.87 (1.04)

***

**

*

***

***

***

3.30 (0.38) -2.72 (0.44) 4.08 (1.09) -7.09 (0.94) -3.16 (1.09) 0.64 (0.40)

2.74 (0.51) 6.11 (0.47) 2.61 (0.32) 4.69 (0.36) 3.16 (0.35) 9.06 (0.58)

0.55 (0.63) -8.82 (0.64) 1.47 (1.13) -11.79 (1.00) -6.33 (1.14) -8.42 (0.70)

***

***

*

***

9.60 (0.27) 2.46 (0.45) 3.71 (0.31) 3.99 (0.54) 5.42 (0.35) 6.74 (0.32) 7.43 (0.33)

-2.71 (0.39) 4.30 (0.27) 3.40 (0.37) 3.41 (0.27) 3.05 (0.29) 0.60 (0.36) 1.10 (0.33)

12.31 (0.47) -1.84 (0.52) 0.30 (0.48) 0.58 (0.61) 2.37 (0.46) 6.14 (0.48) 6.33 (0.46)

***

*

***

***

Panel B: G7 local stock returns

USA

GBR

FRA

DEU

JPN

CAN

ITA

11.19 (1.42) -1.56 (1.28) 3.77 (2.04) -37.10 (3.57) 1.96 (1.52) 9.81 (0.98)

2.46 (2.90) 7.83 (0.91) 8.84 (1.73) 13.16 (1.33) 9.37 (2.33) 15.16 (2.77)

8.73 (3.23) -9.39 (1.57) -5.07 (2.68) -50.26 (3.81) -7.40 (2.78) -5.35 (2.94)

**

***

5.84 (1.06) 4.38 (2.60) 11.40 (2.46) -6.37 (2.42) -9.84 (4.47) 1.44 (1.06)

1.74 (1.03) 7.12 (1.07) -2.84 (1.07) 4.11 (1.00) 6.58 (0.87) 4.98 (1.42)

4.11 (1.48) -2.74 (2.81) 14.24 (2.68) -10.48 (2.62) -16.42 (4.55) -3.54 (1.77)

*

12.30 (0.91) 0.82 (1.29) 0.86 (1.41) -4.50 (2.80) -0.79 (2.17) 6.87 (1.18) 10.62 (1.68)

1.34 (1.22) 6.17 (1.03) 1.09 (1.50) 2.63 (1.19) -0.70 (1.30) 0.95 (1.23) -6.91 (1.44)

10.96 (1.52) -5.35 (1.66) -0.24 (2.06) -7.13 (3.04) -0.09 (2.53) 5.92 (1.70) 17.53 (2.22)

**

**

Panel C: VWR estimate

USA

GBR

FRA

DEU

JPN

CAN

ITA

10.15 (0.96) -0.46 (1.81) 11.11 (0.88) -26.89 (3.81) -5.98 (1.79) 10.06 (0.79)

1.73 (2.12) 10.54 (1.14) 6.23 (1.25) 12.34 (0.95) 13.80 (1.05) -5.33 (3.87)

8.42 (2.33) -11.00 (2.14) 4.88 (1.52) -39.23 (3.93) -19.78 (2.08) 15.39 (3.95)

*

***

***

4.00 (0.75) -2.26 (1.09) 7.77 (1.50) -7.06 (2.39) -2.75 (2.74) 0.86 (0.74)

2.83 (0.92) 6.47 (0.72) 2.25 (0.69) 4.85 (0.59) 3.58 (0.63) 9.06 (1.09)

1.17 (1.18) -8.73 (1.31) 5.52 (1.65) -11.91 (2.46) -6.32 (2.82) -8.20 (1.31)

**

*

**

9.08 (0.90) 0.61 (1.55) 4.76 (1.02) 1.11 (1.80) 6.42 (1.25) 5.43 (1.03) 6.68 (1.01)

-2.30 (1.31) 5.46 (0.89) 2.35 (1.26) 4.13 (0.93) 2.69 (0.99) 1.66 (1.30) 1.45 (1.20)

11.39 (1.59) -4.86 (1.79) 2.41 (1.62) -3.02 (2.03) 3.74 (1.59) 3.77 (1.66) 5.24 (1.57)

**

Table 7. G7 Political Cycles and Return Di˙erence The table reports mean excess returns di˙erence based on G7 political cycles. Panel A presents the pooled estimate of 23 countries. Panel B presents the local stock returns of each G7 country. Panel C presents the value-weighted returns across 23 countries. The standard errors below the coeÿcients are obtained from the Newey-West heteroskedasticity and serial-correlation robust statistics. The asterisk indicates signifcance tests on the null hypothesis that stock returns are the same during Left and Right presidential (or prime-minister) terms. p-value < 10% *, 5% **, 1% ***.

39

1927:02-1956:12 1957:01-1986:12 1987:01-2018:07

Election Left Right Di˙ Left Right Di˙ Left Right Di˙

Panel A: Pooled estimate

USA

GBR

FRA

DEU

JPN

CAN

ITA

0.85 (0.29) -0.79 (0.67) -1.23 (0.77) 1.39 (0.05) 1.88 (0.03) 0.83 (0.23)

1.44 (0.05) 1.79 (0.04) 1.81 (0.04) 0.98 (0.23) 0.58 (0.31) 2.14 (0.05)

-0.59 (0.30) -2.58 (0.67) -3.04 (0.77) 0.41 (0.24) 1.29 (0.31) -1.31 (0.24)

*

0.67 (0.09) 0.71 (0.10) 11.53 (1.98) 0.63 (0.12) 1.95 (0.18) 1.84 (0.40)

4.50 (0.72) 4.44 (0.72) 0.65 (0.06) 3.53 (0.54) 3.15 (0.48) 6.45 (0.75)

-3.82 (0.73) -3.73 (0.72) 10.88 (1.98) -2.90 (0.56) -1.19 (0.51) -4.61 (0.85)

*

*

*

*

*

-1.88 (0.15) 0.76 (0.18) 1.90 (0.15) 0.85 (0.22) 2.06 (0.19) 0.03 (0.14) -2.44 (0.15)

3.79 (0.14) 1.02 (0.13) 0.07 (0.15) 0.93 (0.12) 0.61 (0.12) 1.71 (0.16) 3.02 (0.14)

-5.68 (0.20) -0.26 (0.22) 1.83 (0.21) -0.08 (0.25) 1.45 (0.22) -1.68 (0.21) -5.46 (0.20)

***

**

**

**

***

Panel B: G7 local FX returns

USA

GBR

FRA

DEU

JPN

CAN

ITA

0.01 (0.09) 1.05 (0.02) 3.57 (0.07) 2.66 (0.12) 0.04 (0.06)

0.76 (0.04) 1.26 (0.13) -7.28 (2.72) 3.29 (0.26) 1.80 (0.02)

-0.75 (0.10) -0.21 (0.14) 10.85 (2.72) -0.63 (0.29) -1.76 (0.06)

**

***

2.36 (0.14) 7.49 (1.55) -0.78 (0.06) 8.10 (0.19) 0.86 (0.13)

2.02 (0.50) 1.38 (0.12) -0.09 (0.30) 0.97 (0.29) 0.95 (0.08)

0.34 (0.53) 6.11 (1.55) -0.69 (0.31) 7.14 (0.35) -0.09 (0.15)

***

0.86 (0.75) 2.36 (0.80) 0.20 (1.17) 0.72 (1.20) 0.90 (0.48) -2.45 (0.75)

1.58 (0.70) -0.75 (0.73) -0.11 (0.62) -1.02 (0.63) 0.71 (0.57) 2.60 (0.77)

-0.72 (1.03) 3.11 (1.08) 0.31 (1.33) 1.74 (1.36) 0.18 (0.74) -5.06 (1.07)

*

Panel C: EWR – dollar factor

USA

GBR

FRA

DEU

JPN

CAN

ITA

0.91 (0.28) -0.66 (0.63) -1.32 (0.78) 1.39 (0.05) 1.90 (0.01) 0.86 (0.23)

1.44 (0.03) 1.79 (0.02) 1.80 (0.02) 1.01 (0.23) 0.62 (0.29) 2.14 (0.01)

-0.53 (0.28) -2.45 (0.63) -3.13 (0.78) 0.37 (0.23) 1.28 (0.29) -1.29 (0.23)

*

0.69 (0.06) 0.64 (0.08) 11.45 (0.97) 0.63 (0.11) 1.95 (0.11) 1.64 (0.22)

4.08 (0.33) 3.92 (0.37) 0.69 (0.07) 3.04 (0.22) 2.76 (0.21) 5.28 (0.68)

-3.39 (0.34) -3.28 (0.38) 10.76 (0.98) -2.41 (0.25) -0.81 (0.23) -3.64 (0.71)

***

**

***

***

*

-1.88 (0.49) 0.76 (0.65) 1.95 (0.46) 0.85 (0.85) 2.06 (0.67) 0.04 (0.50) -2.44 (0.53)

3.83 (0.55) 1.06 (0.42) 0.07 (0.54) 0.96 (0.40) 0.64 (0.43) 1.76 (0.53) 3.06 (0.48)

-5.71 (0.74) -0.30 (0.78) 1.88 (0.71) -0.11 (0.94) 1.41 (0.80) -1.73 (0.73) -5.50 (0.71)

**

**

Table 8. G7 Political Cycles and Currency Return Di˙erence The table reports currency excess re-turns based on G7 political cycles. Panel A presents the pooled estimate of 23 countries. Panel B presents the local FX returns of each G7 country. Panel C presents the dollar factor, i.e. equal-weighted FX returns across 23 countries. The standard errors below the coeÿcients are obtained from the Newey-West heteroskedasticity and serial-correlation robust statistics. The asterisk indi-cates signifcance tests on the null hypothesis that stock returns are the same during Left and Right presidential (or prime-minister) terms. p-value < 10% *, 5% **, 1% ***.

40

Dependent variable: 4-year cumulative stock returns denoted in local currency

(1) (2) (3) (4) (5) (6) (7) (8) (9)

DEM 0.5496*** 0.5370*** 0.5371*** 0.5604*** 0.5461*** 0.5336*** 0.5549*** 0.5455*** 0.5377*** (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

TS1 -0.0206** (0.0050)

DP1 -0.0191** (0.0038)

RREL1 0.0000 (0.6641)

TS6 -0.0195*** (0.0007)

DP6 -0.0228** (0.002)

RREL6 -0.0003** (0.0047)

TS12 -0.0112* (0.0361)

DP12 -0.0188* (0.0141)

RREL12 0.0000 (0.7342)

Constants -0.0564 -0.0956** -0.1333*** -0.0672 -0.0816* -0.1311*** -0.0959* -0.0863* -0.1334*** -0.1439 -0.0089 -0.0005 -0.0723 -0.0239 -0.0006 -0.017 -0.0189 -0.0004

N 159 159 159 159 159 159 159 159 159 R-sq 0.461 0.451 0.432 0.454 0.456 0.433 0.439 0.448 0.432

Table 9. Long-horizon predictability and real business cycle This table analyzes the long-horizon return predictability of the real business cycle variables. The long-term horizon is of four years, equivalent to the tenure of a presidential term. DEM is a dummy variable indicating the Democrat presidential terms. TS is referred to the term spreads and the numbers indicate the number of lagged-month. For instance, TS1 is the term spreads observed at month t − 1 and TS6 is the term spreads at month t − 6 and so on. DP is referred to the dividend-to-price ratio also known as dividend yields. RREL is referred to the relative short rate and is calculated as the current month’s short rate minus a moving average of the past 12 months’ rate. The coeÿcients is expressed in percentage. p-values in parentheses * p<0.05 ** p<0.01 *** p<0.001.

41

Panel A: Equity excess returns denominated in local currencies

(1) (2) (3) (4) (5)

DEM

HHNW

US HHNW

HHNW Fin

0.0132*** (0.0000)

0.0134*** (0.0000) -0.1330 (0.4356)

0.0122*** (0.0000) -0.1459 (0.3471) 0.6003*** (0.0000)

0.0134*** (0.0000)

-0.0519 (0.5879)

0.0121*** (0.0000)

0.6002*** (0.0000) -0.0611 (0.4860)

N R-sq

3636 0.014

3636 0.014

3636 0.022

3636 0.014

3636 0.022

Panel B: Equity excess returns denominated in USD

(6) (7) (8) (9) (10)

DEM

HHNW

US HHNW

HHNW Fin

0.0073*** (0.0007)

0.0071*** (0.0003) 0.1217 (0.6521)

0.0055** (0.0036) 0.1056 (0.6810) 0.7445*** (0.0000)

0.0071*** (0.0004)

0.0760 (0.5954)

0.0055** (0.0041)

0.7444*** (0.0000) 0.0646 (0.6353)

N R-sq

3636 0.003

3636 0.003

3636 0.013

3636 0.003

3636 0.013

Panel C: Currency risk premia

(11) (12) (13) (14) (15)

DEM

HHNW

US HHNW

HHNW Fin

-0.0059*** (0.0000)

-0.0063*** (0.0000) 0.2547 (0.0901)

-0.0067*** (0.0000) 0.2515 (0.0931) 0.1442*** (0.0005)

-0.0063*** (0.0000)

0.1279 (0.0736)

-0.0066*** (0.0000)

0.1442*** (0.0005) 0.1257 (0.0780)

N R-sq

3636 0.010

3636 0.012

3636 0.013

3636 0.011

3636 0.013

Table 10. Wealth e˙ect I: household net worth This table studies the wealth e˙ect through household net worth. Panel A presents the regression of equity returns denominated in local currencies as the dependent variable, Panel B presents the equity returns in USD and Panel C presents the currency risk premia. HHNW is the percentage change of household net worth and HHNW Fin is the logarithm growth of household fnancial net worth. Both data are obtained from the OECD website. US HHNW is the growth of real, per capita net worth of the US households and is from the Board of Governors of the Federal Reserve System. DEM is a dummy variable indicating the Democrat presidential terms. The coeÿcients is expressed in percentage. p-values in parentheses * p<0.05 ** p<0.01 *** p<0.001.

42

Panel A: Equity excess returns in local currencies Panel B: Equity excess returns in USD

(1) (2) (3) (4) (5) (6) (7) (8)

DEM 0.0100*** 0.0100*** 0.0088*** 0.0088*** 0.0046*** 0.0045*** 0.0040** 0.0040**

Market Cap

US Market Cap

Constant

(0.0000)

-0.0021*

(0.0000) 0.0003 (0.4874)

-0.0024**

(0.0000)

0.0098*** (0.0001) -0.0120***

(0.0000) -0.0011** (0.0014) 0.0111*** (0.0000) -0.0123***

(0.0008)

0.0016*

(0.0010) 0.0006 (0.3368)

0.0011

(0.0026)

0.0049* (0.0111) -0.0034

(0.0026) -0.0000 (0.9746) 0.0049* (0.0145) -0.0034

(0.0103) (0.0099) (0.0000) (0.0000) (0.0199) (0.1954) (0.0910) (0.0921)

N 8084 8084 8084 8084 8084 8084 8084 8084 R-sq 0.008 0.008 0.011 0.011 0.001 0.001 0.002 0.002

Panel C: Currency risk premia

(9) (10) (11) (12)

DEM -0.0054*** -0.0054*** -0.0048*** -0.0048***

Market Cap

US Market Cap

Constant

(0.0000)

0.0037***

(0.0000) 0.0003 (0.4100)

0.0034***

(0.0000)

-0.0049*** (0.0000) 0.0087***

(0.0000) 0.0011 (0.0656) -0.0061*** (0.0000) 0.0089***

(0.0000) (0.0000) (0.0000) (0.0000)

N 8084 8084 8084 8084 R-sq 0.010 0.010 0.013 0.015

Table 11. Wealth e˙ect II: market capitalization This table studies the wealth e˙ect through market capitalization. The market capitalization is the total market value of each country as the end of the month divided by the country’s GDP from the previous year. Note that the GDP is at annual frequency while all other variables are at monthly frequency. DEM is a dummy variable indicating the Democrat presidential terms. Market Cap is the market capitalization of the non-US countries while US Market Cap is the market capitalization of the US. p-values in parentheses * p<0.05 ** p<0.01 *** p<0.001.

43

Panel A: Mean excess returns

Local currencies USD Foreign exchange

(1) (2) (3) (4) (5) (6) (7) (8) (9)

DEM

Cent

DEM×Cent

α

0.010*** (0.000)

-0.002 (0.178)

0.010*** (0.000) -0.000 (0.720)

-0.000 (0.977)

0.008** (0.006) -0.000 (0.499) 0.000 (0.309) 0.001 (0.762)

0.004** (0.007)

0.002 (0.091)

0.004** (0.008) 0.000 (0.907)

0.002 (0.711)

0.003 (0.256) -0.000 (0.982) 0.000 (0.710) 0.002 (0.644)

-0.005*** (0.000)

0.004*** (0.000)

-0.006*** (0.000) 0.000 (0.215)

0.002 (0.280)

-0.004*** (0.000) 0.000 (0.090) -0.000 (0.131) 0.001 (0.637)

N R-sq

749 0.068

749 0.068

749 0.069

749 0.021

749 0.021

749 0.021

749 0.124

749 0.126

749 0.129

Panel B: Volatility of excess returns

Local currencies USD Foreign exchange

(10) (11) (12) (13) (14) (15) (16) (17) (18)

DEM

Cent

DEM×Cent

α

-0.004* (0.011)

0.053*** (0.000)

-0.005** (0.007) 0.000 (0.534)

0.049*** (0.000)

-0.006* (0.040) 0.000 (0.701) 0.000 (0.490) 0.050*** (0.000)

-0.002 (0.310)

0.058*** (0.000)

-0.002 (0.387) -0.001 (0.357)

0.063*** (0.000)

-0.002 (0.604) -0.001 (0.387) 0.000 (0.957) 0.063*** (0.000)

0.001 (0.057)

0.022*** (0.000)

0.002* (0.047) -0.000 (0.388)

0.024*** (0.000)

0.005** (0.002) 0.000 (0.797) -0.000** (0.009) 0.022*** (0.000)

N R-sq

749 0.128

749 0.128

749 0.129

749 0.089

749 0.089

749 0.089

749 0.479

749 0.479

749 0.485

Table 12. Regression analysis on trade network centrality This table reports the regression results of Equation (4). The sample period is from 1987 to 2018 in order to match the availability of centrality measure in Figure 7. We annualize the time series by calculating average and standard deviation of monthly observations of stock returns and currency risk premia. The stock returns are denoted in both local currencies and USD. The dependent variables in the Panel A is the average of stock excess returns and in Panel B are their volatilities. p-values in parentheses * p<0.05 ** p<0.01 *** p<0.001.

44

Panel A: Equity excess returns denominated in local currencies

(1) (2) (3) (4) (5) (6) (7) (8)

DEM 0.0085*** 0.0086*** 0.0142*** 0.0085*** 0.0087*** 0.0087***

Deg Cent

DEM×Deg Cent

Bet Cent

(0.0000) 0.0002 (0.0818)

(0.0000) 0.0003* (0.0458)

(0.0000) 0.0008*** (0.0000) -0.0008*** (0.0001)

(0.0000)

0.0000

(0.0000)

0.0001

(0.0000)

0.0001

DEM×Bet Cent (0.6110) (0.2414) (0.2467)

-0.0000 (0.9730)

R-sq N

0.007 6479

0.002 6479

0.008 6479

0.010 6479

0.007 6479

0.001 6479

0.008 6479

0.008 6479

Panel B: Equity excess returns denominated in USD

(9) (10) (11) (12) (13) (14) (15) (16)

DEM 0.0032* 0.0034* 0.0075** 0.0032* 0.0035* 0.0033*

Deg Cent

DEM×Deg Cent

Bet Cent

(0.0381) 0.0005*** (0.0003)

(0.0283) 0.0006*** (0.0002)

(0.0025) 0.0010*** (0.0000) -0.0006* (0.0126)

(0.0381)

0.0001*

(0.0259)

0.0001*

(0.0492)

0.0001*

DEM×Bet Cent (0.0364) (0.0213) (0.0268)

0.0000 (0.6557)

R-sq N

0.002 6479

0.003 6479

0.004 6479

0.004 6479

0.002 6479

0.002 6479

0.002 6479

0.002 6479

Panel C: Currency risk premia

(17) (18) (19) (20) (21) (22) (23) (24)

DEM -0.0053*** -0.0052*** -0.0067*** -0.0053*** -0.0052*** -0.0054***

Deg Cent

DEM×Deg Cent

Bet Cent

(0.0000) 0.0003*** (0.0000)

(0.0000) 0.0003*** (0.0000)

(0.0000) 0.0002 (0.1185) 0.0002 (0.0556)

(0.0000)

0.0001***

(0.0000)

0.0001**

(0.0000)

0.0001*

DEM×Bet Cent (0.0003) (0.0053) (0.0118)

0.0000 (0.3269)

R-sq N

0.011 6479

0.004 6479

0.013 6479

0.014 6479

0.011 6479

0.003 6479

0.011 6479

0.012 6479

Table 13. Fund fow centrality This table analyses the fund fow network and its interaction with the US political cycle. Panel A presents the regression of equity returns denominated in local currencies as the dependent variable, Panel B presents the equity returns in USD and Panel C presents the currency risk premia. DEM is a dummy variable indicating the Democrat presidential terms. Deg Cent is referred to the degree centrality rank of the country. Bet Cent is referred to the between centrality rank of the country. p-values in parentheses * p<0.05 ** p<0.01 *** p<0.001.

45

Panel A: Country FE and Year FE

(1) (2) (3) (4) (5) (6) (7)

DEM 0.154** 0.195** 0.205** 0.093 0.057 0.154** 0.088

Ave. Tari˙ (0.004) (0.005)

-0.001 (0.003) -0.001

(0.097) (0.415) (0.004) (0.152)

DEM × Ave. Tari˙ (0.114) (0.197)

-0.001

MFN Tari˙ (0.207)

0.012 -0.001

DEM × MFN Tari˙ (0.212) (0.943)

0.016

Com. Tari˙ (0.345)

-0.003 -0.015

DEM × Com. Tari˙ (0.741) (0.157)

0.027* (0.031)

N 7661 4451 4451 5982 5982 7661 7661 R-sq Country FE Year FE

0.090 Y Y

0.084 Y Y

0.085 Y Y

0.104 Y Y

0.105 Y Y

0.090 Y Y

0.091 Y Y

Panel B: Time-varying Country FE

(8) (9) (10) (11) (12) (13) (14)

DEM 0.127*** 0.125*** 0.134*** 0.116*** 0.025 0.127*** 0.041

Ave. Tari˙ (0.000) (0.000)

-0.001 (0.000) -0.001

(0.000) (0.549) (0.000) (0.185)

DEM × Ave. Tari˙ (0.155) (0.247)

-0.000

MFN Tari˙ (0.293)

0.009 -0.026

DEM × MFN Tari˙ (0.084) (0.086)

0.040*

Com. Tari˙ (0.012)

-0.001 -0.020*

DEM × Com. Tari˙ (0.849) (0.033)

0.032** (0.002)

N 7661 4439 4439 5982 5982 7661 7661 R-sq Country x month FE

0.056 Y

0.069 Y

0.069 Y

0.065 Y

0.067 Y

0.056 Y

0.058 Y

Table 14. Trade tari˙s and risk premia denominated in local currencies This table analyses the role of tari˙ and its interaction with the US political cycle. The dependent variable throughout the table is the equity excess returns denoted in local currencies. Panel A presents the regression controlling for both country fxed e˙ect and year fxed e˙ect. Panel B presents the regression controlling for time-varying country fxed e˙ect. DEM is a dummy variable indicating the Democrat presidential terms. Ave. Tari˙ is referred to the average tari˙ calculated from the ratio of collected duties to tax revenue. MFN Tari˙ is referred to the Most-Favored-Nation tari˙ reported by WTO and Com. Tari˙ is referred to the combination of both Ave. Tari˙ and MFN Tari˙. p-values in parentheses * p<0.05 ** p<0.01 *** p<0.001.

46

Panel A: Country FE and Year FE

(1) (2) (3) (4) (5) (6) (7)

DEM 0.004 -0.013 -0.017 0.039 0.111* 0.004 0.047

Ave. Tari˙ (0.920) (0.829)

0.000 (0.776) 0.000

(0.443) (0.029) (0.921) (0.297)

DEM × Ave. Tari˙ (0.692) (0.847)

0.000

MFN Tari˙ (0.321)

-0.007 0.018*

DEM × MFN Tari˙ (0.176) (0.023)

-0.032***

Com. Tari˙ (0.000)

-0.003 0.005

DEM × Com. Tari˙ (0.440) (0.232)

-0.018*** (0.000)

N 7661 4451 4451 5982 5982 7661 7661 R-sq Country FE Year FE

0.057 Y Y

0.051 Y Y

0.052 Y Y

0.058 Y Y

0.060 Y Y

0.057 Y Y

0.058 Y Y

Panel B: Time-varying Country FE

(8) (9) (10) (11) (12) (13) (14)

DEM -0.066*** -0.049*** -0.052*** -0.070*** 0.015 -0.066*** -0.015

Ave. Tari˙ (0.000) (0.000)

0.000 (0.000) 0.000

(0.000) (0.310) (0.000) (0.211)

DEM × Ave. Tari˙ (0.547) (0.682)

0.000

MFN Tari˙ (0.407)

-0.009** 0.025***

DEM × MFN Tari˙ (0.005) (0.000)

-0.037***

Com. Tari˙ (0.000)

-0.002 0.008*

DEM × Com. Tari˙ (0.342) (0.023)

-0.019*** (0.000)

N 7661 4439 4439 5982 5982 7661 7661 R-sq Country x month FE

0.054 Y

0.062 Y

0.062 Y

0.064 Y

0.068 Y

0.055 Y

0.057 Y

Table 15. Trade tari˙s and currency risk premia This table analyses the role of tari˙ and its interaction with the US political cycle. The dependent variable throughout the table is the currency risk premia. Panel A presents the regression controlling for both country fxed e˙ect and year fxed e˙ect. Panel B presents the regression controlling for time-varying country fxed e˙ect. DEM is a dummy variable indicating the Democrat presidential terms. Ave. Tari˙ is referred to the average tari˙ calculated from the ratio of collected duties to tax revenue. MFN Tari˙ is referred to the Most-Favored-Nation tari˙ reported by WTO and Com. Tari˙ is referred to the combination of both Ave. Tari˙ and MFN Tari˙. p-values in parentheses * p<0.05 ** p<0.01 *** p<0.001.

47