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Stock Returns Pasquale Della Corte ∗ Hsuan Fu † Imperial College London Laval University January 11, 2020 PLEASE DO NOT CITE WITHOUT AUTHORS’ PERMISSION Abstract This paper demonstrates the US political cycle is important to understand the risk premia of the stock and currency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
47
Embed
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
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
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PLEASE DO NOT CITE WITHOUT AUTHORS’ PERMISSION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]
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
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.
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.
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
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.
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.
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.
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˙
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˙
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
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
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**
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.
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.