Are market returns higher under left-wing presidencies in France? Yvan Villemagne
Abstract
Based on U.S. findings, we investigate whether the presidential puzzle applies in France. An OLS regression is performed to find average returns (log total, log excess and log real) based on the political affiliation of parties in presidency. Preliminary evidence suggests that market returns are higher under left-wing presidents. To fully unveil the presidential puzzle in the France, we test whether volatility under left-wing parties explains higher returns. We find no evidence of higher volatility under left-wing parties and as such, the presidential puzzle is also occurring in France since 1988. In a parallel analysis, we reverse our dependent and independent variable to question whether the real predictor of political outcome is in fact the returns of the market. We find that recessions increase the probability of right-wing
government, while booms increase the likelihood of left-wing party. Such results could be plausible explanations of the presidential puzzle.
Keywords: asset pricing, stock market, large firms, political cycle, political extremism, voting
I. Introduction
The election of Macron led to an unexpected turn in the history of politics in France. From his recently formed party “La République En Marche” (LREM) a year earlier, the election of Macron disrupted the political scene for the traditional incumbents of French politics. Newly elected president Macron advocated for economic projects that would favor the supply-side of the economy. For instance, reduction of taxes to French corporations would soon be reduced by 16 billion Euros by scaling the current taxes of 33.3% to the European average of 25%. Also, employer’s social contribution would be cut to 0% and financial assets such as stocks and bonds would benefit an exemption from wealth tax (Sénécat, Le Monde, 2018). Measures in favor of the liberalization the economy, the
reduction of taxes and obligations of firms are typically what right-leaning parties would try to implement nationally, or at least defend in their electoral programs. This is in contrast with French left-leaning parties such as the predecessor François Hollande of the “Parti Socialiste” (PS) that consistently maintained high-level of taxes and supported a socially oriented welfare model. Clearly, the divide is clear between left- and right-wing parties and a question that subsequently arises is whether market returns respond differently to left- or right-wing political constituencies and if so, do they react accordingly to the expectations of political party’s preconceived business orientation.
In the U.S, Riley and Luksetisch (2000) observe that DJIA market returns are higher right after a Republican win between 1900 and 1968 supporting the idea, according to the author, that “market prefers Republicans” (p.15), at least on the short-term. However, conflicting results have been observed on a longer-term horizon. For instance, Santa-Clara and Valkanov’s (2003) have found that average returns of U.S companies are significantly higher under Democrats between 1927 to 1998. This disparity in returns between the
political constituencies of the U.S contrasts with the idea that returns should be higher under Republicans, often referred to as the party of business since it is more liberal. Since the
authors find no evidence of increased risk environment or business cycle effects that could explain the differential in returns between political constituencies, they are left with a “presidential puzzle”. Other authors attribute the presidential puzzle to econometric issues (Sy & Al Zaman, 2011; Campbell & Li, 2004) while other non-US based studies such as the ones conducted in New-Zealand have shown no significant evidence to the claim of the presidential puzzle (Calan et al., 2005). Overall, the presidential puzzle is an ongoing investigation and because large part of the studies solely focuses on the American context, these studies raise the question as to whether the presidential puzzle holds in other contexts, places and people. Therefore, is France’s stock market subject to the presidential puzzle? Are companies performing better under left- or right-wing parties in France? if so, can this be explained by the increased volatility of the market?
In contradiction with previous studies and in an effort of completeness, this study will also formalize the potentiality of reverse causality. While the effect of political
constituencies on market return has been show by several authors, a departure from this would be to redirect the causal arrow using the market returns as a predictor of political party in power. By approaching the subject in a pluri-dimensional way, results could potentially be an explication of higher returns under left-leaning parties. For instance, the extreme case of recessions has shown to increase voter’s intention in favor of nationalistic parties, which are typically on the right of the political scale (Weatherford, 1978; Brückner and Grüner, 2010). Therefore, since the causal effect direction is not necessarily evident, we extend the study to investigate whether good market conditions could significantly increase the probability of having a specific party in the future. In other words, do market returns influence the political outcomes in France since 1988?
The rest of the paper is structed as follows. Section 2 discusses the previous literature on the effect of political information on markets and the premises of the presidential puzzle theory. Section 3 will stress out the variables used throughout the paper that will be used to answer our hypotheses. We aim to understand why the dependent and independent
variables selected are meaningful for the conduct of this study. Section 4 provides the results of regression analysis. Finally, section 5 provides a discussion on the results and possible limitations of the current study leading to suggestions for future research.
II. Literature review
One of the most important aspect of capitals market is to ensure that information is rightly incorporated into asset prices. Fama’s (1970) efficient market hypothesis has been essential in having a deeper understanding of market’s incorporation of new information. Ideally, markets would be efficient if they would integrate all available information in asset prices. In the context of elections, studies seem to confirm a transmission mechanism
between political information and asset prices. For example, Knight (2000) tests a sample of 70 politically sensitive U.S. firms and finds that during the 2000 U.S. election, the policies of campaigners were rightfully incorporated into stock price. His findings show that favorable public policies influence directly by 3 to 6% of a firm’s total value. In the same alignment, Wagner et al. (2018) show that the Trump’s tax policy of lowering corporate taxes and a more restrictive trade policy has had an observable effect on stock markets. In general, high-tax paying companies with large deferred high-tax liabilities (DTLs) were positively affected while firms with large deferred tax asset such as net operating loss carryforward (NOL DTAs) were negatively impacted. Furthermore, not only do U.S policies affect domestic returns, the effect of such tax policies has also been observed on foreign market yields, implying a transnational effect of U.S. domestic policies. Indeed, Gaertner et al. (2020) examine the Tax Cuts and Job Act (TCJA) and the effect on non-U.S. markets and finds that the Chinese market is
impacted by negative returns while the rest of the world experiences positive returns. Finally, capitalization of political information in asset prices have also occurred in foreign countries. The Taiwan-based study of Imai and Shelton (2011) on the 2008 Presidential Campaign shows the effect that political influence can exert on market. In the bipolar
Taiwanese political scene, two party emerge with one party leaning towards the unification with Mainland China and the opposing party promoting independence of Taiwan (DPP). The authors observe that increases in expected votes for the DPP leads to a simultaneous
decrease of share prices for Taiwanese firms with mainland investments. Therefore, Taiwanese stocks that are affiliated to China respond negatively to the surge of the independence seeking party. This provides further evidence of a relatively efficient
transmission mechanism that impounds accurately political information in asset prices that occurs not only in the U.S but in other settings. Overall, there is a consensus that parties’ policies information is priced into share prices in a relatively efficient way.
While the previous studies are mainly event studies comprising short- to medium- run timespan, to fully uncover the presence of the presidential puzzle, returns between left- and right- wing parties must be thoroughly analyzed on a long-term basis and by analyzing aggregated averages. For instance, in the US, Santa-Clara and Valkanov (2003) find that excess returns of the value-weighted CRSP portfolio under Democrat presidencies are 9% higher on average than that of Republican presidencies over the period from 1927 to 1998. By using political dummies, they compare the differences in aggregated averages between left- and right-wing governments. The authors have first verified if these differences in returns were not correlated with business cycles. Their results show that they are not caused by the business cycle. Second, they further try to confirm if such differential can be explained with a higher risk environment in markets. Indeed, in efficient markets, one would predict that such difference in return could be explained by the presence of a democratic premium implying a higher risk environment and therefore a higher compensation for risk (Fama, 1970). Surprisingly, they find no evidence of increased risk when observing the volatility of the U.S. market. To that extent, the conflicting and unexplainable results are referred by the authors as the “presidential puzzle”. Other authors have also documented such differentials between Democrats and Republicans. For instance, Huang (1985) also finds some evidence of higher returns in cases of democratic administrations in year 1 and 2 in period 1961-80
and year 3 and 4 in period 1921-80 considering a 4-year presidential mandate. Not only do these results show a differential of returns, a political cycle on market returns seems to occur as well.
However, the results of Santa-Clara and Valkanov (2003) remain controversial in the existing literature and some authors relate the puzzle to an econometrical issue. For
instance, Campbell and Li (2004) demonstrate that the difference is not as important when employing statistical techniques such as WLS and GARCH. Also, Sy and Al Zaman (2011) provide with additional evidence in the explanation of the presidential puzzle. They state that when the risk is properly taken into account, by using a conditional version of the three-factor pricing model of Fama and French (1996) model, which enables Betas to vary across political cycle and take into account market, size and value factors, the efficient market hypothesis holds. That is, they find evidence of higher risk under democratic presidency which explain higher return through the presence of a premium to compensate for the risk. Furthermore, New-Zealand have shown to present the opposite patterns which shows that the presidential puzzle does not necessarily transfer in a similar two-party democracy (Calan et al., 2005). Overall, “the relation between the political cycle and the stock market remains a puzzle” (Santa-Clara and Valkanov, 2003, p. 4). Therefore, it seems logical to question whether our variables are really causal, that is, if our independent variables do indeed predict our dependent variable. In their study, Santa-Clara and Valkanov (2003) test whether business cycles might be driving the differential in returns since lower GDP growth is observed under Republicans while Democrats show significant higher inflation (Alesina & Rosenthal, 1995). They control for any variations related to business cycles by including macroeconomic variables (that have been shown to predict market returns) as control variables in the regression of market returns. However, the added control variables do not explain the differential in returns and the puzzle remains.
Interestingly, previous studies hypothesized the effect of economic conditions on political outcomes, also suggesting that parties may not predict returns but instead, that
extreme economic conditions such as recession may actually affect political outcomes. Weatherford (1978) analyzes the American voters during the Recession of 1958 and 1968. The author finds that recessions lead to the increase of right-party voting intentions.
Furthermore, the effect is amplified for working classes compared to middle classes. He also observes that an economic downturn seems to disfavor the incumbent party at the following election. In the same alignment, Brückner and Grüner (2010) find evidence of a negative relationship between economic growth and the rise of political extremism. Specifically, recessions increased the number of extremist votes. The authors observe that a percentage point increase roughly leads to a simultaneous increase of a percentage point in the shares of votes casted to right-wing or nationalist parties. These pieces of evidence suggest that the reversal of our causal arrow is appropriately motivated.
Overall, the fact that this is a puzzle makes it even more insightful to reproduce it in the French market context since additional evidence is needed. The study could validate the theory as in the U.S or instead find no evidence such as in New-Zealand. Therefore, to attempt to uncover the presidential puzzle in France, we define the hypothesis that will be tested in the results section:
Hypothesis 1. Returns in France are higher under left-wing presidencies than that of
right-wing presidencies
Hypothesis 2. Volatility of market is higher under left-wing presidencies than that of
right-wing presidencies
Hypothesis 3. Good market conditions increase the likelihood of left-wing governments in
III. Methods
In this section, we present the variables used for our regression analysis. The monthly return data comprises the sample period from 1988:02 to 2020:03, containing a total of 386
monthly observations, 6 mandates, 4 right-wing governments and 2 left wing governments. Due to the lack of data available, the Mitterrand presidency will be truncated from 1981:05-1995:05 to 1988:02 – 1981:05-1995:05. This is reasonable considering that it entails the whole period of his presidency (7-year mandate prior to 2002). The market returns obtained represent 147 monthly returns under left-wing governments (38% of monthly return data) and 239 right-wing government returns (62% of monthly return data). Therefore, 386 monthly returns are analyzed. Table 1 outlines the timespan covered in the study and governance period under each president along with the political affiliation. In general, while there are more returns available for right-wing mandate returns, the diversity of presidents (Mitterrand, Chirac, Sarkozy, Hollande and Macron) and larger mandates (5-year mandate, 7-year prior to 2002) provide us with an interesting starting point for the potential unveiling of the presidential puzzle in France. Data has been found to be overall consistent, with no outliers and no missing values (Table 2).
To answer our hypotheses, three separate regressions and a t-test by group will be conducted. First, an OLS regression using the political affiliation category as independent variable will be used to find evidence of higher returns under left-wing parties. Then, the OLS regression model will be extended to include control variables that also affect our dependent variable. According to Durbin Watson test, we find positive correlation in both the simple and multiple regressions with test results in the range of [1.79: 1.95], implying that the preceding month’s return has an impact on the following months. This is typically inherent to the use of time-series in general as the errors are often correlated to one another (autocorrelation). To correct for the positive serial correlation and heteroskedasticity in the error term of the model, we use an OLS regression using Newey-West (1987) procedure. This will allow to find consistent estimates and more accurate estimates of standard errors. Then, we run a t-test by group on the FTSE volatility data of the French market enabling us to confirm if a differential in risk could be the reason why the returns diverge between political constituencies. Finally, to address potential reverse causality and to tackle possible
endogeneity issues, we extend our research to a Probit regression. Here, we depart from Santa-Clara and Valkanov (2003) procedure by reverting the causal arrow instead of adding macroeconomic variables to the main regression. Taking the political dummy variable as our dependent variable and market returns as our independent variable, we investigate whether market returns increased the likelihood of a specific party in place.
In the next subsections, we define the variables that will be subsequently used in the remainder of the study.
A. Dependent variable
Our dependent variable will be three variations of return which originates from the FTSE France index available on FactSet. It represents a total of 85 mid- and large-capitalization companies in France with a total market value of 1 258 900 million of euros. The index
constituents come from various industries such as consumer goods (24.36%), industrials (22.21%), health care (12.51%), financials (10.75%), oil & gas (7.29%) and technology firms (7.10%) amongst others. With a representative sample in our index, we introduce our first form of the dependent variable as the logarithmic total return, expressed as RETt+1. Index value data is retrieved from FactSet and converted to total logarithmic return in the following way:
𝑅𝐸𝑇 = ln ( 𝐹𝑇𝑆𝐸 𝐼𝑁𝐷𝐸𝑋 𝑡−1 𝐹𝑇𝑆𝐸 𝐼𝑁𝐷𝐸𝑋 𝑡 )
Where t = month
We use data that includes dividends as variation in returns should not be influenced by any disbursements (since dividends are included in the total return given to
shareholders). Second, following Santa-Clara and Valkanov (2003), the excess return ERt+1 is used which is the log market return minus the log one-month French Treasury Bill rate. The data is available on the OECD website. To find the excess returns, we de-annualize the percentage rate which is given per annum on the data source. Once we have converted the risk-free rate into a monthly rate and subsequently transformed it into logarithmic form, subtraction over the monthly log total return provides us with a consistent estimate of the monthly excess returns.
Additionally, we aim to include in our analysis RRt+1 as real returns. That is, log markets return in excess of log inflation. Once the data on Consumer Price Index (CPI) is found on the OECD website, the monthly log inflation rate is subtracted to the monthly log total return. Real returns are used because it is a convenient way of taking into account the effects of inflation and monetary policy. There is a belief that inflation is higher under Democrats in the US according to some political macroeconomists. Likewise, Fama (1981), Geske and Roll (1983) find evidence of the effect of inflation on stock market returns.
B. Political variable:
In our regression analysis, data between left- and right- leaning parties returns are separated by the use of a political dummy variable that takes on the value of a 1 when right-leaning parties are in office (i.e. =1), 0 otherwise (i.e. =0). In the French context, the political variables are:
• =0 if a left-leaning party is in office at the time; • =1 if a right-leaning party is in office.
Figure 1 below shows the different parties that have been in office since 1988 and
identifies their political affiliation on a linear scale. Left-wing parties are placed on the left of the linear scale while right-wing parties are placed on the right. When =1, the returns include the right-wing governance represented by the LREM and the UMP party. On the other hand, =0 indicates that the president is PS affiliated. The dummy has been created by assigning a value of 0 for each monthly return in the data spreadsheet that are under left-wing presidencies, while associating a value of 1 in case of right-left-wing presidencies. LREM is often defined as centrist but the policies implemented make the president closer to a form of neo-liberalism which we define on the right-leaning part of the scale. Elections dates and information about French presidencies were publicly available on the website of the French presidency information page. Also, no coalition government were observed during the timespan covered.
Figure 1. Representation of the French parties in office since 1988 on a linear scale.
C. Control variables:
In our OLS regression analysis, omitting an explanatory variable which is related to another independent variable and that affects our dependent variable leads to the omitted variable bias. As a result, the dependent variable would be correlated with the error term. Therefore, control variables must be included to prevent misspecification of the regression model and to produce unbiased and consistent estimates. The correlation matrix shown below outlines the important relationships of the selected variables included in the study:
We follow Santa-Clara and Valkanov’s (2003) choice for control variable as the authors believe they are uncontroversial since they have been extensively used in previous studies. First, we use the dividend yield of firms DYIELD t as control variable. Fama and French (1998) previously investigated whether dividend yields forecast stock returns and find significant evidence on the effect of dividend yield’s effect on market returns. In a similar way, Shiller (1984) finds that investor’s reaction to company announcements such as dividends or earnings news has an effect on stock valuation and market returns. For this reason, dividend yield DYIELD (Dividend/ Price) is a suitable control variable as dividends have shown to be a determinant of demand for a company’s stock and thereby affecting its value.
Additionally, we also supplement our regression equation with a control variable to account for the effects of international markets, specifically the S&P500 t. it is reasonable to suggest that worldwide market volatility affects European stock markets. In line with Cahan et al. (2005) in their study of the presidential puzzle in the context of New-Zealand, we control for worldwide volatility by introducing in our regression the S&P 500 index. the U.S market index variable controls for variability in local market return due to the effect of a larger U.S. market. We then gather data from the CRSP database and select value-weighted returns including distributions. The value-weighted index depicts a more representative image of the financial markets by taking into account the weights of each index constituents.
Additionally, Santa-Clara and Valkanov (2003) use the macroeconomic variable term spread TSPR t to control for business cycles, which is the difference between 10-year treasury bond and three-month Treasury-bill rate. The annual rates in France is found on the OEDC website. After gathering the different data on short-term rates (i.e. 3months) and the long-term rates (i.e. 10 years), we transform the annual rates found on the data source into
monthly rates. Once we have subtracted the monthly short-term bond rate from the monthly long-term bond rate, we get an estimate of the term spread in France. The variable term
spread shows usefulness in predicting stock return and has been widely used as a control variable when analyzing market returns (Campbell, 1987).
To test our second hypothesis, we search whether volatility might explain the differential in returns by running a t-test by group with unequal variances. To do so, we use the volatility measure of the FTSE France index on FactSet as our dependent variable VOL t. The subscript t is not lagged as we assume that the incorporation of new information and induced volatility is rapidly translated in asset prices. Furthermore, the total number of observations is different from the market returns as data was limited on the FactSet website. The volatility is measured in 377 months (n=377) and covers the period from 1988:11 to 2020:03.
Table 2. Summary Statistics of Financial and Control Variables. The table reports the number of observations n (Obs), the sample average (Mean) and the standard deviation (Std.Dev.). All the returns (RET, ER,
III. Results and analysis
Figure 1 below gives the average excess return ER and real returns RR for each presidency. The data suggests that annualized returns were on average negative for Nicolas Sarkozy (UMP) and the current President Macron (LREM), respectively resulting in negative real returns of -4.46% and -1.99% and negative excess return of -4.71% and -1.27% per year. This is rather surprising considering the pro-business policies that Sarkozy and Macron advocated for. It is worthwhile to say that the financial crisis of 2007-2009 occurred during Sarkozy’s mandate. Jacques Chirac (UMP) is the only right-leaning president that
experienced average positive returns since 1988. Interestingly, François Hollande (PS) is the president that produces the highest average returns among all the presidents in the period studied. Indeed, during his 5-year mandate, markets yields reached 6.88% per year in excess returns and 6.61% in real returns. The U.S. data also present similar negative excess returns patterns for 4 out of 18 U.S. presidents (e.g. Nixon/ Ford with negative 5% excess returns). Of the four presidents experiencing negative excess returns, three of them are affiliated to the Republican party.
Figure 2. Annualized excess and real returns observed under each president since 1988. ER represent excess returns and RR as real returns. In Mitterrand’s presidency, large interest rates are observed. This trend has harmonized in the remaining presidencies. The two lines indicates the averages over the whole
To test our first hypothesis and confirm whether returns are higher under left-wing presidencies, we run an OLS regression equation without the control variables:
(1) FTSE t +1 = t + t + t+1
The Intercept t shows the average return, either in the log total return, log excess return or log real return under left-wing parties. On the other hand, ( + ) shows the average log total, log excess or log real return under right-wing parties. If there are no differences in returns between left- and right-wing parties, we should expect = 0, thus implying ( + ) = . Inversely, if returns are higher under left-wing parties, we expect 0. Our results indicate that total returns of left-wing president average 6.16% per year since 1988 (see Table 3 below for estimates and significance). The rates provided are the monthly coefficients which are annualized. Interestingly, total returns are 3.73% lower under right-wing parties (i.e. for UMP and LREM) with total returns averaging 2.43% per year. However, results are not significant at p<.05 and p<.10. A t-test by group with unequal variance is performed to find if the left-wing constituencies produce significantly higher total returns and we find no evidence at 5%, while significant at 10% for RET and RR. When analyzing excess returns and real returns, we find similar evidence with left-wing parties averaging 3.90% and 5.35% market returns per year whereas right-wing parties obtain 2.72% and 3.63% lower market returns on average per year. While the differential in returns are higher in the U.S, these results seem to concur with Santa-Clara and Valkanov’s (2003) finding in the American context as the PS produces larger market returns Generally speaking, it is seen with > ( + ) for each type of returns (log total, log excess and log real returns).
Table 3. Annualized Average Returns Estimates under Left and Right-Wing Presidents Using Simple OLS Regression.
Second, we introduce in the regression our control variables to account for external effect that could potentially be also affecting market returns. By, including these variables, we reduce the risk of biased estimators and inconsistent standard errors. Therefore, the following model is estimated:
(2) FTSE t +1 = t + t + 1DYIELDt + 2TSPRt + 3SP500t+1 + t+1
The addition of the control variables accentuates the coefficients previously observed. Indeed, the total log returns, log excess returns and log real returns are
respectively 6.84%, 4.47% and 6.01% for left-wing parties and 2.12%, 0.89% and 1.39% for right-leaning parties. The results are significant at 5% level for each type of returns studied (see table 4 below for monthly estimates). When using total returns as the dependent variable, the extension of the model has increased the R² from o.45% to 51.59%, suggesting that our multiple regression model is a better fit and has enhanced explanatory power. Although the differential is less pronounced in the French market, results concur with the
findings between Republicans and Democrat in the U.S. Our first hypothesis is therefore confirmed, the returns (total log, log excess return and log real returns) under left-wing government are consistently higher than right-wing constituencies between 1988 to 2020 in France.
Next, to unveil if the presidential puzzle occurs in France, the volatility observed under the PS should be higher than the volatility under UMP and LREM presidencies. To answer this, we run a t-test by group (=0 and =1) with unequal variances to confirm if higher risk environment could be the reason why the returns diverge between political constituencies. However, the results show no significant difference of volatility (diff = -.003, t = -0.605). Therefore, not rejecting the null hypothesis leads to the conclusion that volatility is not statistically different between the two opposing political constituencies. In alignment to the findings in the U.S., higher volatility is not a valid explanation of higher returns for left-leaning parties since similar volatilities are observed between political constituencies. Because higher returns were consistently produced under the PS since 1988 and these higher returns were not explained by the volatility of the markets, evidence suggest the existence of the presidential puzzle in France.
Table 3. Monthly Average Returns Estimates under Left and Right-Wing Presidents Using Multiple OLS Regression.
We finally extend our research to a Probit regression to address potential of reverse causality. Taking the political dummy variable as our dependent variable and market returns as our independent variable, we investigate whether market returns increased the likelihood of a specific party in place:
(4) Pi= Pr (t+x =1|FTSE t) = Φ (β0+β1FTSE t) t = {1 for right − wing party
0 for left − wing party x = {6 𝑚𝑜𝑛𝑡ℎ𝑠
12 𝑚𝑜𝑛𝑡ℎ𝑠
The probability Pi represents the likelihood of having a right-leaning president (=1). Also, the dependent variables are lagged in two different time horizons (i.e. half-year and one-year horizon). The use of different time horizons allows us to test whether there the likelihood of having a specific party increases over time based on current market conditions. In terms of total logarithmic returns, excess logarithmic returns and real logarithmic returns, a negative coefficient β is observed while not statistically significant (See table 3 below for estimates and significance). The negative relationship indicates that as the returns of the markets increases, the likelihood of having a right-leaning government is diminished. This could show that when market returns are higher, people will lean towards left-wing constituencies while on the other end, recession could lead to the election of a right-leaning candidate. For instance, assuming a current average total return of the market of 3%, we estimate that if an economic boom occurs with a 7% sudden upward shift in total market returns, the
probability of having a right-leaning government in the next 6 months is subsequently reduced by 6.11% (see appendix for computation). For the 12-month horizon, the likelihood of a right-leaning president is reduced by 4.47%. While these results are not significant, they nonetheless show the possible endogeneity of our variables with an inverted causal
Table III. Probit Regression Results estimating the Likelihood of a Right-leaning Government in Response to Market Conditions in Different Time Horizons in France. The table reports the results of
the regressions Pi= Pr (t =1|FTSE t) = Φ (β0+β1FTSE t). The result below the coefficient represent the z-values.
IV. Discussion
Initially, the expected results of the study were unknown since previous results have shown the existence of the puzzle in the U.S. while New-Zealand did not produce similar patterns. Interestingly, the results in France indicate higher returns under left-wing presidents which confirms the findings between Democrats and Republicans in the U.S. Furthermore, preliminary evidences of the presidential puzzle in France was confirmed since the volatility of the FTSE index between the political parties were approximately equal.
While this slightly diverts from Santa-Clara and Valkanov’s (2003) findings as they find significant evidence of lower volatility under Democrats, our results do however lead to the same conclusions. That is, by not having a difference in volatility, this would suggest that higher return pattern under the PS is also not explainable since volatility is approximately the same between the different political constituencies. Therefore, the puzzle also occurs in France, at least since 1988.
In general, one would think that right-wing parties are the typical parties of business since their policies often support the supply-side of the economy (e.g. Macron’s liberal measures). In reality, left-leaning presidents have consistently experienced higher returns than right-leaning presidents. A possible explanation could lie in the trade-off that
governments face. Phillips’ (1958) model provides insights on the goals of government and demonstrates the trade-off that governments in place face between unemployment and inflation at the domestic level. That is, when left-leaning parties put emphasis on policies against unemployment, they consequently face an increase in inflation. Furthermore, the rise in inflation leads to a lower real interest rate environment which, in theory, would lead to a rise of business activity (increase of aggregate investments). As economy is booming, market valuations are enhanced yielding higher nominal returns. Therefore, log total returns, log real returns and log excess return are then expected to increase. Based on this reasoning, one would expect left-wing government such as the PS to yield favorable returns, at least on the short-term. Of course, while this remains theoretical and evidences remain limited, it remains nevertheless an insightful economic outlook on the potential reasons that could explain the higher returns observed under left-wing presidencies.
After finding evidences of the presidential puzzle in France since 1988, the Probit regression posit the possibility that political parties may not be the driving force of returns. Instead, it could simply be a “proxy” for another unknown variable that actually causes the higher returns. For instance, the current economic condition or business cycle might be the reason of higher returns observed under left-leaning parties. In line with the findings of
Weatherford (1978) and Brücker and Grüner (2010), we also find that the likelihood of having a right-leaning president increases with decreasing market returns, although not significantly. It seems that individuals are strongly rooted in the idea that right-wing
governments are beneficial to the economy when a downturn is persistent and the likelihood of voting for a far right-wing party therefore increases. However, our study shows that market returns averages are higher under the PS and contradicts the idea that right-wing parties are value enhancing, at least in the U.S since 1927 and in France since 1988. Fear of persisting recession, the increasing need of radical change and the political game could be the origins of the trend observed.
Lastly, it is also reasonable to think that current economic well-being influences incumbent on the next elections. If a president does poorly and the economic situation is worsening, then the likelihood of being re-elected at the following elections is meager. Inversely, a president whose term is brightened by staggering economic figures is likely to be appraised and subsequently re-elected at the coming elections. In France, only Francois Mitterrand and Jacques Chirac were re-elected since 1988. The higher returns observed under their respective mandates corroborates the idea that incumbents are more likely to be re-elected if market returns produce high yields. Curiously, Francois Hollande’s market’s average yield was the highest, but his unpopularity tarnished the impressive returns observed. According to TNS Sofres, a French research institute, only 24% of French
V. Conclusion
The intent of this paper was to find evidence of the presidential puzzle in France. First, we wanted to test the hypothesis that market returns are higher under left-leaning parties. Regression (1) and (2) confirmed that log total returns, log excess returns and real returns were on average higher under the PS in France since 1988. Furthermore, to confirm the existence of the puzzle, we have performed a t-test by group between political
constituencies by using the FTSE volatility measure and find that the volatility between left-and right-wing parties in France are not significantly different. That is, one could not explain the higher return under the PS observed in France considering that the market is deemed efficient. We are therefore left, with the occurrence of a presidential puzzle in France and this comforts the findings in the U.S. Lastly, we address the potentiality of economic environment as predictor of political outcomes and allows us to expand the scope of the problem by addressing it in a multi-dimensional way. Interestingly, our results show that higher returns increase the likelihood implying that recession would favor right-wing parties.
Overall, our results have shown to be promising for the unveiling of the presidential puzzle in France. The current literature for countries outside the U.S is meager and finding the occurrence of such phenomenon in outside financial markets is beneficial to a wide variety of economic agents. Not only do they provide us with some insights on the mechanism of asset-pricing in response political information, they sharpen our
understanding of financial markets in general. Additionally, our study extended the possible path to explain the differential in returns, and the response might be by reversing the causal arrow. Indeed, papers suggested that economic conditions deemed to be a valid predictor or political extremism. However, results are not significant in our study. Overall, the
conjunction of the finance motivated theory (politics affecting returns) and the politically motivated theory (returns affecting politics) led to valuable insights on a problem that is after all, a puzzle.
Having that said, in trying to understand the puzzle, uncertainties remain, and possible unobserved factors could be missing. The fact that political variables might be merely proxying a real missing explanatory variable, such as business cycles, is not to be excluded. Although The inclusion of macroeconomic variable TSPR in our multiple
regression did not alter the initial differential observed, the hypothesis that a business cycle is occurring is very likely, in theory and in practice. Lastly, misspecification of the regression models is not to be excluded. Previous authors (Campbell & Li, 2004; Sy and Al Zaman, 2011) showed the use of different models and different explanatory variables consequently altered the U.S. findings. Once again, the question whether the model is rightly specified is debatable and also applies to the current study. Additionally, due to the lack of reliable data, the study period only comprises the monthly returns since 1988. To fully confirm whether the presidential puzzle really applies in France, a larger sample of presidents is needed. Therefore, either reproducing the current study in the future or with a more extensive set of returns covering a larger period would be necessary.
Overall, understanding whether political party do influence our financial markets on the long-term horizon remains an open question. The effect of business cycles, the
endogeneity issues, the lack of comparable studies, the uncertainty of the model leaves us with numerous paths for further studies to tackle the puzzle.
Appendix
Assume a 3 % current monthly market return. Economic boom increases by 7% total market return (total return = 10%):
Time horizon = t+6 Pr (t+6 =1|RET) = Φ (0.2974 -2.4296 RETt) Change in Y = Φ (0.2974 -2.4296 * 3%) – Φ (0.2974 -2.4296 * 10%) = Φ (0.37) – Φ (0.54) = (0.64431) - (0.7054) = -6.11% Time horizon = t+12 Pr (t+6 =1|RET) = Φ (0.2788 -0.17405 RETt) Change in Y = Φ (0.2788 -0.17405 * 3%) – Φ (0.2788 -0.17405 * 10%) = Φ (0.33) – Φ (0.45) = (0.6297) - (0.6744) = -4.47% References
Bruckner, Markus, and H. P. Gruner. “Economic Growth and the Rise of Political Extremism: Theory and Evidence.” SSRN Scholarly Paper. Rochester, NY: Social Science Research Network, March 1, 2010.
Li, Canlin, and Sean D. Campbell. “Alternative Estimates of the Presidential Premium.” Board of Governors of the Federal Reserve System (US), 2004.
Campbell, John Y., and Robert J. Shiller. “Stock Prices, Earnings, and Expected Dividends.” The
Cahan, Jared, Christopher B. Malone, John G. Powell, and Udomsak Wong Choti. “Stock Market Political Cycles in a Small, Two-Party Democracy.” Applied Economics Letters 12, no. 12 (2005): 735–740.
Fama, Eugene F. “Stock Returns, Real Activity, Inflation, and Money.” The American Economic
Review 71, no. 4 (1981): 545–565.
Herbst, Anthony F., and Craig W. Slinkman. “Political-Economic Cycles in the US Stock Market.”
Financial Analysts Journal 40, no. 2 (1984): 38–44.
Huang, Roger D. “Common Stock Returns and Presidential Elections.” Financial Analysts
Journal, 1985, 58–61.
Gaertner, Fabio B., Jeffrey L. Hoopes, and Braden M. Williams. “Making Only America Great?
Non-Us Market Reactions to Us Tax Reform.” Management Science 66, no. 2 (2020): 687–
697.
Imai, Masami, and Cameron A. Shelton. “Elections and Political Risk: New Evidence from the 2008 Taiwanese Presidential Election.” Journal of Public Economics 95, no. 7–8 (2011): 837–849.
Knight, Brian. “Are Policy Platforms Capitalized into Equity Prices? Evidence from the Bush/Gore 2000 Presidential Election.” Journal of Public Economics 90, no. 4–5 (2006): 751–773.
Phillips, Alban W. “The Relation between Unemployment and the Rate of Change of Money Wage Rates in the United Kingdom, 1861–1957 1.” Economica 25, no. 100 (1958): 283–299.
Riley, William B., and William A. Luksetich. “The Market Prefers Republicans: Myth or Reality.”
Journal of Financial and Quantitative Analysis 15, no. 3 (1980): 541–560.
Geske, Robert, and Richard Roll. “The Fiscal and Monetary Linkage between Stock Returns and Inflation.” The Journal of Finance 38, no. 1 (1983): 1–33.
Sénécat, Adrien. “Les mesures de la première année d’Emmanuel Macron avantagent-elles vraiment les plus riches ?” Le Monde.fr, May 5, 2018. https://www.lemonde.fr/les- decodeurs/article/2018/05/05/les-mesures-de-la-premiere-annee-d-emmanuel-macron-avantagent-elles-vraiment-les-plus-riches_5294884_4355770.html
Sy, Oumar, and Ashraf Al Zaman. “Resolving the Presidential Puzzle.” Financial Management 40, no. 2 (2011): 331–355.
Wagner, Alexander F., Richard J. Zeckhauser, and Alexandre Ziegler. “Company Stock Price Reactions to the 2016 Election Shock: Trump, Taxes, and Trade.” Journal of Financial
Economics 130, no. 2 (2018): 428–451.
