Should Investors Pay Attention to Domestic and US Election Regimes ? A Canadian Perspective

Based on Canadian historical returns from 1951 to 2011 and mean-variance frontier analysis, we document better stock market opportunities in the late versus early part of the mandates of the Canadian federal governments or the American presidents, as well as when Democratic versus Republican American presidents are in power. Better bond market opportunities are found in majority versus minority Canadian parliaments and in Conservative versus Liberal federal governing parties. We investigate the role of controls for the state of the economy to explain these results. We conclude that both domestic and American electoral regimes significantly affect investment opportunities and optimal asset allocation.


Introduction
In countries with a democracy, political parties spend large amounts of time, money and effort to convince electors that their policies are the most appropriate for their country.In particular, they argue that their economic policies are the best for the finances and growth of the country, that they are the best equipped to foster solid relations with their international partners and that they should be given "strong" mandate to operate efficiently and reduce uncertainty.Political analysts and economists regularly comment on these claims and further analyze the outside influence of foreign politics.The media and ultimately many citizens show tremendous interest for elections and their results.Should investors pay attention to election regimes?
In the United States (US), there is a growing academic literature that answers 'yes' to this question by looking at the relationship between electoral regimes and returns.For examples, Huang (1985), Hensel and Ziemba (1995), Johnson, Chittenden and Jensen (1999), Santa-Clara and Valkanov (2003) and Booth and Booth (2003) show that large and small-capitalisation equities yield higher returns under Democratic presidencies and in the last two years of a presidential term, while US Treasury bonds and bills produce higher returns under Republican presidencies.As no corresponding differences in volatility or macroeconomic conditions are found, the "Democratic equity premium" and "presidential cycle effect" have been called puzzles, although explanations based on a longer-run analysis (Beyer, Jensen, & Johnson, 2004), international comparison (Bohl & Gottschalk, 2006), spurious econometrics problem (Powell, Shi, Smith, & Whaley, 2007, 2009) and time-varying risk premiums (Sy & Zaman, 2011) have been proposed.
This study extends this literature by examining the question for another country, namely Canada.Apart from the importance of a clear answer for investors in Canadian capital markets, it makes two contributions.First, by focusing on Canada, a country similar to the US for the stability and functioning of its capital markets, and with a political system that has similarly resulted in only two parties being in power (a left-leaning one and a right-leaning one), this study provides a useful out-of-US-sample check on the US results.Second, and more importantly, this study gives a novel assessment of the outside influence of American politics on foreign capital markets by investigating whether US election effects spill over to the capital markets "north of the border".Canada is a natural choice for detecting such influence as it has strong ties to the US, being its most important economic and political partner in the last century.
Specifically, using monthly returns on Canadian bills, bonds and stocks from 1951 to 2011, this paper investigates five sub-questions.Are investment opportunities different in: 1) Left-leaning Liberal versus right-leaning Conservative governments?
3) The early versus late parts of the federal mandates?4) Left-leaning Democratic versus right-leaning Republican presidential administrations?
5) The early versus late parts of the American presidential mandates?
We answer these questions about investment opportunities in different electoral regimes with a traditional mean-variance analysis.Using bonds and stocks, we compute the investor opportunity set, which is delimited by the mean-variance frontier, conditional on the electoral regimes.We then evaluate the Sharpe ratio performance of the individual assets and optimal portfolios, and formally test for the equality of Sharpe ratios across regimes.While there are many other performance measures, the Sharpe ratio has a long history of relevancy and is the most natural measure to complement mean-variance analysis.Due to its simplicity and intuitive appeal, it is widely used both in practice and in academic studies.Next, we check the robustness of the results with controls for the state of the economy.Finally, we examine the optimal asset allocation between bills, bonds and stocks across regimes by computing the asset weights for selected optimal portfolios.For Canada, Foerster (1994) and Chrétien and Coggins (2009) are the only two references that provide some evidence on these questions.Focusing on estimates of expected return and standard deviation, they document a "prime ministerial cycle effect" as well as a Democratic equity premium and a presidential cycle effect in Canadian stocks, but no robust "Liberal equity premium" or minority government differential.This paper expands on their results by considering a more complete dataset and by looking at mean-variance frontiers, Sharpe ratios and optimal asset allocations across regimes, offering a clearer overall picture of investment opportunities.It also puts more emphasis on the impact of American political regimes on Canadian investments.
The empirical results indicate that investors and portfolio managers should pay close attention not only to their domestic electoral regimes, but also to the American ones.With respect to the domestic regimes, the Canadian investment opportunities are significantly better in Conservative versus Liberal governing parties, although mainly a reflection of the strength of the bond market.This is consistent with Alesina and Sachs (1988), who suggest that leftist parties generate higher inflation, and with Booth and Booth (2003), who document a similar finding for the US bond market.While this difference is robust to the use of Canadian information variables to control for the state of the economy, it does not subsist when US information variables are included as controls.For the stock market, in contrast to the findings of the US literature, there is only weak evidence that it yields better opportunities under a left-leaning leadership, although the difference becomes significant when we use US information variables as controls.The exclusion of minority parliaments also reinforces these results.
The Canadian investment opportunities are also better in the late parts of the federal election cycle than in the first two years, with significantly higher Sharpe ratios for the bond market, the stock market and the optimal portfolio.With an optimal Sharpe ratios more than three times higher in the late versus early mandates (0.955 versus 0.311), these results confirm a prime ministerial cycle effect stronger than the presidential cycle effect in US returns, perhaps because Canadian governments have the additional option of calling an election at the "right moment", as election dates are not fixed in our sample period.
With respect to the American regimes, we find that Canadian investment opportunities are significantly better in Democratic versus Republican administrations.The performance spread comes entirely from the stock market and is particularly striking: The value-weighted portfolio of stocks earn a Sharpe ratio of 0.830 in Democratic regimes versus -0.007 in Republican ones.Having ideologically aligned leaderships in Canada and the US preserves but does not reinforce the effects found independently.The Canadian stock market also performs significantly better in the late versus early parts of the US presidential cycle.Furthermore, stocks yield even worst risk-adjusted returns when the Canadian and US governments are simultaneously in the early parts of their mandates, with a negative Sharpe ratio of -0.385.Thus, the puzzling Democratic equity premium and presidential cycle effects strongly spill over "north of the border".Hence, we document an important outside influence of American politics on Canadian capital markets that is consistent with the ties between both countries, as exemplified by correlations varying from 0.7 to 0.8 between their capitals markets.
These findings on the significant effects of the Canadian election cycle, the US President party and the US US election cycle: The EARLY US regime includes the months in the first two years after a fixed-date US presidential election, while the LATE US regime has the months in the last two years of the four-year mandate.Thus, about half the months are in each regime.

Measures of the Investment Opportunities
The Canadian investment opportunities we investigate are composed of four assets representing three common financial asset classes: bills (denoted RF, based on three-month Treasury bills), bonds (denoted RGOV, based on long-term government bonds) and stocks (two assets denoted RVW and REW, based on respectively value-weighted and equally-weighted portfolios of all exchange-traded stocks).While RVW is similar to the S&P/TSX Composite Index and is highly weighted in large-cap stocks, REW can be thought as representing the small-and medium-cap equity asset classes as they are the dominant portfolio components.The series of monthly realized returns for these assets are obtained from the TSX Canadian Financial Markets Research Centre (CFMRC).
Table 1 presents the annualized mean return (monthly value 12), the annualized standard deviation (monthly value √12) and correlations for the four assets (Note 2).The historical risk-reward opportunities look good compare to the ones of the last few years.More importantly for our purpose, we can observe a risk-return trade-off between the assets as expected.Note.This table presents the annualized means, standard deviations (St Dev) and correlations (Corr) for the monthly returns of three-month Treasury bills (RF), the long-term government bonds (RGOV), the value-weighted equity portfolio (RVW) and the equally-weight equity portfolio (REW).
To examine how the investment opportunities vary across electoral regimes, we rely on mean-variance (MV) analysis.Specifically, we first estimate the means, standard deviations and correlations, conditional on being in a given regime.We then compute the corresponding MV frontiers, which represent the limit of the investors' opportunity set.The MV frontier is defined as the portfolios that have minimum variance for a given mean return.
Using the notation and demonstrations of Roll (1977), let be the 3 × 1 vector of mean returns for RGOV, RVW and REW, let be their corresponding variance-covariance matrix, and let be the unit vector.Then, for a given mean return of , the MV frontier portfolios have the following variance: where , and ′ .
Next, we compare the Sharpe ratios of the individual assets and of the MV tangency portfolio (the portfolio with the maximum Sharpe ratio) across regimes.The Sharpe ratio (Sharpe, 1966(Sharpe, , 1994)), also called the reward-to-variability ratio, is a portfolio's excess return over the risk-free rate divided by its standard deviation, This commonly used performance measure is intuitively interpreted in the mean-standard deviation space as the slope of a line from the risk-free asset to a specified portfolio.The higher is the slope, the better located is the portfolio.The highest possible slope leads to the MV tangency portfolio, with return denoted , which has the following maximum possible or optimal Sharpe ratio: (3) where: (4)

2
(5) As discussed by Ferson and Siegel (2003), the sample maximum Sharpe ratio is biased upward when the number of assets is large relative to the number of observations .We thus report adjusted maximum Sharpe ratio by using their proposed correction: In this paper, while varies depending on the regimes under consideration, we only consider three assets ( 3) for the MV frontiers, which implies that the bias is small.Furthermore, our tests focus on the difference between Sharpe ratios across regimes, which mitigates the effect of the bias further.Hence our results are similar whether we use adjusted Sharpe ratio or not.
We formally test for the equality of Sharpe ratios across regimes with a statistic proposed by Jobson and Korkie (1981), revisited by Lo (2002) and Memmel (2003).(See also Ledoit & Wolf, 2008, Leung & Wong, 2008, for further discussions.)Specifically, let , et , be the estimated Sharpe ratios for portfolio in regimes and , two mutually exclusive regimes (for example MINOR and MAJOR).These estimates are obtained by using the sample mean and standard deviation of the portfolio returns in the two regimes, with samples of and observations, respectively.Then, under the null hypothesis that the Sharpe ratios are equal, and assuming that returns are identically and independently distributed, the estimated Sharpe ratio difference has the following asymptotic distribution, which allows a test on its significance: This distribution leads to a z-test on the equality of Sharpe ratios across regimes.When applied to the maximum Sharpe ratio portfolio, this test becomes a test on the equivalence of the optimal MV opportunities across regimes.
As a robustness check, following Santa-Clara and Valkanov (2003), we also use predetermined information variables to control our results for the anticipated state of the Canadian economy.Specifically, we run the following regression with the returns of each risky asset class (RGOV, RVW and REW): where is a vector of state of the economy control variables that have been demeaned.Then, we re-estimate the measures of investment opportunities using a new set of returns that exclude the effect of the information variables: (9) Our information variables are Canadian versions of the ones used in the study of Santa-Clara and Valkanov (2003), namely the annualized log dividend-price ratio, the term spread, the default spread and the relative interest rate (Note 3).The predictive value of these variables has been studied extensively in the literature, starting with Keim and Stambaugh (1986), Campbell (1987), Campbell and Shiller (1988), and Fama andFrench (1988, 1989).Rapach, Wohar and Rangvid (2005), Hjalmarsson (2010) and Chrétien and Coggins (2014) provide evidence in a Canadian context.To account for the possibility that the Canadian information variables might be governme (Note 4).2, the optimal Sharpe ratio is significantly higher for left-leaning leaderships.Overall, having ideologically aligned leaderships preserves but does not reinforce the effects found previously.Third, in Panel C of Table 7, the effect of the combined Canadian and US election cycles is stronger than each effect taken separately, especially for the stock market.For example, RVW has Sharpe ratios of 0.56 when both countries are in late mandate and -0.39 when they are in early mandate, a difference statistically significant at the 1% level.The stock market tends to perform very poorly when both countries are simultaneously in the first two post-election years.

Controls for the State of the Economy
Our analysis has thus far documented numerous significant Sharpe ratio differences between the electoral regimes.Following Santa-Clara and Valkanov (2003), this section uses predetermined information variables to control our results for the anticipated state of the Canadian economy, using the methodology described previously.Table 8 documents that most of our results are robust to controls for the state of the economy.Our findings on the effect of the Canadian governing party regimes are the most affected by the controls.In particular, compare to the ones in Table 2, the results in Panel A of Table 8 show that the higher Sharpe ratio for the bond market in Liberal versus Conservative governments is robust to Canadian controls, but not to US controls.The addition of US controls also renders the Sharpe ratio of the stock market under Liberals significantly higher than the one under Conservatives.
With respect to electoral strength, Panel B of Table 8 confirms that the bond market performs better in majority versus minority governments, although the statistical significance weakens with the use of US controls.While the other Sharpe ratio differences were not significant in Table 3, the inclusion of Canadian or US controls makes the stock market and optimal Sharpe ratios significantly higher in minority versus majority governments.For the stock market, it thus appears that, after controlling for the state of the economy, there is a higher reward per unit of risk in minority governments.
As the rest of Table 8 shows, whether using Canadian or US control variables, our main findings on the Canadian election cycle (Panel C), the US President party (Panel D) and the US election cycle (Panel E) are robust, so that they do not appear to have been expected due to measurable business cycle variations.In particular, the stock market has significantly higher Sharpe ratios in the late versus early part of the Canadian or While historical in nature, and thus subject to the difficulty of extrapolating from past returns, our performance results involve investable, low turnover, portfolio strategies, using start-of-the-month information to invest for the month.As no apparent variation in business cycle risk accounts for the results, the large differences in returns per unit of risk that we document are somewhat puzzling.Given that electoral information is public and easily available, the efficient market theory states that investors should not be able to profit from it, yet portfolio managers following some of the electoral signals would have made important gains.Since rational explanations for our results are not well developed and are left for future research, it remains to be seen if such opportunities will materialize again in the coming years.

Notes
Note 1.During our sample period, there is no fixed election date in Canada.While the dissolution of Parliament can occur at any time within the five-year electoral mandate, the shortest observed majority term is 41 months, a longer period than the longest observed minority term.In 2007, the Parliament adopted a bill for fixed election dates, with the first scheduled fixed date set for 2015.
Note 2. For simplicity, we assume that RF has a standard deviation of zero and correlations of zero with the other assets.These assumptions are commonly made and do not affect our results, given that three-month Treasury bill returns have a standard deviation and correlations with other assets that are close to zero.
Note 3.More precisely, the Canadian information variables are constructed as follow.The annualized log dividend-price ratio is the difference between the one-year total return of the S&P/TSX Composite Index and its one-year price return, multiplied by the value of the index one year ago, and divided by its current value, with data from CFMRC.The term spread is the difference between the average yield-to-maturity of the Canada Treasury bonds with a maturity of ten years or more (CANSIM series V122487) and the yield-to-maturity of the three-month Canada Treasury bill (CANSIM series V122541).The default spread is the difference between the yield on long-term corporate bonds and the government bond yield from series V122487.To construct a long history for the corporate bond yield, we combine three different series.From February 1950 to October 1977, we use the CANSIM series V35752, the Scotia-McLeod Canada Long-Term All-Corporate Yield Index.From November 1977 to June 2007, we take the Scotia Capital Canada All-Corporations Long-Term bond yield CANSIM series V122518.From July 2007, we take the yield from the Merrill Lynch Canada Long-Term Corporate Bond Index (F9C0) from Bloomberg.The relative interest rate is the deviation of the three-month Treasury bill rate from its one-year moving average using the previously introduced series V122541.
Note 4. The US variables are defined as in Santa-Clara and Valkanov (2003).They are the annualized log dividend-price ratio on the S&P 500, the term spread between yields of 10-year US Treasury notes and three-month US Treasury bills, the default spread between yields of BAA-and AAA-rated US bonds, and the relative interest rate given by the deviation of the three-month US Treasury bill rate from its one-year moving average.The data are available on the Professor Goyal's website, with the sources described in Welch and Goyal (2008).
Note 5.It is useful to observe that our approach is not designed to capture the abnormal returns associated with changes in regimes, which would necessitate, for example, an event study methodology.Instead, it focuses on measuring the risk-return opportunities and asset allocations across various electoral regimes, exploiting the large number of observations in each regime for more precise estimation.A study of the market responses to election events is beyond the scope of the paper.
figure shows the B) versus Conser vative (Liberal) re
Figure 5 the Cana conseque 11.2% an market p investmen higher un fall outsid

(
LIB) versus Conservative (CON) governing party regimes in majority (MAJOR) parliaments (Panel A), for the governing party / US president party joint left-leaning Liberal / Democratic (LIB-DEM) versus right-leaning Conservative / Republican (CON-REP) regimes (Panel B), and for the joint Canada-US late (LATE-LATE) versus early (EARLY-EARLY) election cycle regimes (Panel C), and p-values on tests for the equality of Sharpe ratios across regimes (Diff p-val).***, ** and * indicate significance at the 1%, 5% and 10% levels, respectively.The number of observations (T) for each regime is in parentheses.
, while the LATE regime has the other months.The restriction to majority months ensures governmental control on policy implementation and election calling in the early part of the mandate, creating conditions where "tough" long-term policies could be implemented more easily.With such a definition, 34.6% of the months fall in the EARLY part of the mandate (Note 1).US President party:The DEM regime regroups the months under a Democratic President, while the REP regime includes the months with a Republican President.Democratic and Republican Presidents are in the White House for 41.0% and 59.0% of the months, respectively.

Table 7 .
Additional results on sharpe ratios and electoral regimes Panel A. Governing party in majority electoral strength regime Note.This table presents Sharpe ratios for RGOV, RVW, REW and their MV efficient frontier tangency portfolio (Maximum) for Liberal

Table 8 .
Table 8 presents the results when considering either Canadian controls (left side of the table) or US controls (right side of the table)as instruments for variations in the Canadian business cycle.Sharpe ratios and electoral regimes with controls for the state of the economy This table presents Sharpe ratios for RGOV, RVW, REW and their MV efficient frontier tangency portfolio (Maximum) for different electoral regimes after controlling for the state of the economy, and p-values on tests for the equality of Sharpe ratios across regimes (Diff p-val).The variables for the electoral regimes are described in Tables2 to 6.The control variables are the dividend-price ratio, term spread, credit spread and relative interest rate.The results under Canadian and US Controls use respectively Canadian and US versions of the variables.***, ** and * indicate significance at the 1%, 5% and 10% levels, respectively.
US presidential election cycles.Controls for the state of the economy do not explain these last three findings.