Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
Elections, Political Control and Duration of Stock Market Cycles
Fan Wang
Abstract
This study uses duration analysis to investigate whether the occurrence and the outcome of
presidential elections can affect the timing of turning points of the U.S. stock market cycle.
Results suggest that political alignment between the White House and the Congress (political
control) plays an important role in the timing of turning points relative to the elections.
Specifically, our estimates show that the increased hazard for a bear market prior to an election
predicted by our Hypothesis One is only found to be significant when political control is absent
or when the incumbent is a Republican. However, the predicted rise in the hazard for a bull
market in the period immediately after an election is not found in our data. In fact, the estimates
reveal a smaller probability for a market peak to occur in the post-election period than at other
times if the party of president elected did not control the Congress. A further examination of the
post-election effects provides evidence that is consistent with our Hypothesis Two that there is
no difference in the likelihood of a peak or trough after the election of a Democratic president as
compared to other times. Furthermore, the estimates show that market troughs are less likely to
occur in the wake of a Republican presidential victory than at other times and political control
attenuates this post-Republican effect. Finally, although the implied post-Republican surge in the
hazard for a bull market is not found in our data, political control does seem to boost the hazard
for a bull market after an election of a Republican.
1. Introduction
"By the Law of Periodical Repetition, everything which has happened once must happen again, and again, and
again -- and not capriciously, but at regular periods, and each thing in its own period, not another’s and each
obeying its own law... The same Nature which delights in periodical repetition in the sky is the Nature which
orders the affairs of the earth. Let us not underrate the value of that hint." -- Mark Twain
Every four years, money and power collide at the intersection of Wall Street and
Pennsylvania Avenue. In fact, some of the most interesting historical patterns in the Stock
Trader’s Almanac relate to the occurrence and the outcome of presidential elections in the United
States. Over the years, academics and market pundits have conducted numerous studies on the
subject of the election cycles in an attempt to make a better prediction of future market trends. It
has been found that the U.S. stock market generally tends to ascend with a coming election and
descend once the election is over. Historically, returns in the U.S. stock market, on average, are
higher during Democratic presidencies than during Republican presidencies. Some suggest that
these periodic patterns in the stock market are the reflection of theories of political business
cycles in the real economy. It is widely accepted that real economy affects stock market as stock
prices reflect investors’ forecasts of the future state of economy and firm. Since a major factor
that affects real economy is the economic policies of an incumbent government, its strategic
decision for reelection and/or partisan preferences could and should affect stock market
movement. Theories of political business cycles predict that the quadrennial election cycle in the
United States could affect the timing of the peaks and troughs of the U.S. business cycles. For
instance, opportunistic political business cycle theories suggest that, compared to other times, a
business trough is more likely in the period before an election as an incumbent attempt to
maximize the chance of reelection, and a business peak follows soon after an election as pre-
election stimulus is reversed. Alternatively, partisan political business cycle theories suggest that
the likelihood of a peak or a trough following a presidential election depends upon which party
was victorious in an election. For example, rational partisan theory (Alesina 1987) suggests that
a business peak is more likely to occur in the wake of a Republican presidential victory, while a
business trough is less likely to occur after a Republican has won a presidential election than at
other times.
One critical task for all investors is to predict the future market direction and then decide
when to enter or exit the market. It is impossible for one to predict future market movement
correctly all the time, but that doesn’t mean the stock market doesn’t have patterns that could
meaningfully add to investors’ profits. That said history can be a good guide and a helpful tool in
considering entry and exit points in the stock market. The importance of the presidential
elections in the U.S. financial markets has long been recognized by financial intuitions. In a
client note, Goldman Sachs offers “3 Reasons Why US Investors Should Take Election Cycles
Very Seriously.” First, the political stakes in presidential, parliamentary, or legislative elections
often translate into changes in policies that can reshape the economic environment. Second, the
regularity with which elections take place in most countries may give place to cyclical patterns in
government and investment behavior. And third, elections can markedly increase political and
social uncertainty. These three factors have the potential to affect all asset classes, especially
equities, given their strong sensitivity to changes in the economic outlook. (Business Insider,
February 2012)
Given the close tie between real economy and stock market, it is natural to wonder will
the quadrennial elections can affect the timing of peaks and troughs of stock market cycles in the
United States as suggested by the PBC theories? Specifically, we are trying to answer the
following questions: Whether the U.S. stock market is more likely to reach a trough (the
beginning of a bull market) in period leading up to an election and the market is more likely to
reach a market peak (the beginning of a bear market) in period shortly after an election? Whether
there is any party difference in the hazards?
Previous studies on the subject of election cycles in the U.S. stock market only address
the timing issue indirectly by focusing on the amplitude of returns and volatilities before and
after the elections or across the tenure of different parties. In this paper, we provide a more direct
test of the temporal links between the political event and turning points in the U.S. stock market.
Duration analysis is used to test whether the likelihood of the occurrence of a turning point in the
U.S. stock market (that is, either the end of a bear market or the end of a bull market) can be
significantly affected by the occurrence and the outcome of a primary election.
Duration analysis is well suited for analyzing the temporal links between elections and
stock market turning points. It allows for directly testing the determinants of the likelihood of the
end of a market cycle phase in any period conditional upon the phase lasting up until that period.
The determinants of the timing of peaks and troughs that we focus on in this study are the
occurrence and the outcome of elections. Duration analysis enables us to estimate the effect of
elections on the likelihood of the end of a stock market cycle phase holding constant other
factors. In particular, duration analysis controls for duration dependence that arises when there is
a changing probability of the end of a stock market cycle as the cycle itself progresses.1
Results presented in the study suggest that political control for the Congress has an
important role in the timing of turning points relative to the elections. The pre-election surge in
the hazard for a bear market predicted by Hypothesis One is found to be significant only when
political control is absent for the incumbent. We do find, however, a significant increase in the
likelihood of the end of a bear market in the 24-month period and the 16-month period before an
election when the incumbent is a Republican, even without controlling for political control. On
the other hand, the predicted post-election surge in the hazard for a bull market in the period
immediately after an election is not found in our data. In fact, after controlling for political
control, the estimates indicate a smaller hazard for a bull market in the period immediately after
an election if the party if the new president did not have the complete control of the Congress.
A further examination of the post-election effects by disaggregating post-election periods
according to the come out of the elections provides evidence that is consistent with Hypothesis
Two. The statistically insignificant coefficients on the Democratic covariates indicate that there
is no difference in the likelihood of a peak or trough after the election of a Democratic president
as compared to other times. However, estimates on the coefficients on the Republican covariates
indicate that a market trough is significantly less likely to occur in the wake of a Republican
presidential victory than at other times. Meanwhile, political control is found to attenuate this
post-Republican effect. Alternatively, the implied post-Republican surge in the hazard for a bull
market is not found in our data even though control of the Congress seems to exacerbate the
hazard for a bull market after an election of a Republican.
1 Various studies have provided evidence of duration dependence in stock market. For example, Zhou and Ridgon
(2011) find evidence of negative duration dependence in all samples of bull markets and evidence of positive
duration dependence in complete, peacetime and post-World War I sample of bear markets. Using a duration-
dependent Markov-switching model, Maheu and McCurdy (2000) find declining hazard functions (negative duration
dependence) in both the bull and bear market states using monthly data from 1834-1995.
In the next section of the paper, we provide a brief review of related literature and offer our
hypotheses. The empirical approach used to test our hypotheses is discussed in Section 3, which
is followed by the empirical results presented in Section 4. Section 5 presents the results when
political control is considered. Finally, before concluding the study in Section 7, we conduct a
robust analysis of our results in Section 6.
2. Related Literature and Our Hypothesis
2.1 Related Literature
Based on their assumptions across different dimensions, theories of political cycles can be
classified into a two by two matrix. One of the dimensions concerns the nature of the economy
itself. For example, the traditional models in the early literature, such as that by Nordhaus
(1975), Lindbeck (1976) and Hibbs (1977), assume that the economy is characterized by a stable
inflation-output tradeoff, policy-makers have a direct control over inflation, and inflation
expectations are adaptive. More recent work expands the early studies by incorporating rational
expectations into their models. The principal assumptions of Persson and Tabellini (1990),
Rogoff and Sibert (1988), Rogoff (1990) and Alesina (1987), for example, are that economic
agents are forward-looking and make decisions based upon all information available to them at
the time. The connection between policy and outcome becomes more tenuous under these
rational expectations assumptions than under the traditional assumption of a stable Phillips
curve. In particular, there is a little scope for pre-election stimulation of the aggregate economy
and the post-election effects are more short-lived when people are rational and forward-looking
than when there is a stable Phillips curve in the traditional models.
The motivation of policy-makers represents another important dimension along which
models of political business cycles can be categorized. Opportunistic political business models,
such as Nordhaus (1975), assume that the goal of all policy-makers is to maximize the chance to
be reelected and policy is used towards this end. In the rational opportunistic political business
model of Persson and Tabellini (1990), the forwarding-looking behavior of economic agents
mitigates the extent to which the economy can be manipulated by policy and makes the voters’
goal to elect the most “competent” candidate regardless of ideology.
Instead of trying to be reelected, the goal pursued by policy-makers in partisan political
business cycle models is to pursue their political ideology. In the work of Hibbs (1977, 1987), in
which politicians can exploit a stable output-inflation tradeoff, this leads to differences across the
tenure of left-wing (Democratic) and right-wing (Republican) governments. The rational partisan
theory of Alesian (1987) preserves the assumption of policy-makers pursuing ideological
motives but tempers their ability to realize their goals by modeling an economy characterized by
rational wage-setters who are temporarily bound by nominal contracts. In the model, wages are
set equal to expected inflation. In the period before an election, the expected inflation is a
weighted average of the likelihood of the election of the party more sensitive to costs of inflation
(the Republicans) and the party less sensitive to inflation’s costs (the Democrats). The election
outcome determines the actual inflation rate and therefore whether real wages are unexpectedly
high (due to a Republican victory) and there is a contraction or whether real wage is
unexpectedly low (due to a Democratic victory) and there is an expansion. The length of the
deviation of output from its natural rate in the model is the length of the wage contract, not the
entire tenure of the administration as the Hibbs’ model suggests.
These theories present different implications regarding the temporal relationship between the
elections and the timing of turning points in business cycles. The opportunistic political business
cycle theory predicts an increased likelihood of a business trough (i.e. the end of a contraction)
with the coming of an election. It also predicts that the onset of contraction (i.e. a business cycle
peak) to offset the pre-election simulative policy is more likely following an election than at
other times. These predictions stand regardless of presidents’ partisanship. Alternatively, the
party in power is the key to the timing of stock market cycle turning points drawn from the
insights of the partisan theory. Rational partisan predicts that the likelihood of a business cycle
peak (trough) marking the end of a stock market expansion (contraction) is higher after the
election of a Republican president than at other times and there is no difference in the likelihoods
after the election of a Democratic president as compared to other times.
Klein (1996) first study the predictions from the political business cycle theories by using
the dates of turning points of the United States business cycles identified by the National Burau
of Economic Research. His results show a strong support for the post-election downturn
predicted by opportunistic political business cycle theories and there is evidence supporting an
increased likelihood of the end of a contraction in the two-year period before an election only
when the president is a Democrat. A closer examination of the post-election effect in his study
shows that consistent with rational partisan theory, there is a greater (smaller) likelihood of the
end of an expansion (a contraction) occurring in any given month in periods following
Republican presidential victories than at other times.
Alongside with the study on politic business cycles, there are increasing researches focusing
on the elections and the stock market. On one hand, more and more studies find that stock
market performance has an important role in predicting the outcome of presidential elections.
Chan and Jordan (2004) find that the equity market’s performance for ten months prior to an
election is a better predictor than GDP growth of the incumbents’ election results in recent years.
Prechter et al (2012) find that compared to GDP, inflation and unemployment rate, stock market
performance is a more powerful predictor for the incumbent’s reelection bids. D�̈�pke and
Pierdzioch (2006) find that stock market returns affect the popularity of the German government.
On the other hand, since new information can be reflected quickly in the stock prices, it is easier
for an incumbent to generate a stock market expansion than to change the course of the economy
before an election. Stock returns and return volatility also reflect investors’ perceptions about the
future course of the economy which depends on the ideological composition of the government if
partisan politics is in effect. In view of these arguments, whether electoral cycles or differences
in partisan politics exist in returns and/or volatility has been studied extensively. Allvine and
O’Neill (1980) conclude that stock prices rise relative to trend over the two years prior to a
presidential election. Lobo (1999) finds that stock returns are lower, and volatility is higher in
election years relative to nonelection years. Pantzalis et al. (2000) find positive abnormal returns
during the two-week period prior to the election week of thirty-three countries in their sample.
Herbst and Slinkman (1984), Huang (1985), Gartner and Wellershoff (1995, 1999) all find
evidence in support of a four-year presidential election cycle in the United States.
There are also differences in returns and volatility along the partisan lines. Riley and
Luksetich (1980) and Hobbs and Riley (1984) find that stock returns are higher under
Republican administrations in the United States, whereas Huang (1985) and Gartner and
Wellershoff (1995) are unable to find significant differences. Johnson et al. (1999) fail to find
any differences between Republican and Democratic presidential administrations for an index of
large-capitalization stocks, but they do find that small-capitalization stocks perform significantly
better during Democratic administrations, and the debt market performs significantly better
during Republican administrations. Lobo (1999) confirms the finding for the small-cap stocks
and also finds that the jump risk obtained from the volatility of the stock market is higher during
Democratic administrations. Santa-Clara and Valkanov (2003) discover that excess returns for
the value-weighted and equal-weighted portfolios are 9% and 16% higher, respectively, under
Democratic than Republican administrations. Vuchelen (2003) presents evidence that the
ideological composition of the government has a significant effect on the Belgian stock market.
Fuss and Bechtel (2008) show that small-firm stock returns were positively (negatively) linked to
the probability of a right- (left-) leaning coalition winning the election in Germany. Finally,
Siokis and Kapopoulos (2007) find that different political regimes affect the volatility of the
stock market index in Greece.
2.2 Our Hypotheses
The aforementioned studies point out that the occurrence and the outcome of presidential
elections affect the timing of turning points of the U.S. business cycles, and presidential elections
affect stock market performance. Existing studies, however, fail to investigate the important
potential link between the quadrennial election cycle and the timing of the turning points (peaks
and troughs) in the U.S. stock market. Based on the existing literature, we hypothesize the
following:
Hypothesis One: Compared to at other times, the likelihood of a market trough (the end of a
bear market) is higher in the period leading up to an election; and the likelihood of a market peak
(the end of a bull market) is higher in the period shortly after an election.
Hypothesis Two: The likelihood of the end of a bull market (a bear market) is higher (lower)
in the wake of a Republican election victory than at other times. There is, however, no difference
in the likelihood of a peak or trough after the election of a Democratic president as compared to
other times.
3. Empirical Approach
The timing of turning points of the stock market cycles relative to the quadrennial
presidential elections in the U.S. lends itself to an empirical investigation using duration analysis.
The data used in duration analysis consist of spells. In our data, a spell represents the number of
months in either a bear market or a bull market. The focus of duration analysis is the hazard
function. The hazard function at time 𝑡, ℎ(𝑡, 𝑥(𝑡)), is an estimate of the probability of the
completion of a spell during the time interval (𝑡, 𝑡 + 𝑑𝑡), given that the spell has lasted up until
time 𝑡. We estimate the hazard function for the probability of a peak (trough) in the U.S. stock
market during the next month given that the market has been in a bull (bear) market up until the
beginning of that month. The hazard function allows for duration dependence if its value at any
moment is a function of the time already spent in a spell. The hazard function may shift due to
exogenous factors, represented by the vector 𝑥(𝑡), which are called covariates. In a continuous
time framework, the hazard function is defined as
ℎ(𝑡, 𝑥(𝑡)) = lim𝑑𝑡→0
Pr (𝑡 ≤ 𝑇 < 𝑡 + 𝑑𝑡|𝑇 ≥ 𝑡, 𝑥(𝑡))
𝑑𝑡
The hazard function can be understood as the probability of a turning point in the short
interval 𝑑𝑡 after 𝑡, conditional on the current phase of the business cycle having lasted until
time 𝑡.
There are various potential candidates for the functional form used to implement this
analysis. The focus of attention in this study is the effect of the election covariates on the hazard
rather than the estimation of the duration dependence of expansions or contractions. Therefore,
we estimate the Cox proportional hazard model. Cox model factors the hazard into an arbitrary
and unspecified baseline hazard, ℎ0(𝑡) , and a function that depends upon a vector of explanatory
variables, 𝑥(𝑡), and the associated vector of coefficients, 𝑆, as follows;
ℎ(𝑡, 𝑥(𝑡), 𝛽, ℎ0) = ℎ0(𝑡) exp(𝑥(𝑡)𝛽)
This specification satisfies the requirement of non-negativity of the hazard without
imposing any restrictions on the coefficients 𝛽. The exponent of the coefficient on pre-election
or postelection covariates can be interpreted as the shift of the hazard during the relevant period
as compared to the other times. 2
The focus of this study is a set of covariates representing specified periods before or after
elections. These pre-election and post-election periods are identified by dummy variables that
enter as time-varying covariates, that is, covariates that can change over the course of a spell.3
Each of these covariates represents one of three different time frames and correspondingly is set
equal to one in the eight-month, sixteen-month, or twenty-four-month period either before or
after an election. 4 The use of different time frames for the covariates allows for investigation of
the length of the period of the political effect on the hazard. The coefficients across different
specifications of time frames are directly comparable since the hazard is the estimate of the
likelihood of the completion of a market cycle phase in the next month conditional on its lasting
up until that month.
To shed light on how the hazard rates may depend on the underlying state of the
economy, some specifications also include interest rates as a time-varying covariate.5 Interest
rate levels may be affected by a low-frequency component and therefore might not contain the
same information over a sample as long as ours, whereas interest rates changes are more likely to
track business cycle variation across the full sample. For this reason, we include both levels of
and changes in interest rates. Our analysis of nominal stock prices uses nominal interest rates,
2 The arbitrary baseline hazard of the Cox proportional hazard model can have any shape and it is not estimated. The
proportional hazard specification is well suited for investigating the effect of covariates on the relative risk of ending
a spell, but it does not lead itself to an investigation of duration dependence. 3 It is important to use time-varying covariates rather than simply identify those business cycles in which there was
an election with a dummy variable that serves as a constant covariate because the longer the business cycle the more
likely that there would be an election during it. Therefore, the use of elections as a constant covariate would give
rise to spurious results. 4 In an alternative specification of the postelection covariates begin in the month following the new president’s
inauguration if he represented a different political party from his predecessor and began in the month after the
election otherwise. This alternative is specification is more consistent with the notion that once in power, the
government was able to affect the economy, while the specification based only on election dates is more consistent
with the notion the effect of the election on the stock market was due to the “news” revealed by the outcome of the
election. In any case, the results using either specification were very similar. 5 Interest rates have been widely document to closely track the state of the business cycle and appear to be a key
determinant of stock returns at the monthly horizon (see, e.g., Kandel and STambaugh 1990; Fama and French
1988)
whereas our analysis of real stock prices is based on real interest rates. Both interest rates are
collected from Sheller’s database.6
The coefficients on the pre-election and post-election covariates represent the shift in a
hazard during the specified period before or after an election, respectively, holding constant the
effect of duration dependence and controlling for the underlying economy if interest rates are
included. Pre-election and post-election covariates used in tests of Hypothesis One do not
distinguish between political parties. Hypothesis One predicts positive coefficients on the pre-
election covariates in hazard estimates for bear markets. This implies that bear markets that have
lasted until the period before an election are more likely to end at that time than at other times.
The prediction of a post-election downturn from the hypothesis is consistent with positive
coefficients on post-election covariates in hazard estimates for bull markets.
The set of post-election covariates used to test Hypothesis Two includes separate
covariates for the period after the election of a Republican president and for the period after the
election of a Democratic president. Hypothesis Two suggests that the post-election effect
depends upon the party won the election. It predicts positive coefficients on the post-election
covariates representing the period following the election of a Republican in hazard estimates of
bull markets and negative coefficients on the post-election covariates in hazard estimates of bear
markets. In addition, the coefficients on the covariates that represent the period following the
election a Democratic president are expected to be indifferent from zero in hazard estimates for
both bull and bear markets.
It is reasonable to expect that the effects of presidential elections on the turning points of the
U.S. stock market may differ across sub-periods of the almost century and a half over which we
have data. Accordingly, we estimate hazard functions for the full set of 40 bears and 39 bulls in
the United States since 1872 as well as for subsamples of the 25 bears and 24 bulls in the period
after World War I and the 19 bears and 18 bulls in the period after World War II.
6 http://www.econ.yale.edu/~shiller/data.htm.
4. Empirical Results
The key data in our analysis of the link between political and financial market events are
the dates of turning points of the U.S. stock market cycle. As the name points out that a turning
point is when the stock market trend turns, i.e. goes from being generalized upward moving to
generalized downward moving or vice versa. The upward trend is commonly called a bull market
and the downward period is known as a bear market. Although bull and bear markets are
common words to investors, there is no academic general definition. For this study, we identify
the market regimes by following the algorithm developed by Pagan and Sossounov (2003). The
turning points are peaks and troughs of the identified stock market cycles.7 Since inflation has
varied considerably over our sample period, and it can be argued that drift in nominal prices does
not have the same interpretation during periods of low and high inflation. To deal with effects
arising from this, in addition to the nominal stock prices, we consider the real prices as a proxy
for the general market for robustness check. Both price series are also collected from Sheller’s
database.
Our sample begins with the peak in May 1872 and ends with the trough in February
2016. There are 39 bull markets (periods from troughs to peaks) and 40 bear markets (periods
from peaks to troughs) covered in our entire sample. The duration of each of these bears and
bulls is presented in Table l. The final two columns in the table report whether an election was
held during that phase of the market cycle. It can be seen from Table 1 that 24 of the 36 elections
held over the entire sample occurred during bull markets. This is broadly consistent with the
prediction of Hypothesis One that incumbents attempt to generate a rising market in the period
leading up to elections to increase the likelihood of being returned to office.
Table 2 provides an initial view of the timing of peaks and troughs relative to presidential
elections. We calculate the number of months since the last election for each of the turning
points by subsamples as well as by the party of the president. Except for the Democratic
administrations in the post-World War II period, the average number of months between the last
election and a market peak (the end of a bull market) is consistently smaller than the average
number of months between the last election and a market trough (the of a bear market) across
7 We account a peak and a trough as the final month of the bull market or bear market, respectively
samples and across political parties. This is consistent with the prediction that bull markets tend
to end soon after a presidential election and bear markets tend to end before presidential
elections. Hypothesis Two predicts differences across Republican and Democratic
administrations. Comparing the second and third panels of Table 1 shows that, for the full
sample and the subsample covers the post-World War I era, troughs occur later, on average, after
the election of a Republican than after the election of a Democrat. Additionally, on average,
peaks occur sooner in Republican administrations than in Democratic administrations for the
subsamples beginning after the World War I and the World War II respectively. The standard
deviations for all these statistics, however, are quite large relative to the averages.
Further information about the distribution of the number of months between market cycle
turning points and elections is provided in the histograms in Figures 1 and 2. Partially confirming
the results in Table 1, the histograms of troughs show that there are more troughs occurred in the
pre-election period than in the post-election period. Interestingly, however, the histograms of
peaks reveal that the mass points occur in the pre-election period rather than in the post-election
period. This pattern seems to be more pronounced with samples that cover only the later periods.
Nevertheless, one of the spikes in the histograms of peaks is within the 8-month period after an
election, which indicates that the period immediately after an election is still one of the periods
when peaks are most likely to occur. Histograms the differentiates across presidential parties are
presented in Figure 2. These histograms show that, compared to Democratic presidents, troughs
are more likely to occur in the pre-election period for Republican presidents. Furthermore,
despite there are more peaks in the pre-election period for Republican presidents, there is a clear
pattern that peaks are more likely to occur in the period immediately after an election of a
Republican than after the election of a Democrat.
While the above summary statistics are suggestive, more powerful tests of our hypotheses
are provided by duration analysis. Unlike the unconditional estimates in Table 2, duration
analysis allows us to test the effects of elections on stock market cycles holding constant the
effects of the underlying economy and the time on the market cycle itself. The estimates of the
Cox proportional hazard models for bear markets with the various pre-election periods serving as
time-varying covariates are presented in Table 3. Although all the coefficients on the pre-election
dummy variables are of the expected positive sign, the only instance of a coefficient that
significantly differs from zero is the 16-month covariate for the post-World War II subsample.
The robust p-value on this coefficient rises to above 0.10, however, when interest rates are
included.8 Therefore, the results provide weak evidence, at best, for the prediction of Hypothesis
One that there is an increased likelihood of a bear market ending in the period before an election
than in other periods (conditional upon its having lasted until that period). Kelvin (1996) also
finds that there is no evidence to suggest that the likelihood of an end of contraction in the
business cycle in the period before an election is significantly higher than in other periods.
Hypothesis One does not suggest that there should be a difference across political parties
in attempts to engineer a market expansion in the period before an election. It may be, however,
that the likelihood of bear market ending in the period before an election depends upon the
administration holding power at the time. This possibility is investigated in Table 4. In that table,
the covariates representing the period before an election distinguish between those times when a
Republican holds the presidency and those times when a Democrat sits in the White House. As
above, positive and significant values of the estimated coefficients would demonstrate a greater
likelihood of the end of a bear market in the period leading up to an election than at other times.
As seen in Table 4, there is no evidence that a bear market is more likely to end in the
period before an election than at other times when the incumbent is a Democrat. Interestingly,
however, the coefficients on the covariates representing the pre-election period with a
Republican incumbent are not only all of a positive sign but also those for the 16-month and 24-
month covariates are significant at the 10 percent and 5 percent levels of significance for the
subsamples. The estimates suggest that, compared to other times, an ongoing bear market is
2.20 (𝑒𝑥𝑝 (0.79)) [2.41 (exp (0.88))] times more likely to end in any given month during the 24-
month period [the 16-month period] before an election when the incumbent is a Republican for
the subsample beginning after World War I. This likelihood of the end of a market contraction in
the next month by virtue of a proximate election and a Republican president rises to
2.92(𝑒𝑥𝑝 (1.07)) [2.83(𝑒𝑥𝑝 (1.04)) ] times when the sample is constrained to the post-World
War II era. A similar finding has also been reported by Wong and McAleer (2009) that,
8 The negative coefficients on the interest rate changes indicate that the probability of a bear market to end decreases
when the interest rate changes increase. And the coefficients are significant at the 5 percent level and the 1 percent
level in the full sample and the post-World War I subsample respectively.
compared to their Democrat counterpart, Republican party tends to have greater cause to engage
in active manipulation in the equity market to win.
In addition to the prediction of a greater likelihood of a bear market ending in the period
before an election, Hypothesis One also predicts a higher likelihood of a bull market ending in
the period immediately after an election than in other periods conditional on its having lasted
until that period. This would be reflected in positive coefficients on the post-election covariates.
Table 5 presents the estimates for bull markets with the post-election periods serving as time-
varying covariates.9 Despite results demonstrate some effect consistent with the prediction, such
as the estimated coefficients on the post-election time period dummy variables representing the
8-month and 16-month periods following an election are of the expected positive sign for the
post-World War I and post-World War II subsamples and the point estimates are largest on the 8-
month covariate, none of the coefficients are statistically significant. Thus, these results provide
no strong evidence for the prediction that there is a greater likelihood of a market expansion
ending in the period immediately after an election than in other periods (conditional upon its
having lasted until period).
A change in the likelihood of a stock market cycle turning point in the period following
an election may depend on the outcome of an election. As discussed above, to test Hypothesis
Two, we use a specification that has separate time-varying covariates for the months since the
election of a Republican and the months since the election of a Democrat. Hypothesis Two
suggests that, compared to other times, there is an increased likelihood of the end of a bull
market and a decreased likelihood of the end of a bear market after the election of a Republican.
There is, however, no difference in the likelihoods after the election of a Democrat than at other
times. In other words, this implies positive coefficients on the covariates representing the period
after an election was won by a Republican and those on the covariates representing the periods
after an election that went to the Democratic candidate are insignificantly different from zero in
hazard estimates for bull markets. Alternatively, in hazard estimates of bear markets, it implies
negative coefficients on the covariates representing the period in the wake of a Republican
9 The negative coefficients on the interest rate changes seem to suggest that increase in interest rate changes not only
reduces the probability of a market trough but also the probability of a market peak.
presidential victory and the insignificant coefficients on the covariates representing the period
after a Democratic presidential victory.
Results in Tables 6 and 7 strongly support the predictions of our Hypothesis Two as
regards the effects of Democratic presidential election victories as there are no instances of a
statistically significant coefficient on the Democratic covariate. In the estimates of the hazards
for bull markets presented in Table 6, the coefficients on the covariates representing the 8-month
and 16-month periods after the election of a Republican are positive across specifications (i.e.
with and without interest rates) for both subsamples. Specifically, the coefficients on the
covariates representing the 8-month period after a Republican election victory are significant at
the 10 percent level for the post-World War II sample.10
Compared to the post-election covariates that do not differentiate between the party of the
president, the coefficients on the Republican covariates are significantly larger in size but for the
covariates representing the two-year period after an election of a Republican. As with the results
in Table 5, the coefficients are largest on covariates representing on the 8-month period
following the election of a Republican president, implying that the post-Republican effect is
largest in the period immediately after an election was won by a Republican candidate. The point
estimate suggests that a bull market is twice (𝑒𝑥𝑝 (0.66)) as likely to end within 8 months of a
Republican presidential victory as at other times in the post-World War I period, controlling for
interest rates and for duration dependence. This likelihood of the end of a bull market in the next
month rises to 2.44 (𝑒𝑥𝑝 (0.88)) when the sample is constrained to the post-World War II era.
This suggests that the post-Republican effect has become more noticeable over time.
Results presented in Table 7 for the estimates for hazard functions for bear markets
display effects consistent with our Hypothesis Two.11 The coefficients on the covariates for the
period following the election of a Republican are all of the expected negative sign and those for
the 24-month covariate are significant at the 10 percent level or higher. The coefficients on the
10 There are only two instances of a peak within 8 months of a Democratic presidential victory in the post-World
War II ear implying that the result for the post-World War II period in Table 5 is largely due to periods following a
Republican presidential victory.
11 There are insufficient observations for estimating the coefficient on covariates representing the 8-month period
after the election of a Republican president.
covariate representing the 16-month period after the election of a Republican are also found to be
significant in the full sample. The point estimates suggest that there are substantial differences in
the hazard of a market contraction between the period after an election of a Republican president
and other times. For example, a market trough is about 28 (𝑒𝑥𝑝(−1.28)) percent as likely to
occur in the 24-month period following a Republican presidential victory as at other times for the
post-World War II sample period after controlling for interest rates and for duration dependence.
5. Political Control and Duration of Stock Market Cycles
Political alignment between the White House and the Congress, another important political
factor in the U.S. stock market, is largely overlooked in the existing research when studying the
presidential election cycles in the U.S. stock market. In this section, we investigate whether
political control of the Congress attenuates or exacerbates the election effects on the likelihood
of a market peak or trough presented early in the paper. To investigate the role of political
control, we construct a dummy variable, Congress, which equals to one if the president’s party
has control (holds the majority of seats) of both the house of representatives and the house of
senate simultaneously (i.e. complete control of the legislative branch in the nation) and interact it
with the covariates representing the pre- and post-election periods. We re-estimate the
regressions presented in the early tables by controlling for political alignment between the White
House and the Congress. The new results are presented in Tables 8-12.
As seen in Table 8, after controlling for political control of the Congress, there is a
significant evidence for the prediction that there is a greater probability of an ongoing bear
market ending in the period prior to an election than at other times. Estimates on the coefficients
on the pre-election time period dummies now are not only all of the expected positive sign but
also are highly significant for covariates representing the 16-month and the 24-month periods
leading up to an election. These new estimates indicate that the pre-election effect on the stock
market cycle is largest in the two-year period before an election, controlling for interest rates and
political control for the Congress. Specifically, the estimates suggest that an enduring bear
market is 3.13 (𝑒𝑥𝑝 (1.14)) times more likely to end in any given month during the two-year
period before an election than at other time if the incumbent’s party did not control the Congress
for the entire sample. This likelihood of the end of a bear market in the next month by the virtue
of a proximate election rise to 5.87 (𝑒𝑥𝑝 (1.77)) and 6.43 (𝑒𝑥𝑝 (1.86)) when the sample is
constrained to the post-World War I era and the post-World War II era respectively.
The negative and significant coefficients for the interaction term, Election*Congress,
however, suggest that political control of the Congress weakens this pre-election surge in the
hazard for a bear market.1213 The magnitude of the coefficients shows a remarkable difference in
the pre-election hazard for a bear market between an incumbent with political control and one
without such control. For instance, going from no political control to complete political control,
there is a 97 percent reduction in the pre-election surge in the hazard for a bear market in the 24-
month period prior to an election for the post-World War II period. Moreover, the positive and
significant coefficients on the level effect of Congress imply that bear markets are more likely to
end when the White House and the Congress are controlled by the same party than at other times.
Estimates presented in Table 9 are largely parallel to the ones presented in Table 8.14
After the inclusion of political control, the estimates on the pre-election covariates become more
significant both economically and statistically. Estimates on the period dummies are all of a
positive sign with the coefficients on covariates representing a Democratic incumbent are most
significant in the full sample period and the coefficients on covariates representing a Republican
incumbent are most significant in the subsamples. These results suggest that, for both Republican
and Democratic presidents, the probability of ending an ongoing bear market before an election
is greater than at other times when the incumbents’ parties do not have the control of the
Congress.
Like the results in Table 8, the positive and significant coefficients on the political
control dummy imply that party dominance in the Congress shortens the duration of bear
12 There are insufficient observations for estimating the coefficients on interaction term, Election*Congress, for the
8-month time frame for the post-World War II period. 13 One possible interpretation for the cause of the reduction in the hazard for a bear market before an election is that
the range of policy tools available to the government to win support from voters may increase when the incumbent
has the complete control of the Congress. The president could choose policies that benefit voters more directly, such
as by increasing transfer payments, and lowering taxes prior to an election rather than trying to manipulate the stock
market. 14 There are insufficient observations for estimating the coefficients on the interaction term, Republican * Congress
for the post-World War II sample. And there are insufficient observations for estimating the coefficients on both
interaction terms when the sample is constrained to the 8-month time frame.
markets. Nevertheless, the negative coefficients on the interaction terms, Republican * Congress
and Democrat * Congress, indicate that political control of the Congress dilutes the pre-election
surge in the hazard for a bear market. Particularly, this dilution is more apparent for Democratic
incumbents. For instance, the estimates in the 16-month time frame suggest that going from no
control of the Congress to complete control of the Congress, the likelihood ending a bear market
reduces by 95 percent when the incumbent is a Democrat or cuts by 28 percent when the
incumbent is a Republican for the full sample.
In contrast to the results in Table 6, the negative coefficients on the covariates for the
post-election dummies in Table 10 suggest market expansions are actually less likely to end in
the period following an election when the newly elected president does not have the full support
from the Congress. Within any one sample period, the coefficient on the 24-month covariate is
larger than the coefficient on either the 8-month or the 16-month covariate and it is significant at
the 5 percent level for the full sample. The point estimate indicates the difference in the hazard
for a bull market between the 24-month period after an election and other periods is largest. In
the full sample period, given its survival up to that time and the absence of political control, the
likelihood of the end of a market expansion is less than 20 (𝑒𝑥𝑝 (−1.63)) percent as likely to
occur within two years following an election as at other time. The larger point estimates for the
post-World War I and the post-World War II subsamples are even more striking; the likelihood
of a market peak to occur within two years after an election is just 15 (𝑒𝑥𝑝 (−1.93)) percent as
likely as in other periods in the post-World War I period and is 13 (𝑒𝑥𝑝 (−2.02)) percent as
likely as in other periods in the post-World War II period. This suggests that the post-election
reduction in the hazard for a bull market has become more pronounced over time.
The significant and negative coefficients on the political control itself indicating that
political control of the Congress enables politicians to substitute a longer bull market when a
president has the political control for the Congress for a shorter one when the political control is
absent. Across samples and specifications, the estimated coefficients for the interaction term,
Election * Congress, remain positive and their economic and statistical significance increase as
the post-election time frame expands from 8 months to 16 months and to 24 months after an
election and as the sample is restricted from the full sample to the post-World War I subsample
to the post-World War II subsample. To give an indication of the economic significance of the
political control effect in the post-election period, the coefficient for the interaction term for the
24-month time frame for the post-World War II period is 3.91, which implies that a market peak
marking the end of a market expansion is almost 50 (𝑒𝑥𝑝(3.91)) time as likely to occur within
two years after an election if the president elected has the control of the Congress as the one
lacks of such control. Furthermore, the significant and negative coefficient on the interaction
term, Election * Congress, implies that the early finding that a bull market is more likely to end
in the post-election period can be largely explained by the period when there is no political
control for the new administration.
Estimates in Table 11 largely resemble the results reported in Table 6.15 Although the
coefficients on covariate representing the 8-month period following an election of a Republican
remain positive after controlling for political control of the Congress, there is no coefficient that
is significantly different from zero. A closer examining on the interaction term shows that,
except for the 8-month time frame for the full sample, the coefficient on the interaction term,
Republican * Congress, are positive across specifications and across sample periods. The
magnitude of estimates suggests that there are striking differences in the hazard between the
post-election period with a Republican political control and the corresponding period without
such political control. For example, the likelihood of the end of a continuing bull market is 7.54
(𝑒𝑥𝑝 (2.02)) times as high in the 16-month post-election period when a Republican president
has the political control as in the same post-election period when the president does not have
such control. Furthermore, consistent with the results in Table 10, the negative coefficient on
Congress seems to suggest that politicians may prolong the duration of a bull market when they
have control over the nation’s executive and legislative branches.
Finally, after the inclusion of political control, results presented in Table 12 reveal an
even superior evidence that market troughs are less likely to happen after the election of a
Republican president.16 Compared to the results in Table 7, the new results display a substantial
improvement in both economic and statistical significance for the estimates on the coefficients
on the post-election covariates representing a Republican election victory. The positive and
15 There are insufficient observations for estimating the coefficient on the interaction item, Democratic*Congress. 16 There are insufficient observations for estimating the coefficients on the Republican dummy and its interaction
term, Republican*Congress, for the 8-month time frame, and there are not enough observations for estimating the
coefficient on the interaction item, Democrat*Congress.
significant estimates on the interaction term, Republican*Congress, imply that political control
of the Congress diminishes the post-election reduction in the hazard for bear markets following
an election of a Republican. The point estimates show that the differences in the hazard between
the post-election periods with Republican-controlled Whitehouse and Congress and the periods
without such control are striking. For example, the point estimate for the 24-month Republican
covariate for the post-World War II sample, controlling for interest rates and for time
dependence, suggests that a market trough, marking the end of a bear market, is about 29
(𝑒𝑥𝑝 (3.37)) times as likely to occur within two years following the election of a Republican
president when Republican party holds the majority seats of the Congress as the corresponding
period when there is no political control for the president.
6. Robustness Analysis
Given our data sample spans nearly century and a half, it is possible that drift in nominal
prices does not have the same interpretation during periods of low and high inflation. To deal
with effects arising from this, we also consider real stock prices as a proxy for the general market
for our robustness check.
Like Table 1, in Table 13 we report the dates of the turning points of the market cycle
identified by using real stock prices and the duration of each of the new bears and bulls and
whether an election was held during that phase of the cycle. Again, we can see that 24 of the 36
elections held over the complete sample occurred during bull markets. This is in line with the
prediction that incumbents are likely to cause a market expansion prior to an election in order to
boost his chance to be re-elected. Similarly, Table 14 provides some simple statistics on the
timing of new peaks and troughs relative to presidential elections. Confirming with the results in
Table 2, it can be seen from Table 14, apart from the Democratic administrations in the post-
World War II period, the average number of months between the last election and a market peak
is consistently lesser than the average number of months between the last election and a market
trough across samples and across parties. This is consistent with the prediction that bull markets
tend to end soon after an election while bear markets end to end before an election. Additionally,
the second and third panels of the table display a clear pattern that, across samples, troughs
occur, on average, later after the election of a Republican than after the election of a Democrat.
And, on average, peaks occur sooner in Republican administrations than in Democratic
administrations. The standard deviations for all these statistics, however, are quite large relative
to the averages. The histograms in Figures 3 and 4 show us alike patterns in the distribution of
the number of months between market cycle turning points and election as in Figures 1 and 2.
The histograms of troughs show that, except for the Democratic presidents for the subsamples,
there are more troughs in the pre-election period than in the post-election period. And the
histograms of peaks show that there are more peaks in the pre-election period instead of in the
post-election period. Compared to Democratic presidents, peaks are more likely to occur in the
period immediately after an election of a Republican president.
Finally, the new estimates from the Cox regressions strengthen our early findings.17
Political alignment between the Whitehouse and Congress remains to be critical in the timing of
turning points relative to the elections. The predicted increase in the hazard for a bear market
prior to an election is found to be significant when the incumbent’s party does not have the
control of the Congress. We do observe, however, a significant increase in the likelihood of the
end of a bear market when the incumbent is a Republican even without control for political
alignment between the Whitehouse and the Congress. Although the new estimates on the pre-
election dummies in the hazard functions for bull markets are all of the expected positive sign,
none of the estimates significantly differs from zero. In fact, after controlling for political
control, the estimates still indicate a smaller hazard for a bull in the period immediately after an
election. A further examination of the post-election effects by disaggregating post-election
periods according to the come out of the elections provides evidence that is consistent with
Hypothesis Two. There are no instances of a statistically significant coefficient on the
Democratic covariates. Alternatively, estimates on the coefficients on the Republican covariates
indicate that a market contraction is less likely to end in the wake of a Republican presidential
victory than at other times and political control lessens this post-Republican effect. Lastly,
compared to the initial results, the positive estimates on the coefficients on covariates
representing the period following an election of a Republican are consistent with the implied
post-Republican surge in the hazard for a bull market and the coefficients on the covariates
representing the 8-month period after a Republican election victory are statistically different
17 To save the space, we did not report the new estimates in the paper. The new estimates are available upon request.
from zero. Political control, however, still is found to exacerbate the post-Republican effect in
hazard for a bull market by using the new data.
7. Conclusion
This paper provides a more direct test of the timing of stock market cycle turning points
relative to the elections than previous research which considers the amplitude of returns and
volatilities before and after presidential elections. Results from the duration analysis show that
political alignment between the White House and the Congress (political control) holds a crucial
role in the timing of peaks and troughs of the U.S. stock market cycles relative to the elections.
There is significant evidence supporting the prediction of our Hypothesis One that there is an
increased likelihood of the end of a bear market in the two-year period before an election when
the incumbents have political control of the Congress or when the incumbent is a Republican.
The prediction that there is a post-election surge in the hazard for a bull market in the period
immediately after an election, however, cannot be found in our data. In fact, our estimates
indicate a smaller probability for a market peak to occur in the post-election period than at other
times after controlling for political control. Further examination of the post-election results
provides evidence that is consistent with our Hypothesis Two that there is no difference in the
likelihood of a peak or trough after the election of a Democratic president as compared to other
times. Bear markets are less likely to end after an election of a Republican than at other times
and political control seems to diminish this post-Republican effect. Lastly, although the implied
post-Republican surge in the hazard for a bull market is not found in our data, political control
seems to intensify the hazard for a bull market after an election of a Republican.
The results presented in this paper do not provide a complete test of the election-cycle effects
in the stock market as there are other implications that cannot be addressed by using duration
analysis. The findings presented in this paper, however, undoubtedly complement the existing
literature and extend our knowledge of how presidential elections can affect the U.S. financial
markets. A better understanding of the relationship between the timing of market turning points
and political events is not only central to our understanding of the empirical relevance of
political business cycle theories in financial markets but also it could help investors to make
better investment decisions when combined with other information.
TABLE 1
DATES OF US STOCK MARKET TURNING POINTS
Trough Peak Bear Bull Bear Bull
5/1872
11/1873 4/1875 18 17 YES NO
6/1877 6/1881 26 48 YES YES
1/1885 5/1887 43 28 YES NO
6/1888 5/1890 13 23 NO YES
12/1890 8/1892 7 20 NO NO
8/1893 4/1894 8 8 YES NO
3/1895 9/1895 11 6 NO NO
8/1896 9/1897 11 13 NO YES
4/1898 4/1899 7 12 NO NO
9/1900 9/1902 17 24 NO YES
10/1903 9/1906 13 35 NO YES
11/1907 12/1909 14 25 NO YES
7/1910 9/1912 7 26 NO NO
12/1914 11/1916 24 23 YES YES
12/1917 7/1919 12 19 NO NO
8/1921 3/1923 25 19 YES NO
10/1923 9/1929 7 71 NO
6/1932 2/1934 33 20 NO YES
3/1935 2/1937 13 23 NO YES
4/1938 11/1938 14 7 NO NO
4/1942 5/1946 41 49 YES YES
2/1948 6/1948 21 4 NO NO
6/1949 1/1953 12 43 YES YES
9/1953 7/1956 8 34 NO NO
12/1957 7/1959 12 19 YES NO
10/1960 12/1961 15 14 NO YES
6/1962 1/1966 6 43 NO YES
10/1966 12/1968 11 26 NO YES
6/1970 4/1971 18 10 NO NO
11/1971 1/1973 7 14 NO YES
12/1974 9/1976 23 21 NO NO
3/1978 9/1978 15 6 YES NO
4/1980 11/1980 19 7 NO YES
7/1982 10/1983 19 15 NO NO
7/1984 8/1987 9 37 NO YES
12/1987 8/2000 4 152 NO
2/2003 10/2007 26 56 YES YES
3/2009 5/2011 17 26 YES NO
9/2011 5/2015 4 44 NO YES
2/2016 9 NO
Notes: Along a row, Bear market refers to period from peak in pervious row to trough in the next row.
Bears market refers to period from trough in that raw to prak in that row.
Date of
YES(2 ELECTIONS)
YES(3 ELECTIONS)
Duration (in months) Was an Election Held during the
TABLE 2
STATISTICS ON TIMING OF PEAKS AND TROUGHS RELATIVE TO ELECTIONS
Months Since Elections
Peaks Troughs
Av'g (s.d.)
Min. Max. Obs. Av'g (s.d.)
Min. Max. Obs.
Full Sample 23.28(14.21) 0 46 39 25.30(13.28) 2 47 40
Post WW1 23.08(15.06) 0 46 24 25.96(12.84) 4 47 25
Post WW2 25.33(16.15) 0 46 18 26.32(13.18) 4 47 19
Republican Administrations
Months Since Elections
Peaks Troughs
Av'g (s.d.)
Min. Max. Obs. Av'g (s.d.)
Min. Max. Obs.
Full Sample 23.39(15.37) 0 46 23 25.80(12.51) 7 46 20
Post WW1 22.85(17.23) 0 46 13 26.50(12.48) 9 44 12
Post WW2 23.55(18.35) 0 46 11 25.67(11.49) 10 44 9
Democrat
Administrations
Months Since Elections
Peaks Troughs
Av'g (s.d.)
Min. Max. Obs. Av'g (s.d.)
Min. Max. Obs.
Full Sample 23.13(12.83) 0 45 16 24.80(14.32) 2 47 20
Post WW1 23.36(12.86) 3 45 11 25.46(13.64) 4 47 13
Post WW2 28.14(12.77) 13 45 7 26.90(15.14) 4 47 10
TABLE 3
ESTIMATES OF HAZARD FUNCTION FOR BEAR MARKETS
1. Full Sample
Equation Variable 24 months 16 months 8 months
A. Election 0.298 0.189 0.189
(0.32) (0.27) (0.40)
B. Election 0.295 0.189 0.0916
(0.33) (0.28) (0.41)
Interest rates 0.035 0.010 0.072
(0.07) (0.07) (0.06)
Interest Rate Changes -0.557 -0.079 -1.068 **
(0.84) (0.66) (0.44)
2. Post-World War I
Equation Variable 24 months 16 months 8 months
A. Election 0.407 0.395 0.183
(0.37) (0.28) (0.49)
B. Election 0.417 0.411 0.009
(0.39) (0.30) (0.46)
Interest rates 0.065 0.048 0.098
(0.07) (0.07) (0.07)
Interest Rate Changes -0.675 -0.234 -1.147 ***
(0.69) (0.58) (0.41)
3. Post-World War II
Equation Variable 24 months 16 months 8 months
A. Election 0.547 0.706 * 0.726
(0.47) (0.40) (0.59)
B. Election 0.483 0.684 0.544
(0.50) (0.47) (0.66)
Interest rates -0.019 -0.041 -0.006
(0.07) (0.08) (0.06)
Interest Rate Changes -0.489 0.220 -0.654
(0.66) (0.77) (0.44)
* Significant at 90 percent level of confidence
** Significant at 95 percent level of confidence
*** Significant at 99 percent level of confidence
Dummy Variables: Months Before an Election
Dummy Variables: Months Before an Election
Dummy Variables: Months Before an Election
TABLE 4
ESTIMATES OF HAZARD FUNCTION FOR BEAR MARKETS BY PARTY
1. Full Sample
Equation Variable 24 months 16 months 8 months
A. Republican 0.313 0.280 0.066
(0.37) (0.37) (0.47)
Democrat 0.276 0.042 0.364
(0.42) (0.43) (0.60)
B. Republican 0.302 0.028 -0.030
(0.38) (0.37) (0.48)
Democrat 0.285 0.044 0.284
(0.43) (0.43) (0.55)
Interest Rates 0.035 0.007 0.073
(0.07) (0.07) (0.06)
Interest Rate Changes -0.554 -0.060 -1.044 **
(0.85) (0.63) (0.43)
2. Post-World War I
Equation Variable 24 months 16 months 8 months
A. Republican 0.799 * 0.893 ** 0.653
(0.47) (0.39) (0.49)
Democrat -0.113 -0.233 -0.523
(0.53) (0.58) (1.12)
B. Republican 0.786 * 0.883 ** 0.434
(0.49) (0.39) (0.50)
Democrat -0.089 -0.209 -0.846
(0.53) (0.58) (0.75)
Interest Rates 0.045 0.024 0.096 *
(0.06) (0.05) (0.06)
Interest Rate Changes -0.544 -0.157 -1.292 ***
(0.64) (0.49) (0.45)
3.Post-World War II
Equation Variable 24 months 16 months 8 months
A. Republican 1.117 ** 1.033 * 0.458
(0.56) (0.55) (0.62)
Democrat 0.024 0.342 1.644 *
(0.59) (0.64) (0.92)
B. Republican 1.071 * 1.036 * 0.402
(0.58) (0.58) (0.69)
Democrat -0.098 0.254 1.344
(0.62) (0.70) (1.16)
Interest Rates -0.042 -0.056 -0.019
(0.05) (0.06) (0.07)
Interest Rate Changes -0.423 0.250 -0.444
(0.63) (0.72) (0.55)
Notes: Numbers in paranthesis are robust standard errors.
* Significant at 90 percent level of confidence
** Significant at 95 percent level of confidence
*** Significant at 99 percent level of confidence
Dummy Variables: Months Before an Election, By Party
Dummy Variables: Months Before an Election, By Party
Dummy Variables: Months Before an Election, By Party
TABLE 5
ESTIMATES OF HAZARD FUNCTION FOR BULL MARKETS
1. Full Sample
Equation Variable 8 months 16 months 24 months
A. Election -0.142 -0.173 -0.096
(0.47) (0.34) (0.34)
B. Election -0.100 -0.130 -0.091
(0.48) (0.35) (0.34)
Interest rates 0.020 0.003 -0.018
(0.07) (0.09) (0.10)
Interest Rate Changes -0.829 -0.748 -0.538
(0.56) (0.68) (1.16)
2. Post-World War I
Equation Variable 8 months 16 months 24 months
A. Election 0.311 0.100 -0.010
(0.55) (0.42) (0.48)
B. Election 0.386 0.174 -0.001
(0.56) (0.44) (0.49)
Interest rates 0.041 0.010 0.005
(0.07) (0.08) (0.10)
Interest Rate Changes -0.927 -1.040 -0.451
(0.63) (0.71) (1.17)
3. Post-World War II
Equation Variable 8 months 16 months 24 months
A. Election 0.323 0.137 -0.395
(0.56) (0.46) (0.51)
B. Election 0.364 0.300 -0.355
(0.56) (0.49) (0.54)
Interest rates 0.045 0.011 0.100
(0.06) (0.08) (0.10)
Interest Rate Changes -0.877 -1.187 0.515
(0.76) (0.74) (1.31)
* Significant at 90 percent level of confidence
** Significant at 95 percent level of confidence
*** Significant at 99 percent level of confidence
Dummy Variables: Months Since an Election
Dummy Variables: Months Since an Election
Dummy Variables: Months Since an Election
TABLE 6
ESTIMATES OF HAZARD FUNCTION FOR BULL MARKETS BY PARTY
1. Full Sample
Equation Variable 8 months 16 months 24 months
A. Republican -0.016 -0.154 -0.232
(0.48) (0.38) (0.37)
Democrat -0.533 -0.211 0.114
(1.04) (0.51) (0.44)
B. Republican 0.024 -0.120 -0.246
(0.48) (0.38) (0.38)
Democrat -0.486 -0.141 0.149
(1.06) (0.53) (0.46)
Interest Rates 0.014 0.003 -0.001
(.0.2) (0.09) (0.11)
Interest Rate Changes -0.855 -0.749 -0.542
(0.56) (0.68) (1.17)
2. Post-World War I
Equation Variable 8 months 16 months 24 months
A. Republican 0.640 0.100 -0.381
(0.56) (0.55) (0.63)
Democrat -0.313 0.100 0.279
(1.10) (0.50) (0.48)
B. Republican 0.659 0.093 -0.488
(0.57) (0.57) (0.67)
Democrat -0.179 0.262 0.429
(1.15) (0.54) (0.53)
Interest Rates 0.027 0.015 0.043
(0.07) (0.09) (0.11)
Interest Rate Changes -0.969 -1.043 -0.604
(0.64) (0.70) (1.15)
3.Post-World War II
Equation Variable 8 months 16 months 24 months
A. Republican 0.914 * 0.213 0.562
(0.51) (0.64) (0.70)
Democrat 0.036 -0.198
(0.55) (0.59)
B. Republican 0.886 * 0.317 -0.602
(0.51) (0.66) (0.76)
Democrat 0.269 -0.057
(0.58) (0.67)
Interest Rates 0.030 0.010 0.034
(0.07) (0.08) (0.12)
Interest Rate Changes -0.959 -1.189 -0.443
(0.79) (0.74) (1.32)
Notes: Numbers in paranthesis are robust standard errors.
* Significant at 90 percent level of confidence
** Significant at 95 percent level of confidence
*** Significant at 99 percent level of confidence
Dummy Variables: Months Since an Election, By Party
Dummy Variables: Months Since an Election, By Party
Dummy Variables: Months Since an Election, By Party
TABLE 7
ESTIMATES OF HAZARD FUNCTION FOR BEAR MARKETS BY PARTY
1. Full Sample
Equation Variable 8 months 16 months 24 months
A. Republican -1.044 ** -0.757 **
(0.48) (0.38)
Democrat -0.424 -0.439 -0.161
(0.71) (0.43) (0.40)
B. Republican -0.944 * -0.813 *
(0.51) (0.45)
Democrat -0.454 -0.477 -0.166
(0.69) (0.42) (0.41)
Interest Rates 0.019 0.021 0.049
(0.08) (0.07) (0.07)
Interest Rate Changes -0.606 -1.010 -0.195
(1.14) (0.89) (0.87)
2. Post-World War I
Equation Variable 8 months 16 months 24 months
A. Republican -0.841 -0.811 *
(0.53) (0.48)
Democrat 0.122 -0.333 0.115
(0.66) (0.51) (0.38)
B. Republican -0.687 -0.986 *
(0.54) (0.61)
Democrat 0.043 -0.430 0.160
(0.67) (0.52) (0.41)
Interest Rates 0.072 0.025 0.073
(0.07) (0.07) (0.07)
Interest Rate Changes -1.142 -1.147 0.082
(0.92) (0.91) (0.87)
3.Post-World War II
Equation Variable 8 months 16 months 24 months
A. Republican -1.012 -1.116 *
(0.75) (0.61)
Democrat 0.484 0.218 0.331
(0.60) (0.48) (0.48)
B. Republican -0.706 -1.279 *
(0.75) (0.78)
Democrat 0.308 -0.063 0.463
(0.64) (0.48) (0.50)
Interest Rates -0.018 -0.094 -0.027
(0.08) (0.10) (0.10)
Interest Rate Changes -0.628 -2.162 0.868
(1.12) (1.53) (1.14)
Notes: Numbers in paranthesis are robust standard errors.
* Significant at 90 percent level of confidence
** Significant at 95 percent level of confidence
*** Significant at 99 percent level of confidence
Dummy Variables: Months Since an Election, By Party
Dummy Variables: Months Since an Election, By Party
Dummy Variables: Months Since an Election, By Party
TABLE 8
1. Full Sample
Equation Variable 24 months 16 months 8 months
A. Election 0.979 * 1.064 ** 0.710
(0.54) (0.48) (0.52)
Congress 1.231 ** 1.288 *** 0.843 **
(0.51) (0.45) (0.40)
Election*Congress -0.941 -1.283 * -0.847
(0.70) (0.76) (0.88)
B. Election 1.144 * 1.089 ** 0.724
(0.67) (0.50) (0.53)
Congress 1.412 ** 1.307 *** 0.909 **
(0.65) (0.47) (0.41)
Election*Congress -1.107 -1.277 * -0.888
(0.82) (0.80) (0.79)
Interest rates 0.069 0.052 0.046
(0.07) (0.08) (0.07)
Interest Rate Changes -0.038 1.244 -0.642
(1.81) (1.69) (1.19)
2. Post-World War I
Equation Variable 24 months 16 months 8 months
A. Election 1.410 *** 1.349 *** 0.704
(0.55) (0.39) (0.57)
Congress 1.452 *** 1.283 *** 0.604
(0.54) (0.46) (0.44)
Election*Congress -1.937 ** -2.079 ** -1.309
(0.80) (1.05) (1.52)
B. Election 1.765 *** 1.466 *** 0.700
(0.65) (0.42) (0.59)
Congress 1.877 ** 1.345 *** 0.674
(0.74) (0.51) (0.45)
Election*Congress -2.397 *** -2.186 ** -1.431
(0.92) (1.08) (1.31)
Interest rates 0.103 * 0.089 0.041
(0.06) (0.07) (0.06)
Interest Rate Changes -0.463 1.175 -0.709
(1.76) (1.57) -1.175
3.Post-World War II
Equation Variable 24 months 16 months 8 months
A. Election 1.602 ** 1.613 *** 0.985 *
(0.68) (0.62) (0.56)
Congress 2.000 *** 1.769 *** 0.906 *
(0.66) (0.60) (0.50)
Election*Congress -2.941 *** -2.362 ***
(0.80) (0.86)
B. Election 1.862 ** 1.687 *** 0.994
(0.73) (0.64) (0.65)
Congress 2.386 *** 1.785 ** 0.929 **
(0.90) (0.79) (0.47)
Election*Congress -3.400 *** -2.518 ***
(0.96) (0.91)
Interest rates 0.053 0.029 -0.056
(0.08) (0.08) (0.07)
Interest Rate Changes -1.136 0.250 -0.163
(2.06) (2.08) (0.93)
* Significant at 90 percent level of confidence
** Significant at 95 percent level of confidence
*** Significant at 99 percent level of confidence
Dummy Variables: Months Before an Election
ESTIMATES OF HAZARD FUNCTION FOR BEAR MARKETS WITH POLITICAL CONTROL
Dummy Variables: Months Beefore an Election
Dummy Variables: Months Before an Election
TABEL 9
1. Full Sample
Equation Variable 24 months 16 months 8 months
A. Republican 0.874 0.956 0.525
(0.67) (0.65) (0.64)
Democrat 1.178 ** 1.335 *** 1.098 *
(0.54) (0.47) (0.65)
Congress 1.235 ** 1.324 *** 0.855 **
(0.52) (0.47) (0.40)
Republican*Congres -0.573 -0.425 -0.367
(0.78) (0.84) (0.69)
Democrat*Congress -1.425 * -2.698 *** -1.504
(0.85) (0.99) (1.47)
B. Republican 1.045 0.100 0.520
(0.75) (0.64) (0.65)
Democrat 1.530 ** 1.554 *** 1.168 *
(0.71) (0.54) (0.68)
Congress 1.479 ** 1.426 *** 0.930 **
(0,68) (0.50) (0.41)
Repuglician*Congress -0.654 -0.328 -0.321
(0.85) (0.83) (0.68)
Democrat*Congress -1.856 * -2.978 *** -1.710
(1.05) (0.93) (1.30)
Interest Rates 0.101 0.100 0.049
(0.08) (0.07) (0.07)
Interest Rate Changes -0.133 1.145 -0.795
(1.86) (1.71) (1.39)
2. Post-World War I
Equation Variable 24 months 16 months 8 months
A. Republican 1.679 *** 1.843 *** 1.056 **
(0.62) (0.54) (0.52)
Democrat 0.904 0.904 -0.834
(0.71) (0.63) (1.14)
Congress 1.518 *** 1.496 *** 0.662
(0.58) (0.51) (0.43)
Republican*Congres -1.375 ** 0.371
(0.68) (0.90)
Democrat*Congress -1.930 * -2.557 **
(1.04) (1.04)
B. Republican 1.935 *** 1.846 *** 1.020 *
(0.68) (0.55) (0.54)
Democrat 1.366 1.166 * -0.939
(0.87) (0.68) (0.85)
Congress 1.898 ** 1.561 *** 0.719 *
(0.75) (0.56) (0.44)
Repuglician*Congress -0.161 ** 0.495
(0.81) (0.87)
Democrat*Congress -2.501 ** -2.776 ***
(1.22) (0.99)
Interest Rates 0.102 0.086 0.038
(0.07) (0.07) (0.06)
Interest Rate Changes -0.472 1.057 -0.717
(1.79) (1.61) (1.11)
3. Post-World War II
Equation Variable 24 months 16 months 8 months
A. Republican 2.369 *** 2.331 *** 0.838
(0.78) (0.79) (0.66)
Democrat 0.862 0.945 1.460 *
(1.01) (1.03) (0.82)
Congress 2.288 *** 2.076 *** 0.854
(0.83) (0.75) (0.54)
Democrat*Congress -2.386 ** -2.023 **
(1.03) (0.98)
Republican 2.756 ** 2.357 *** 0.878
(0.90) (0.87) (0.69)
B. Democrat 0.775 0.873 1.506
(1.16) (1.13) (1.18)
Congress 2.754 *** 2.125 ** 0.880 *
(0.99) (0.93) (0.49)
Democrat*Congress -2.270 * -1.960 *
(1.30) (1.10)
Interest Rates -0.027 -0.016 -0.058
(0.08) (0.07) (0.07)
Interest Rate Changes -2.275 -0.317 0.004
(2.13) (2.19) (1.08)
Notes: Numbers in paranthesis are robust standard errors.
* Significant at 90 percent level of confidence
** Significant at 95 percent level of confidence
*** Significant at 99 percent level of confidence
Dummy Variables: Months Before an Election, By Party
ESTIMATES OF HAZARD FUNCTION FOR BEAR MARKETS BY PARTY WITH POLITICAL
CONTROL
Dummy Variables: Months Before an Election, By Party
Dummy Variables: Months Before an Election, By Party
TABLE 10
1. Full Sample
Equation Variable 8 months 16 months 24 months
A. Election -0.176 -1.015 -1.598 **
(0.65) (0.72) (0.79)
Congress -0.232 -0.552 -1.333 **
(0.38) (0.42) (0.58)
Election*Congress 0.083 1.344 2.676 ***
(0.94) (0.85) (0.97)
B. Election -0.175 -0.993 -1.634 **
(0.66) (0.72) (0.82)
Congress -0.232 -0.493 -1.310 **
(0.40) (0.44) (0.60)
Election*Congress 0.128 1.338 2.727 ***
(0.07) (0.85) (0.99)
Interest rates -0.001 0.001 0.006
(0.07) (0.09) (0.10)
Interest Rate Changes -0.827 -0.556 -0.430
(0.83) (0.66) (1.22)
2. Post-World War I
Equation Variable 8 months 16 months 24 months
A. Election -0.359 -1.264 -1.842
(1.03) (1.11) (1.17)
Congress -0.592 -1.002 * -1.794 **
(0.46) (0.54) (0.81)
Election*Congress 1.105 2.172 * 3.325 **
(1.16) (1.26) (1.47)
B. Election -0.317 -1.227 -1.929
(1.03) (1.12) (1.21)
Congress -0.550 -0.850 -1.729 **
(0.52) (0.60) (0.87)
Election*Congress 1.114 2.153 * 3.441 **
(1.16) (1.26) (1.47)
Interest rates 0.012 0.017 0.026
(0.07) (0.08) (0.11)
Interest Rate Changes -0.837 -0.670 -0.356
(0.78) (0.69) -1.39
3.Post-World War II
Equation Variable 8 months 16 months 24 months
A. Election -0.241 -1.165 -1.795
(0.98) (1.10) (1.16)
Congress -0.721 -1.400 -1.916 *
(0.63) (0.91) (1.16)
Election*Congress 1.193 2.721 * 3.321 *
(1.37) (1.59) (1.79)
B. Election -0.221 -1.076 -2.022
(0.96) (1.10) (1.29)
Congress -0.611 -1.135 -1.972 **
(0.69) (0.94) (0.82)
Election*Congress 1.168 2.693 * 3.912 **
(1.33) (1.44) (1.61)
Interest rates 0.029 0.030 0.008
(0.07) (0.07) (0.10)
Interest Rate Changes -0.567 -0.911 -1.474
(1.02) (0.84) (1.70)
* Significant at 90 percent level of confidence
** Significant at 95 percent level of confidence
*** Significant at 99 percent level of confidence
ESTIMATES OF HAZARD FUNCTION FOR BULL MARKETS WITH POLITICAL CONTROL
Dummy Variables: Months Since an Election
Dummy Variables: Months Since an Election
Dummy Variables: Months Since an Election
TABEL 11
1. Full Sample
Equation Variable 8 months 16 months 24 months
A. Republican 0.188 -0.504 -0.964
(0.60) (0.67) (0.74)
Democrat -0.514 -0.154 0.227
(1.04) (0.53) (0.48)
Congress -0.181 -0.317 -0.506
(0.37) (0.38) (0.41)
Republican*Congres -0.351 0.595 1.214
(1.01) (0.87) (0.85)
B. Republican 0.184 -0.504 -1.155
(0.60) (0.67) (0.88)
Democrat -0.512 -0.128 0.319
(1.06) (0.54) (0.50)
Congress -0.192 -0.258 -0.455
(0.39) (0.41) (0.42)
Repuglician*Congress -0.280 0.622 1.424
(0.08) (0.86) (0.95)
Interest Rates -0.007 0.004 0.059
(0.08) (0.09) (0.12)
Interest Rate Changes -0.836 -0.574 -0.029
(0.63) (0.68) (1.20)
2. Post-World War I
Equation Variable 8 months 16 months 24 months
A. Republican 0.348 -0.612 -1.111
(1.00) (1.13) (1.19)
Democrat -0.273 0.259 0.503
(1.12) (0.54) (0.55)
Congress -0.471 -0.608 -0.715
(0.46) (0.47) (0.53)
Republican*Congres 0.519 1.208 1.319
(1.20) (1.27) (1.38)
B. Republican 0.353 -0.729 -1.557
(1.06) (1.12) (1.37)
Democrat -0.227 0.362 0.716
(1.15) (0.56) (0.60)
Congress -0.469 -0.464 -0.609
(0.52) (0.51) (0.57)
Repuglician*Congress 0.571 1.368 1.854
(1.25) (1.34) (1.53)
Interest Rates -0.001 0.030 0.107
(0.08) (0.09) (0.13)
Interest Rate Changes -0.870 -0.719 0.119
(0.78) (0.68) (1.29)
3. Post-World War II
Equation Variable 8 months 16 months 24 months
A. Republican 0.149 -0.602 -1.201
(0.97) (0.09) (1.16)
Democrat 0.330 0.078
(0.73) (0.74)
Congress -0.770 -0.858 -0.790
(0.60) (0.63) (0.65)
Republican*Congres 1.780 1.313 1.536
(1.29) (1.31) (1.59)
B. Republican 0.116 -0.667 -1.511
(0.99) (1.16) (1.29)
Democrat 0.547 0.290
(0.69) (0.79)
Congress -0.684 -0.652 -0.720
(0.68) (0.65) (0.64)
Repuglician*Congress 1.710 2.016 1.982
(1.29) (1.37) (1.69)
Interest Rates 0.022 0.035 0.067
(0.07) (0.08) (0.12)
Interest Rate Changes -0.599 -0.971 -0.377
(1.04) (0.75) (1.47)
Notes: Numbers in paranthesis are robust standard errors.
* Significant at 90 percent level of confidence
** Significant at 95 percent level of confidence
*** Significant at 99 percent level of confidence
ESTIMATES OF HAZARD FUNCTION FOR BULL MARKETS BY PARTY WITH POLITICAL
CONTROL
Dummy Variables: Months Since an Election, By Party
Dummy Variables: Months Since an Election, By Party
Dummy Variables: Months Since an Election, By Party
TABEL 12
1. Full Sample
Equation Variable 8 months 16 months 24 months
A. Republican -1.996 ** -1.401 ***
(0.97) (0.53)
Democrat -0.636 -0.596 -0.256
(0.71) (0.40) (0.43)
Congress 0.677 ** 0.383 0.213
(0.35) (0.36) (0.41)
Republican*Congres 2.045 * 1.577 **
(0.36) (0.72)
B. Republican -2.054 ** -1.640 ***
(1.02) (0.64)
Democrat -0.714 -0.593 -0.234
(0.71) (0.42) (0.45)
Congress 0.753 ** 0.431 0.221
(0.37) (0.35) (0.41)
Repuglician*Congress 2.062 * 1.785 **
(1.13) (0.77)
Interest Rates -0.036 0.051 0.086
(0.11) (0.08) (0.07)
Interest Rate Changes 1.285 -0.141 -0.591
(1.63) (1.11) (1.90)
2. Post-World War I
Equation Variable 8 months 16 months 24 months
A. Republican -1.806 * 1.276 **
(0.95) (0.55)
Democrat -0.011 -0.356 2.652
(0.65) (0.49) (0.44)
Congress 0.373 0.012 -0.254
(0.43) (0.47) (0.58)
Republican*Congres 2.533 ** 1.594 *
(1.13) (0.98)
B. Republican -1.851 * -1.580 **
(1.00) (0.62)
Democrat -0.007 -0.363 0.346
(0.68) (0.51) (0.49)
Congress 0.433 0.075 -0.287
(0.44) (0.43) (0.60)
Repuglician*Congress 2.581 ** 1.987 *
(1.19) (1.02)
Interest Rates 0.022 0.051 0.101
(0.08) (0.07) (0.07)
Interest Rate Changes 0.222 0.089 -0.209
(1.13) (1.13) (1.73)
3. Post-World War II
Equation Variable 8 months 16 months 24 months
A. Republican -1.565 -1.493 **
(1.03) (0.67)
Democrat 0.123 0.039 0.571
(0.58) (0.54) (0.49)
Congress 0.746 * 0.277 -0.328
(0.43) (0.49) (0.52)
Republican*Congres 2.530 ** 3.027 ***
(1.05) (0.88)
B. Republican -1.461 -1.657 **
(1.07) (0.69)
Democrat -0.061 0.102 0.664
(0.57) (0.49) (0.65)
Congress 0.821 0.433 -0.408
(0.51) (0.54) (0.79)
Repuglician*Congress 2.169 * 3.370 ***
(1.16) (1.23)
Interest Rates -0.078 -0.067 0.046
(0.12) (0.10) (0.09)
Interest Rate Changes 0.584 -1.329 -0.393
(1.69) (1.32) (2.13)
Notes: Numbers in paranthesis are robust standard errors.
* Significant at 90 percent level of confidence
** Significant at 95 percent level of confidence
*** Significant at 99 percent level of confidence
Dummy Variables: Months Since an Election, By Party
ESTIMATES OF HAZARD FUNCTION FOR BEAR MARKETS BY PARTY WITH POLITICAL
CONTROL
Dummy Variables: Months Since an Election, By Party
Dummy Variables: Months Since an Election, By Party
TABLE 13
DATES OF US STOCK MARKET TURNING POINTS (REAL PRICES)
Trough Peak Bear Bull Bear Bull
7/1872
11/1873 2/1876 16 27 YES NO
6/1877 6/1881 16 48 YES YES
6/1882 7/1883 12 13 NO NO
6/1884 11/1886 11 29 NO YES
3/1888 5/1890 16 26 NO YES
12/1890 5/1892 7 17 NO NO
7/1893 4/1894 14 9 YES NO
4/1895 9/1895 12 5 NO NO
8/1896 9/1897 11 13 NO YES
5/1898 3/1899 8 10 NO NO
9/1900 6/1901 18 9 NO YES
10/1903 9/1906 28 35 NO YES
11/1907 8/1909 14 9 NO YES
7/1910 6/1911 11 11 NO NO
12/1914 12/1915 42 12 YES NO
1/1919 7/1919 37 6 NO NO
12/1920 3/1923 17 27 YES NO
10/1923 9/1929 7 71 NO YES(2 ELECTIONS)
6/1932 7/1933 33 13 NO YES
3/1935 2/1937 20 23 NO YES
4/1938 11/1938 14 7 NO NO
5/1942 4/1946 42 47 YES YES
2/1948 6/1948 22 4 NO NO
6/1949 1/1953 12 43 YES YES
9/1953 4/1956 8 31 NO NO
12/1957 7/1959 20 19 YES NO
10/1960 12/1961 15 14 NO YES
6/1962 1/1966 6 43 NO YES
10/1966 12/1968 9 26 NO YES
7/1970 4/1971 19 9 NO NO
11/1971 1/1973 7 14 NO YES
12/1974 9/1976 23 21 NO NO
3/1978 8/1978 18 5 YES NO
4/1980 11/1980 20 7 NO YES
7/1982 6/1983 20 11 NO NO
7/1984 8/1987 13 37 NO YES
12/1987 8/1989 4 20 NO YES
10/1990 1/1992 14 15 NO NO
10/1992 8/2000 9 94 NO YES(2 ELECTIONS)
2/2003 12/2004 30 22 YES YES
10/2005 10/2007 10 24 YES NO
3/2009 2/2011 17 23 YES NO
9/2011 5/2015 7 44 NO YES
2/2016 9 NO
Notes: Along a row, Bear market refers to period from peak in pervious row to trough in the next row.
Bears market refers to period from trough in that raw to prak in that row.
Date of Duration (in months) Was an Election Held during the
FIGURE 1
Timing of Peaks and Troughs Relative to Election
FIGURE 2
Timing of Peaks and Troughs Relative to Election
By Political Party
FIGURE 2, continued
FIGURE 3
Timing of Peaks and Troughs Relative to Election (Real Prices)
FIGURE 4
Timing of Peaks and Troughs Relative to Election
By Political Party (Real Prices)
FIGURE 4, continued
Bibliography
Alesina, A. (1987). "Macroeconomic Policy in a Two-Party System as a Repeated Game."
Quarterly Journal of Economics 102: 651-78.
Alesina, A. (1988). "Macroeconomics and Politics." In National Bureau of Economic Research
Macroeconomics Annual, edited by Stanley Fischer, pp. 13-52. Cambridge, Mass.: M.I.T. Press.
Alesina, A., & Noudel, R. (1992). "Political Cycles in OECD Economies." Review of Economic
Studies 59: 663-688.
Allvine, F. C., & O’Neil, D. E. (1980). “Stock market returns and the presidential election cycle:
Implications for market efficiency.” Financial Analysts Journal 36: 49-56
Chan, A., & Jordan, B. (2004). “The economy, the market, and presidential elections.” Weekly
Economic Review: Banc One Investment Advisors
Döpke, J, & Pierdzioch, C. (2006) “Politics and the stock market: evidence from Germany.”
European Journal of Political Economy, 22(4): 925–943
Füss, R., & Bechtel, M. (2008). “Partisan Politics and Stock Market Performance: The Effect of
Expected Government Partisanship on Stock Returns in the 2002 German Federal
Election.” Public Choice, 135(3/4), 131-150.
Gartner, M. & Wellershoff, K. (1995). “Is There an Election Cycle in American Stock Returns.”
International Review of Economics and Finance, 4(4): 387-410
Herbst, A., & Slinkman, C, (1984). “Political economic cycles in the U.S. stock market.” Financial
Analysts Journal, 4(2): 35-39.
Heckman, J., & Burton S. (1984). "Econometric Duration Analysis." Journal of Econometrics 24:
63-132.
Hibbs, D. (1987). The American Political Economy. Cambridge: Harvard University Press.
Hibbs, D. (1977). "Political Parties and Macroeconomic Policy." The American Political Science
Review 11:1467-87.
Huang, R. (1985). “Common stock returns and presidential elections.” Financial Analyst Journal,
41(2): 58-61
John, M. M., & Thomas, H. M. (2012). “Identifying Bull and Bear Markets in Stock Returns.”
Journal of Business & Economic Statistics, 18(1): 100-112.
Johnson, R., Chittenden, T., & Jensen, R. (1999). “Presidential politics, stocks, bonds, bills, and
inflation.” The Journal of Portfolio Management, 26: 27-31.
Kiefer, N. (1988). "Economic Duration Data and Hazard Functions." Journal of Economic
Literature 26: 646-722.
Klein, Michael W. (1996). “Timing Is All: Elections and the Duration of United States Business
Cycles.” Journal of Money, Credit and Banking 28(1): 84–101.
Lancaster, T. (1990). The Econometric Analysis of Transition Data, New York: Cambridge
University Press.
Lindbeck, A. (1976). "Stabilization Policies in Open Economies 'with Endogenous Politicians."
American Economic Review Papers and Proceedings 66: 1-19.
Lobo, B. (1999). “Jump risk the U.S. stock market: Evidence using political information.”
Financial Economics, 8(2): 149-163
Nordhaus, W. (1975). "The Political Business Cycle." Review of Economic Studies 42: 169-90.
of Economics and Statistics 73 (2): 254-260.
Persson, T., & Guido, T. (1990). Macroeconomic Policy, Credibility and Politics. London:
Harwood Academic Publishers Press.
Pantzalis, C., Stangeland, D. & Turtle, H. (2000). “Political elections and the resolution of
uncertainty: the international evidence.” Journal of Banking and Finance 24(10): 1575–1604
Prechter, R. R. Jr., D. Goel, W. D. Parker, & M. Lampert. (2012). “Social mood, stock market
performance and U.S. presidential elections: a socionomic perspective on voting results,”
September 27.
Rogoff, K. (1990). “Equilibrium Political Budget Cycles." American Economic Review, 80: 21-36.
Rogoff, K., & Sibert, A. (1988). "Equilibrium Political Business Cycles." Review of Economic
Studies, 55: 1-16.
Romer, C. (1992). "Remeasuring Business Cycles." National Bureau of Economic Research
Riley, W., & Luksetich, W. (1980). “The Market Prefers Republicans: Myth or Reality.” The
Journal of Financial and Quantitative Analysis, 15(3), 541-560.
Santa-Clara, P., & Valkanov, R. I. (2003). “The presidential puzzle: Political cycles and the stock
market.” Journal of Finance, 58: 1841-1872.
Sichel, D. (1991). “Business Cycle Duration Dependence: A Parametric Approach.” The Review
of Economics and Statistics,73(2), 254-260
Siokis, F. & Kapopoulos, P. (2007). "Party, Elections and Stock Market Volatility: Evidence From
A Small Open Economy.” Economics and Politics, 19(1): 123-134
Tufte, Edward R. (1978). Political Control of the Economy. Princeton, N.J.: Princeton University
Working Paper no. 4150
Vuchelen, J. (2003). “Electoral systems and the effects of political events on the stock market: The
Belgian case.” Economics and Politics, 15(1): 85-102
Wong, W.K., & McAleer, M. (2009). “Mapping the Presidential Election Cycle in US stock
markets.” Mathematics and Computers in Simulation, 79(11): 3267-3277.
Zhou, H. G., & Rigdon, (201). “Duration Dependence in Bull and Bear Stock markets.” Modern
Economy, 2: 279-286.