Content uploaded by Ray Sturm

Author content

All content in this area was uploaded by Ray Sturm on Dec 16, 2015

Content may be subject to copyright.

Does the Market's Vote

Count?

The

Informational

Content of Post-Presidential

Election Returns

RAY R. STURM

RAY R. STURM

is an

associate

lecturer

of

fmance

in the

Department

ot

Finance

at

College

of

Business

Administration,

University

of

Central

Florida

in Orlando, FL.

rsturm@ucf.edu

T

he 2012 presidential election was

perhaps the most important elec-

tion in the modern history of the

United States. The country was

trying to recover from the worst economy

since The Great Depression, President Obama's

leadership based on his first term in office

was questioned, and a victory by GOP chal-

lenger Mitt Romney was ripe for the picking.

Yet, on Election Day (November 6, 2012),

the GOP was shell-shocked by a resounding

defeat not only in the presidential election

but in most congressional elections as well.

Despite the argued lack of economic growth

over the prior four years, the President was

reelected and the balance of power in Con-

gress remained unchanged. In response, the

S&P 500 fell 2.37% (close-to-close) the day

following the election as the market apparently

expressed disapproval of the election's out-

come.

(In addition, media speculated that the

market was turning its attention to the Greek

economic crises.) Then the following day, it

fell another 1.22% as investors apparently more

fully considered the implications for their port-

folios.

This study seeks to determine whether

market reactions to presidential elections, such

as those following the 2012 election, are valu-

able information for investors.

The market efficiency model posits that

the adjustment of prices following a news

event such as the presidential election reflects

all available information about the market

(Fama [1965]). After an election, the market in

aggregate will therefore predict the incoming

President's ability to stimulate the economy

and the resulting future stock price levels; cur-

rent prices then adjust according to the aggre-

gate market's required return. To illustrate,

consider Exhibit 1, which illustrates the well-

known efficient market discounting process.

Assuming a 12% required return over a two-

period time horizon and in the absence of an

election, prices would adjust in an efficient

market from an assumed $100 starting level

to $U2 at Time 0 and finally to $125.44 at

Time 1. Inserting an election at Time 0, if

the market approves (disapproves) of the elec-

tion's outcome, believing that the incoming

President's fiscal policies will result in 10%

higher (lower) future stock prices ($137.98 and

$112.90,

respectively), then prices immediately

following the election would adjust to $123.20

($100.80). This illustration forms the null

hypothesis for this study: that market returns

after the discounting process are not reliably

different regardless of the market's vote.

But there are at least two possible viola-

tions of this process. The first is the possibility

that the market overreacts to the election's

outcome. Consistent with the seminal paper

by De Bondt and Thaler [1985] and many

others that followed, the test characteristic

of overreaction is a reversal in prices. With

respect to the current example, overreaction

could manifest itself in infinite combinations.

SPRING 2014THE JOURNAL OE WEALTH MANAGEMENT 55

EXHIBIT

1

The Market Vote Model: Perfect Efficiency

$140.00

$120.00

$100.00

$80.00

$137.98

$125.44

$112.90

ApprovalDisapprovalNormal

This study is concerned with the value of the market's

vote,

so Exhibit 2 illustrates one alternative hypothesis

for this study. Specifically, the market's overreaction to

an election would be indicative of the market's inability

to predict the President's effectiveness. In this example,

regardless of the market's approval or disapproval, future

prices would be $125.44. Hence from

Exhibit 2, the market's approval would

result in abnormally lower returns and

vice versa. Even if the market expects

future prices to differ from $125.44, the

net result would not change. The justi-

fication for overreaction is well docu-

mented in the literature.

The second alternative hypothesis

of the efficient market discounting pro-

cess illustrated in Exhibit 1 is that the

market correctly, but not completely,

discounts the president's economic

ability as reflected in future prices. From

Exhibit 3, if prices adjust only 5% instead

of

10%

($117.60 and $106.40 for approvals

and disapprovals, respectively), then the

result will be abnormally positive (nega-

tive) returns for market approvals (disap-

provals)

.

The two alternative hypotheses

are especially plausible with presidential

elections due to the difference between

quantitative and qualitative news.

Intuition declares that some infor-

mation is more easily, and thus more

quickly, discounted into prices than other

information, because the estimated effects

of quantitative news (e.g., earnings releases

by companies) on future price levels can

be calculated relatively easily using the

many mathematical models available.

These models provide investors with a

relatively objective new trading price

level. But presidential elections reflect less

tangible information because the infor-

mation introduced into the market

is

non-

numeric. Consequently, the discounting

process after an election reflects much-

1 less-certain mathematical calculations

based on a qualitative news event.

Therefore, the market's discounting

process can be viewed from two perspec-

tives.

One perspective is that the process is essentially a

binary vote in which the aggregate market

is

forecasting

future economic conditions based on political cam-

paigning, and in the case of the 2012 election, the past

job performance of

the

President. The second perspec-

tive is that the process is a precise forecast of the present

EXHIBIT

2

The Market Vote Model: Overreaction

$140.00

$120.00

$100.00

$80.00

;5.44

•

Approval

DisapprovalNormal

56 DOES THE MARKET'S VOTE COUNT? THE INFOR.MATIONAL CONTENT OF POST-PRESIDENTIAL ELECTION RETURNS

SPRING

2014

EXHIBIT

3

The Market Vote Model: Information Not Fully Reflected

$140.00

$120.00

$100.00

$80.00

$137.98

$125.44

$112.90

'

Approval

DisapprovalNormal

value of future price levels consistent with Exhibits 1-3.

Both perspectives are explored in this study.

This study is certainly not the first to examine

the relationship between the President and the stock

market. Many studies have examined the relationship

between stock returns and the President's political party

(Santa-Clara and Valkanov [2003] and others), and the

relationship between stock returns and the presidential

election cycle (Allvine and O'Neill [1980] and others).

Others, such as Nippani and Medlin [2002] and Ferri

[2008] have analyzed short-term stock return responses

to presidential elections.

This study adds to the literature by providing evi-

dence about the market's ability to accurately discount

the outcomes of presidential elections. Market efficiency

predicts that the adjustment process (i.e., the market's

vote) to the election's outcome should result in no sub-

sequent abnormal returns. Alternatively, the presence

of abnormal returns indicates that the lnarket's vote is

systematically incorrect and therefore may be a source

of

value

to investors.

DATA AND METHODOLOGY

Since presidential elections are only held once

every four years, the first challenge in testing the value

ofthe market's vote is the unavoidably small sample size.

Since data for the Dow Jones Industrial

index are available from 1896, and fol-

lowing Jones and Banning [2009], the

Dow Jones Industrial index (without div-

idends) covering the period 1896-2011

is examined. The advantage of this rela-

tively long time series is that it captures

the maximum number of elections (29).

The disadvantage is that it predates most

of the business cycle variables that have

been shown to explain stock returns.

For example, Aaa bond rates are avail-

able from tbe Federal Reserve back to

the year 1919, but even this long time

interval would reduce the sample size by

21%,

from 29 to 23 elections. For the

purposes of this study, maximizing the

already-small sample size is more impor-

tant than decreasing the sample size to

accommodate the availability of control

variables, because a smaller sample size

with more control variables is less compelling than a

larger sample size with fewer control variables.

Since the election news is qualitative in nature,

intuition suggests that the markets in aggregate may not

fully agree on the news' value—at least not immediately.

Hence, it stands to reason that the news' information may

not be reflected in prices as quickly as with quantitative

news.

For example, if

a

company releases earnings that

are inconsistent with market expectations, it is relatively

straightforward to drop the new numbers into a matb-

ematical model and buy or sell shares to the new level.

By contrast, if the market following the 2012 election

(for example) is effectively voting that President Obama

will not be successful in strengthening the economy over

his remaining four

years,

how ineffective will he be and

what is the present value ofthat ineffectiveness?

Estimating the present value of "ineffectiveness"

(as opposed to crisp numbers) is much less straightfor-

ward and requires a longer discounting process. For

this reason, cumulative returns for the one, two, and

three days following presidential elections (CRl, CR2,

and CR3) will be employed as a proxy for the mar-

ket's approval (positive returns) or disapproval (negative

returns) ofthe election's outcome. These returns, col-

lectively referred to as "post-election returns," capture

the market activity for the remainder of the election

week and are calculated as follows:

SPRING

2014

THE

JOURNAL OF WEALTH MANAGEMENT 57

Ret, =£(Ln(P„/P„.,))(1)

where returns (Ret^) refer

to the

cumulative daily

log-

change

in

closing prices over time period t. That is,

the

cumulative returns (Ret_) represent the trading days

for

the remainder of the election week. (During the sample

period, presidential elections

are

held

on the

Tuesday

following

the

first Monday

in

November.)

Prior

to

1984,

the day of

the presidential election

was closed

for

trading,

so the

first closing price (P^)

is Monday's (i.e.,

the day

before

the

election) closing

price. After 1980, the day of the presidential election has

been open

for

trading,

so the

first closing price used

is

Tuesday's (i.e., Electioti Day's) closing

price.

The excep-

tion is the 2000 electioti;

it

was held

on

November

7,

but

the witmer was

not

declared tintil Monday, November

13.

For

2000, therefore,

the

closing price

on

Friday,

November 10, serves as

the

first closing price (P,)-

Equation

(1)

serves

as a

proxy

for the

market's

vote.

Once

the

vote is determined, returtis over various

periods during the presidential election cycle (PEC)

up

to and including the next presidential election are exam-

ined (see Allvine and O'Neill [1980], Booth and Booth

[2003],

Sturm [2009 and 2011] and others

for

discussion

of

the

presidential electioti cycle). Such

a

methodology

serves

to not

otily control

for the

PEC,

but

also

to

help

cotitrol

for

business cycle influences.

In

particular, five

titne-ititervals

of

subsequent returns

are

exatnined:

Year

E4:

represents average daily returns

for

the

day

following

the

post-election period

(the

fourth trading day after

the

election) through

the

remainder

of

the election year.

Year 1, Year 2, and Year

3:

represent average daily

returns

for

the years one through three of the Pres-

ident's term.

Year

4E:

represents average daily returns for the next

election year up

to

and including Election Day.

For

all the

tests,

the 29

post-election cutiiulative

returns

are

divided into

two

groups, positive returns

and negative returns,

to

proxy

for

the market's approval

or disapproval respectively

of the

election's outcome

(i.e.,

the

tnarket's vote). Then,

the

difference between

average subsequent returns across the five time-intervals

is examined.

Wheti

the

market votes

on the

economic value

of the country's vote

for

President, there

are two

fore-

casts

at

issue:

the

direction of future price levels and

the

magnitude

of

those price levels. Therefore,

two

tests

are conducted. First,

the

nonparametric binomial test

of proportions

is

used

to

examine whether the market's

approval

or

disapproval

of

the election's outcome

con-

tains information about

the

proportion

of

subsequent

winning or losing daily returns over the five subsequent

time periods.

The

test statistic

is

calculated

as

follows:

(= {u-p)/{pq/n)

(2)

where

u is the

proportion

of

daily successes

in the

test group,

p is the

proportioti

of

daily successes over

the subsequent return time interval beitig exatnined,

q

is (1

—

p),

and n is the

number

of

returns

in the

test

group.

A

"success"

is

defined

as the

condition when

post-election returns correctly predict

the

president's

effectiveness as proxied

for

by subsequent returns.

Thus within

the

test groups,

w

is the proportion

of

positive returns following market approvals (i.e., positive

post-election returns [RetJ) and the proportion of nega-

tive returns following market disapprovals (i.e., negative

post-election returns [RetJ).

For

example, during

the

Year

E4

across

the

entire sample, 54.0%)

of

the daily

returns were positive and 46.0% were negative. Accord-

ingly,

(p) is

54.0% when testing market approvals

for

this group and 46.0% when testing market disapprovals.

When testing Year

1, (p) is

51.8%

for

approvals

and

48.2%

for

disapprovals,

and

so forth.

The second test,

the

difference

of

means test,

is

employed to test the market's ability to discoutit the mag-

nitude of futtire prices,

and

therefore, returns as follows:

where R^ and R^ are

the

average daily returtis

of

stibse-

quent returns following

the

market's approval

or

disap-

proval respectively,

G_

and

G^

are the variances

oíR^

and

R.

respectively, and

n^

and

n^

are the number of observa-

tions of R^ and

R^

respectively. To be clear. Equation

(3)

will

be

calculated

by

first sorting

the

average daily

returns following

the

post-election period

in

order

of

the post-electioti returns. Then,

the

average daily

sub-

sequent returtis following market approvals during

the

post-election period (i.e., positive returns) will be cotii-

pared

to the

average daily subsequent returns followitig

58

DOES

THE

MARKET'S VOTE COUNT?

THE

INFOR.MATIONAL

CONTENT

OF POST-PRESIDENTIAL ELECTION

RETURNS

SPRING

2014

market disapprovals during the post-election period (i.e.,

negative returns).

As illustrated

in

Exhibit

1,

market efficiency pre-

dicts that no relation should exist while intuition, given

the qualitative nature

of

the news, predicts otherwise.

In the case of presidential elections, investors are buying

and selling based

on

their confidence

in the

President's

ability

to

create economic conditions conducive

to

posi-

tive stock market returns via his

fiscal

policy. (See Sturm

[2011]

for

an analysis of fiscal policy and the presidential

election cycle.) After

the

election

and

from

a

market

efficiency perspective, investors will

buy or

sell shares

to

new

levels that eliminate abnormal return expecta-

tions.

Thus, Equations

(2) and

(3) serve

to

test whether

the conditions subsequent to the post-election returns as

illustrated

in

Exhibit

2 or

Exhibit

3

are present.

THE MARKET'S VOTE

Exhibits

4 and 5

present

the

results

of

testing

the

difference

in

returns subsequent to presi-

dential elections and, therefore,

the

pre-

dictive power

of

the market's approval

or disapproval

of

the election. Exhibit

4 presents

the

nonparametric results

of

testing

the

market's direction, while

Exhibit 5 presents

the

parametric results

of testing

the

market's quantification

of

the election results.

Specifically, Exhibit

4

presents

the

results of testing the proportion of

posi-

tive and negative daily returns following

market approvals

and

disapprovals using

cumulative returns from Equation

(I)

over the one, two, and three days

(CRl,

CR2,

and

CR3, respectively) following

each presidential election.

In the

first

row

of

each panel

is the

proportion

of

successes

(p)

following market approvals,

which

in

the case of approvals is the pro-

portion

of

positive daily close-to-close

returns across the five subsequent return

periods.

The

second

row

presents

the

i-stat

for

approvals using Equation

(2).

In

the

third row of each panel is the pro-

portion of successes

(p)

following market

disapprovals, which

in the

case of

disap-

provals is the proportion of negative daily close-to-close

returns across

the

five subsequent return periods.

The

fourth row

in

each panel presents

the

f-stat

from Equa-

tion

(2) for

market disapprovals.

From Exhibit

4,

the results initially seem to indicate

that

the

market fully discounts the presidential election

results efficiently with respect

to

binomial probabili-

ties.

That is, when

the

market approves

of

an election,

there

are not an

abnormal number

of

positive returns,

and when

the

market disapproves

of

an election, there

are

not an

abnormal number of negative returns. Upon

closer examination, however, there does appear

to be a

slight "reversal effect"

in

Year

2 of

the

PEC

following

market disapprovals.

During Year 2, 48.0% of the daily returns over the

entire 1896-2011 sample period

are

negative (descrip-

tive statistics are not reported). Following market disap-

provals as proxied

for

by

CRl

(Panel A), 48.5% of daily

returns

in

Year

2 are

negative, which

is not

statistically

different from 48.0% (f

=

0.60). Following market disap-

EXHIBIT

4

Binary

Vote

This

exhibit presents

the

results

of

Equation

(2)

as follows:

i={tt-p)/{pq/ri)"'

where

» is the

proportion

of

successes

in the

test group,

p is the

proportion

of

successes

over

the subsequent return time interval being examined,

q\s{'\

-p),

and

/;

is the number

of returns

in the

test group.

A

"success"

is

defined as the post-election returns correctly

predicting

the president's effectiveness as proxied

for

by subsequent returns. Panels A,

B,

and

C

employ post-election returns

CRl,

CR2,

and

CR3 (respectively)

as a

proxy

for

the

market's approval/disapproval

of an

election.

Panel

A:

CRl

Approved

Disapproved

Panel

B:

CR2

Approved

Disapproved

Panel

C:

CR3

Approved

Disapproved

%

Pos.

t

%

Neg.

t

%

Pos.

t

%

Neg.

t

%

Pos.

t

%

Neg.

t

Year

E4

57.2

1.43

49.0

1.38

55.2

0.62

48.0

0.78

53.3

0.68

48.8

0.99

Yearl

52.3

0.53

48.6

0.51

51.2

-0.83

47.1

-1.06

51.7

-0.13

48.0

-0.19

Year

2

52.6

0.63

48.5

0.60

51.6

-0.53

47.3

-0.67

51.1

-1.30

46.0

-1.91*

Year

3

52.4

-1.34

45.4

-1.29

53.3

-0.27

46.1

-0.35

53.2

-0.49

45.7

-0.73

Year

4E

52.6

0.17

47.7

0.17

52.8

0.35

48.0

0.47

52.8

0.35

48.2

0.55

*indicates sigtiificattcc

at

tite

0.Í0

level.

SPRING

2014

THE

JOURNAL

OF

WEALTH MANAGEMENT

59

provals as proxied for by CR2 (Panel B),

47.3%

of daily

returns are negative, which is still not reliable

{t

=

—0.67),

but (i) has turned negative and moved slightly further

away from zero. Finally, following market disapprovals

as proxied for by CR3 (Panel C), only 46.0% of daily

returns are negative—reliable at the 0.10 level

(f =

-1.91).

While modest, this pattern does hint at a bias that will

be confirmed in Exhibit 5. (There is also a slight pat-

tern in Year 4E disapprovals, but nothing close to being

reliable at the 0.10 level.)

Exhibit 5 employs Equation (3) and presents the

main results of testing whether the market's vote counts.

The first row (second row) presents the average per-

centage daily returns for the various periods following

the market's approval (disapproval) during the post-elec-

tion period; the third row presents the difference and the

fourth row presents the z-statistic of

the

difference.

EXHIBIT

5

Difference of Means

This

exhibit presents the results

of

Equation

(3) as follows:

where

R^

and

R^

are the average daily returns

of

subsequent

returns following the market's

approval

or disapproval respectively,

C^

and

CJ^

are the variances

of

R^ and

R^

respectively,

and

n^

and »j are the number

of

observations

of

R^ and R^ respectively. Panels A, B, and

C

employ post-election returns CRl, CR2, and CR3 (respectively) as a proxy for the

market's

approval/disapproval

of an

election.

Panel

A: CRl

Ave.

%Ret. following:

Approval

Disapproval

Differenee

Z

Panel

B:

CR2

Ave.

%Ret. following:

Approval

Disapproval

Difference

Z

Panel

C: CR3

Ave.

%Ret. following:

Approval

Disapproval

Difference

Z

Year

E4

0.09

-0.05

0.13

2.07**

0.02

0.01

0.Ö1

0.13

0.02

0.01

0.01

0.21

Yearl

0.01

0.02

-0.02

-0.59

0.00

0.03

-0.03

-1.12

0.01

0.03

-0.02

-0.59

Year

2

0.00

0.01

-0.01

-0.27

-0.01

0.03

-0.04

-1.37

-0.01

0.05

-0.06

-2.52**

Year

3

0.01

0.06

-0.05

-1.47

0.03

0.04

-0.01

-0.42

0.03

0.05

-0.01

-0.52

Year

4E

0.02

0.02

0.00

-0.06

0.02

0.02

0.00

-0.11

0.02

0.02

0.00

-0.07

^indicates

significance at the 0.05 level.

The first result to be discussed in Exhibit 5 is the

relation between CRl and average daily returns for

the remainder of the election year (Year E4). When the

market approves of an election as proxied for by the next

day's return (CRl), the average daily returns for the

remainder of

the

election year have been about 0.09%.

By contrast, when the market disapproves of an election,

the average daily losses for the remainder of the elec-

tion year have been about 0.05%—a difference of about

0.13%

(0.05 level). Hence, it appears that following a

presidential election, the election's information is not

fully reflected after CRl, consistent with the notion

that qualitative information is more difficult to discount

than qualitative information.

The second result to be discussed from Exhibit 5 is

that over the president's subsequent term, all the differ-

ences in means are negative, consistent with Exhibit 2.

That

is,

across all post-election periods (CR1-CR3) and

across Year 1-4E of the PEC, average

daily returns following disapprovals

exceed average daily returns following

approvals. While most of the differences

are not statistically reliable, the pattern is

still intriguing because it exists without

exception. The lack of statistical signifi-

cance may be an artifact of the inherently

small sample size. This pattern suggests a

long-term reversal effect during most of

the subsequent return periods, as illus-

trated in Exhibit 6.

Exhibit 6 provides a means for

observing average price levels (as cal-

culated from returns presented in

Exhibit 5) over the President's subse-

quent term following the elections. For

consistency with Exhibits 1-3, Exhibit 6

begins with an arbitrary $100 price level

one year before the election (Time -1).

At the election date (Time 0), the data

are separated into market approvals and

disapprovals using CR3 as the proxy.

CR3 is chosen because it captures the

largest difference in returns for Year

2.

For comparison, the expected price

levels in the absence of an election are

also shown, calculated from average

returns over the entire sample period.

60 DOES THE MARKET'S VOTE COUNT? THE INFORMATIONAL CONTENT OF POST-PRESIDENTIAL ELECTION RETURNS

SPRING

2014

EXHIBIT

6

Reversal Effect Following CR3

$150.00

$140.00

$130.00

$120.00

$110.00

$100.00

$90.00

Approvals

•

Disapprovals

Expected

From

a

casual observation of Exhibit

6, it

is clear that

a

reversal effect is present consistent with, but much more

pronounced than

in.

Exhibit

2. The

presence

of

this

reversal effect suggests that the market overreacts

to the

election's outcome.

Recall that

the

purpose

of

employing

CR2 and

CR3 is to consider the possibility that information about

the elections' results

are not

immediately agreed upon

by

the

market,

in

aggregate resulting

in a

slower

dis-

counting

process.

From a market efficiency perspective,

the difference

in

Year E4 results over

the

post-election

period (Exhibit

5,

Panels A,

B, and C)

suggest tbat

the

market requires more than

one day to

fully reflect

the

information provided by the presidential election, as evi-

denced

by the

abnormal difference

of

0.13%

in

returns

(consistent with Exhibit

2)

following CRl.

But by the

end

of

the second

day

following

the

election

(CR2 in

Exhibit

5,

Panel B),

the

market has already discounted

the information fully,

as

evidenced

by the

non-reliable

difference

of

0.01%

(consistent with Exhibit 1).

How-

ever, the opposite effect is present with respect to Year 2

of the PEC.

The Year

2

results suggest that

the

markets

may

slowly overreact to the elections' results, somewhat con-

sistent with Exhibit

2.

Comparing the Year

2

results across

the post-election return period (i.e.. Panels A, B, and C),

the daily returns

in

Year 2 are flat following

CRl approvals

and

0.01% following disap-

provals. Following

CR2

approvals, returns

for Year

2

become slightly negative (-0.01%)

and more positive following disapprovals

(0.02%).

However,

for

market approvals

as

proxied for by the three days after the election

(CR3),

returns

are

still around

-0.01%

but

returns following disapprovals have jumped

to 0.05%. Moreover, the statistical reliability

of the differences across Panels

A, B, and

C increases from

a

very insignificant

-0.27

z-score following CRl,

to

-1.37 following

CR2,

and

then

to a

statistically significant

-2.52 (0.05 level) following CR3.

The Year

2

results

are the

most

intriguing

for

three reasons. First, they

are

consistent with

the

pattern identified

in the

binomial tests of Exhibit

4.

Second, the Year

2 returns

in

Panel

C

present the greatest sta-

tistical reliability of all the results. And third,

they are

not

consistent with

the

results

for

Year E4

fol-

lowing

CRl

(Exhibit

5,

Panel A). Hence,

the

returns

for

the

President's second year

in

office merit further

investigation.

RETURNS EOR THE PRESIDENT'S SECOND

YEAR IN OFFICE

The contradictory evidence between Year

E4

and Year

2 is

difficult

to

reconcile within

the

efficient

market hypothesis. While

it

is straightforward

to

accept

that prices

may not

fully reflect information immedi-

ately (hence the use of CR2 and CR3), once the infor-

mation

is

fully reflected, abnormal returns should

not

be observed, consistent with Exhibit

1.

Even if abnormal

returns

are

observed, tbey would

be

expected

to be

eventually arbitraged away, resulting

in

normal returns

thereafter.

(See

Black [1971]

and

many others

for an

explanation

of

the arbitrage process.)

It

appears that

this arbitrage process

is

observed

in

Exhibit 5, Panel

A.

Abnormal returns

for

Year E4 are observed, suggesting

that prices do not fully and immediately reflect all infor-

mation about

the

elections.

Yet no

abnormal returns

are observed after

tbe

Year E4 time period, suggesting

that

the

information

is

fully reflected

by tbe end of

Year E4.

SPRING

2014

THEJOURNAL

OF WEALTH MANAGEMENT

61

By contrast, the returns during Year 2 do not pro-

vide evidence of

a

clean discounting process. To inves-

tigate the Year 2 conundrum further. Exhibit 7 presents

the results of expanding the

tests.

In particular. Exhibit 1,

Panel

A

presents Year

2

returns following market approvals

and disapprovals as in Exhibit 5, except the post-election

period is extended through eight trading days following

the election. That

is,

(n) in Equation (1) is extended from

three to eight to capture returns for the remainder ofthe

election week, as well as the following week. Panel B

in Exhibit 7 replicates the results of Exhibit 5 for Year 2

using CR3 as a proxy for the market's vote, except that

the sample is divided into the sub-periods 2011-1952

and 1952-1896.

EXHIBIT

7

Year

2 Returns

This

exhibit presents the results

of Equation

(3) for Year 2 returns

as

follows:

where

R and

R^

are the average daily returns

of subsequent

retLirns

following

the market's approval or disapproval respectively, a_ and

C5j

are the variances

of

R^ and R^ respectively, and

«^

and

»^

are the

number

of observations of

R_

and R^ respectively. Panel A sorts

returns

over post-election periods from the day after the election

(CRl)

through the following week

(CR8).

Panel B replicates the

results

in Exhibit

5

over the sub-periods 1896-1951 and 1952-2011

using

CR3 as a proxy for the market's

vote.

Returns

Panel

A: Year 2

Ave.

%Ret. following:

Approval

Disapproval

Difference

Z

Ave.

%Ret. following:

Approval

Disapproval

Difference

Z

Panel

B: Subperiods

following

CR3

Ave.

%Ret. following:

Approval

Disapproval

Difference

Z

CRl

0.00

0.01

-0.01

-0.27

CR5

0.00

0.00

0.00

0.02

CR2

-0.01

0.03

-0.04

-1.37

CR6

0.01

0.00

0.01

0.45

2011-1952

-0.03

0.04

-0.07

-2.19**

CR3

CR4

-0.01

0.00

0.05 0.01

-0.06

-0.01

-2.52**

-0.28

CR7

CR8

0.00 0.01

0.02 0.00

-0.02

0.01

-0.66

0.34

1952-1896

-0.01

0.07

-0.08

-2.85***

*indiccite significance

at

the

0.05 and

0.01

levels

respectively.

From Exhibit 7 Panel A, the results for CR1-CR3

reproduce the results in Exhibit 5. For the remainder of

the extended post-election period, the previously dis-

cussed pattern leading up to CR3 does not continue

beyond CR3, implying that with respect to Year 2, the

market requires three days after the election (i.e., the

remainder of the election week) to discount the elec-

tion's news. Afterwards, Year 2 returns are not abnor-

mally different, consistent with Exhibit 1.

With the unavoidably small sample size, statistical

intuition would question the results as simply a small-

sample anomaly. To explore this possibility further. Panel

B of Exhibit 5 employs CR3 as the proxy for the mar-

ket's vote and divides the sample into two sub-periods.

While this procedure obviously reduces the already-

small sample size, the statistical reliability ofthe results

is not affected. In particular, the reversal effect holds up

over both sub-periods, with a difference in means of

-0.07%

for the most recent period and

-0.08%

for the

earlier period, both statistically reliable at the 0.05 and

0.01 levels, respectively. This result serves to not only

mitigate the small-sample concerns, but also speaks per-

suasively to the effect's consistency.

Another possible explanation for the Year

2

results

is that outliers skew the results. Exhibit 8 presents the

average daily returns in Year

2

as a function ofthe CR3

returns. Exhibit

8

identifies four potential outliers in the

data: the returns following the elections of

1912,

1928,

1952,

and 1972. Considering each potential outlier one

at a time and in order of extremity, the returns fol-

lowing the 1912 election appear to be the most obvious

potential outlier. Dropping this data point from the

analysis reduces the z-value from —2.52 (Exhibit 7)

to -2.24, still a reliable result. (Results for the outlier

tests not reported.) If both 1912 and 1928 are dropped,

the z-value drops further to —1.94, also still a reliable

result. Finally, if 1972 is to be considered an outlier, so

too should 1952. Dropping all four of

these

data points

(1912,

1928, 1952, and 1972) results in

a

z-value of-2.14.

Therefore, the results do not appear to be significantly

influenced by outliers. Moreover, taking Exhibits 7 and

8 together strongly suggests that a Year 2 effect exists

that is robust to methodological considerations.

Since a Year 2 effect appears to be clearly present,

what drives the effect? Answering this question is not

easy, especially since a conclusive explanation for the

PEC has not been offered in the literature. (See Sturm

[2011] for

a

literature review and discussion.) The Presi-

62 DOES THE MARKET'S VOTE COUNT? THE INFORM.ATIONAL CONTENT OF POST-PRESIDENTIAL ELECTION RETURNS

SPRING

2014

EXHIBIT 8

Outlier

Effects

on

Average

Daily Year

2

Returns

0.20%

0.15%

£

0.10%

s

K

0.05%

2

0.00%

="

-0.05%

cd

Q

gjo

-0.10%

2

>

-0.15%

-0.20%

-0.25%

M.

X

1972

X

1928

X

1912

-8%

-6%

-4%

-2%

0%

CR3

2%

4%

6%8%

dent's influence over fiscal policy would seetn to be an

obvious driver of the PEC, but Sturm [2011] examines

and finds no relation between fiscal policy and the PEC.

However, he does consider tax legislation as a potential

driver, which tnay also help explain the Year 2 reversal

effect docutnented in this study.

Since the majority of tax legislation is usually

passed within the first half of the President's term

(Sturm [2011]), the Year 2 effect may reveal the dif-

ference between election campaigning and actual eco-

notnic policies. For example, regardless of whether the

market approves or disapproves of ati election's outcome,

the market's vote is based mostly on information dis-

seminated via campaigning at that time. Because it takes

time for the new President to itnplement policies, the

second year of

his

tertii is the most likely time interval

for the full details of his actual economic/tax policies

(as opposed to the policies presented during the election

campaign) to be revealed. Once they are, the market

recalculates the present value of future price levels, con-

sistent with Exhibits 1—3.

From the evidence in Exhibits 4, 5, and

7,

it appears

tbat the tnarket is usually disappointed in Presidents of

whom it approved and relieved by Presidents of whom it

disapproved. That

is,

the market overreacts. Admittedly,

this explatiation

is

conjecture, but it

is

plausible iti light of

a belief that presidential candidates either have a hidden

economic agenda during their campaign, do not (or are

not able to) follow through on campaign promises, or do

not truly understand the consequences of their planned

economic policies. In the absence of such a

belief,

other

explanations would need to be considered.

CONCLUDING REMARKS

The model of market efficiency predicts that

market prices will adjust completely and immediately

following the release of

a

news event. One of the most

important news events for the markets certainly must be

the election of the Presidetit of the Utiited

States.

So frotn

an efficient market perspective, the stock tnarket will

effectively "vote" on the newly elected President's ability

to create future value in the markets after the election's

results are determined. This study seeks to determine

whether this vote contains value for investors.

The key piece of evidence determining whether

prices fully reflect infortnation

is

the presence or absence

of a difference in returns subsequent to the news release.

Using daily data for the Dow Jones Industrial Average

over the period 1896-2011, this study examines how

SPRING

2014

THE

JOURNAL OF WEALTH MANAGEMENT 63

efficiently the market discounts the results of presidential

elections. Returns for the three days following the elec-

tion are employed as a proxy for the market's approval

or disapproval of the elections' results. Positive returns

indicate the market voting in favor of the incoming

President and vice versa. Then, returns following this

post-election period are examined from the perspective

of the presidential election cycle for the presence of

a

difference.

The results expose three abnormalities that may be

a source of information to investors. First,

a

momentum

effect appears to be present over the remainder of the

election

year,

implying that the market reacts "correctly"

but not fully to the election's outcome. Second, a slight

reversal effect appears to be present across the President's

term. Finally,

a

persuasive "Year

2

reversal effect" appears

in both the direction and the magnitude of average daily

returns during the President's second year in office. The

market experiences an abnormal number of negative

days during the President's second year in office fol-

lowing market approvals, and an abnormal number

of positive days during the President's second year in

office following market disapprovals. Moreover, daily

average returns are

—0.01%

following market approvals

and 0.05% following market disapprovals—a statistically

reliable difference of about 0.06% that is robust across

sub-periods.

Whether the results are an artifact of the unavoid-

ably small sample size or

a

pattern of market inefficiency

is hard to say. But the most intriguing result is the statis-

tically reliable Year 2 reversal, which may be driven by

the difference between the political rhetoric with respect

to economic policy during the election and the actual

facts as revealed during the President's first two years

in office. Given the robustness of this result combined

with the other results, the market's vote may indeed be

a source of value to investors. Therefore, the market's

vote does appear to count.

Black,

F.

"Implications of the Random Walk Hypothesis for

Portfolio Management."

Finatuial

Analysts

Journal

(1971), pp.

16-22.

Booth,

J.R., and

L.C.

Booth. "Is Presidential Cycle in Secu-

rity Returns Merely a Reflection of

Business

Conditions?"

Review

of

Financial

Economics,

12 (2003), pp. 131-159.

De Bondt, W.F.M., and R.H. Thaler. "Does the Stock Market

Overrezctl"

Journal

of

Finance,

40 Quly 1985), pp. 793-805.

Fama, E.F. "The Behavior of Stock-Market Prices." The

Journal

of

Business,

Vol. 38, No. 1 (1965), pp. 34-105.

Ferri, M.G. "The Response of

U.S.

Equity Values to the 2004

Presidential Election." JoMrna/ of

Applied

Finance,

Vol. 18, No.

1 (2008), pp. 29-37.

Jones,

S.T., and K. Banning. "U.S. Elections and Monthly

Stock Market Returns." The

Journal

of

Economics and

Finance,

22 (2009), pp. 273-287.

Nippani, S., and W.B. Medlin. "The 2000 Presidential Elec-

tion and the Stock

Market."

JoMn/a/

of

Economics and

Finance,

Vol. 26, No. 2 (2002), pp. 162-169.

Santa-Clara, P., and R. Valkanov. "The Presidential Puzzle:

Political Cycles and the Stock Market." The

Journal oj

Finance,

Vol. 58, No. 5 (2003), pp. 1841-1872.

Sturm, R.R. "The 'Other'January Effect and the Presidential

Election Cycle."

Applied Financial

Economics,

Vol. 19, No. 17

(August-September, 2009), pp. 1355-1364.

Sturm, R.R. "Economic Policy and the Presidential Elec-

tion Cycle in Stock Returns."JoMmal of

Economics and Finance

(2Gll),DOI:10.1007/sl2197-011-0179-6.

To

order reprints

of

this

article,

please contact

Dewey

Palmieri

at

dpalmieri@iijournals.com or

212-224-3675.

REFERENCES

Allvine, F.C., and D.E. O'Neill. "Stock Market Returns and

the Presidential Election Cycle/Implications for Market Effi-

ciency."

Financial

Analysts

Journal,

Vol. 36, No. 5 (Sept./Oct.

1980),

pp. 49-56.

64 DOES THE MARKET'S VOTE COUNT? THE INFORMATIONAL CONTE.\'T OF POST-PRESIDENTIAL ELECTION

RETURNS

SPRING 2014

©EuromoneyInstitutionalInvestorPLC.Thismaterialmustbeusedforthecustomer's

internalbusinessuseonlyandamaximumoften(10)hardcopyprint-outsmaybemade.No

furthercopyingortransmissionofthismaterialisallowedwithouttheexpresspermissionof

EuromoneyInstitutionalInvestorPLC.Source:JournalofWealthManagementand

http://www.iijournals.com/JPPM/Default.asp