Can the president really affect economic growth? Presidential effort and the political business cycle.
Rohlfs, Chris ; Sullivan, Ryan ; McNab, Robert 等
I. INTRODUCTION
Presidential election is often seen as a referendum on the
performance of the economy. Ronald Reagan won the presidency in 1980
with a campaign that asked the question, "Are you better off now
than you were 4 years ago?" and Bill Clinton's 1992 campaign
centered on the reminder, "It's the economy, stupid."
Mitt Romney, the 2012 Republican Presidential nominee, campaigned on his
business acumen, strong economic performance of Massachusetts while he
was governor, and his ability to "turn the economy around"
(Lowry 2012). These slogans and arguments reflect the widespread belief
among voters that economic growth is substantially affected by actions
taken by the president.
Extensive research has shown that economic factors affect votes to
reelect a president (Fair 1996; Silver 2011), but little concrete
evidence exists on the causal relationship between presidential actions
and economic performance. Assassinations of country leaders do not
affect long-term economic growth (Jones and Olken 2009) and changes in
the presidency in the United States do not appear to significantly
influence growth in gross domestic product (GDP) (Alesina and Rosenthal
1995). (1) Historians rank presidents based partly upon their ability to
manage the economy (Schlesinger 1997; Wikipedia 2011). Economic
performance varies widely from year to year, however, and executives who
presided over periods of growth may have simply been lucky. The degree
to which GDP responds to presidential behavior is of enormous practical
importance to voters and policymakers and for general understanding
about how the economy functions.
This article uses a novel estimation strategy to measure the extent
to which a president's actions influence economic growth. If
governors and mayors, for example, exert effort to hire additional
police officers during reelection years (Levitt 1997), then we may
reasonably argue that presidents may exert effort to improve GDP growth
in similar circumstances. Additionally, unlike unemployment, which is
constrained by the "stickiness" of the labor market, GDP
growth may be relatively malleable in the short term, as it can be
affected by changes in hours per worker or usage of existing equipment.
In a reelection year, the president faces an incentive to divert energy
away from other objectives (e.g., improving national security) and
toward increasing economic growth or to time growth-inducing policies to
take effect in election years. Given modern term limits, this incentive
only applies to first-term presidents. If the president's effort
and policies influence economic growth, we should see especially high
levels of growth in election years for first-term presidents, and we
should observe no such effect among second-term presidents.
A large literature exists on the "political business
cycle," and the extent to which growth rises in election years.
Previous work on the United States has generally used data from only a
few decades and has found mixed results, with some researchers finding
large effects (Drazen 2000; Grier 2008; Nordhaus 1975) and others
finding no such effect (Alesina, Roubini, and Cohen 1997). Those who
find an effect attribute it to "artificial" increases caused
by monetary or fiscal policy. However, such studies have failed to
identify robust election cycles in inflation or government spending or
evidence of post-election year crashes (Alesina et al. 1997 and Drazen
2000 provide reviews). This study is the first to propose that the
election year effect reflects a true increase in economic performance
that is caused by an increase in presidential effort. Unlike in previous
studies, data are used on every U.S. president from George Washington to
Barack Obama. New cuts of the data are presented to examine how this
effect varies over time, with the characteristics of the president, and
across sectors of the economy. Unlike in many of these previous studies,
the regressions presented here do not control for potentially endogenous
macroeconomic variables and focus purely on the essentially random
cyclical variation in election years.
While somewhat imprecise, the estimates indicate that increased
presidential effort in election years leads to substantial increases in
growth. For first-term presidents since 1933, real GDP growth has been
from 1.0 to 2.7% percentage points higher in the third and fourth years
of office than in the first 2 years. This effect matches up roughly with
the election cycle, beginning I year before the election in the third
quarter of the third year. No crashes occur immediately afterward in the
beginning of presidents' second terms, and no election year effect
appears for second-term presidents. In some specifications, growth
declines significantly in the fourth year of the second term, when the
president's departure is imminent. A similar negative effect of
election years is observable from 1837 to 1932, when reelection was
uncommon, and for second-term presidents from 1791 to 1836. (2)
This article also provides suggestive evidence that, contrary to
findings by Alesina and Rosenthal (1995), the competence of a president
is a substantial predictor of growth and interacts with effort. One
might expect the returns to effort to be increasing in the ability of
the president, so that the positive effect of election years on growth
would be larger for more capable presidents than for less capable ones.
It is difficult to measure this difference in an objective way, as
historians' rankings of presidents may be biased by random factors
that influenced growth in the presidents' terms and made certain
presidents appear to be particularly effective or ineffective in
managing the economy. Using this measure, however, produces the result
that GDP growth and the effect of election years are substantially
higher for high-quality presidents than for low-quality ones. Effort
might also be expected to be more effective when the president's
party has majorities in the House of Representatives and the Senate. The
data provide suggestive evidence to support this view, and growth is
higher and the electoral cycle is clearer for presidents whose terms
began with undivided governments than for those beginning with divided
governments. One might also expect the amount of effort that a president
exerts to be greatest for elections that are expected to be close. In
the data, however, the effect of election years on growth is larger for
those who won their first elections handily than for those whose initial
elections to office were close.
It is not clear how, exactly, the effort of a presidential
administration translates into results. We find that the effect of
reelection years on growth is spread across multiple sectors, with the
largest effects in the services and government sectors, smaller effects
in the finance, insurance, and real estate and wholesale and retail
trade sectors, and large negative effects in the manufacturing sector.
Given the decline in manufacturing and the rise in services, the actions
taken by the president might include policies to modernize the economy
through building infrastructure and establishing rules and property
rights for new markets. Some of the relevant actions may have also
involved trade agreements, deregulation, or tax reductions. Recent
evidence indicates that considerably more laws are passed in election
years than in other years (Fritze 2010). This effect of effort does not
appear to be associated with a single common strategy that can be
measured in macroeconomic data. The rise in growth is not accompanied by
rises in inflation or in the growth rate of government spending or
private investment, although federal spending appears to shift slightly
away from national defense. (3) There is some evidence of a reduction in
unemployment in reelection years--an effect that may also result from
efforts made by the president.
II. A BRIEF HISTORY OF THE PRESIDENTIAL ELECTION PROCESS
The time-series data in this study are divided into three eras,
based upon changes over time in the likelihood of reelection. Table 1
shows the fraction of first-term presidents who ran for reelection in
the general election, the fraction who won reelection, and an index of
historians' judgments of president quality for three separate eras
in American history. In the first era, from George Washington in 1789
through Andrew Jackson in 1836, every first-term president ran for
reelection in the general election, and 71.4% were reelected. The
average of the zero-to-one index of quality for this era is 0.779. In
the second era, from Martin Van Buren in 1837 through Herbert Hoover in
1932, all three variables are considerably lower. Only 55.0% of
first-term presidents from this era ran for reelection in the general
election, only 30.0% were reelected, and the average rating of quality
was 0.407. Grover Cleveland, who was unseated after his first term but
won 4 years later, is coded as having run but not won reelection.
Primary elections did not occur in the early years of the country, and
candidates from the same party sometimes ran against one another in the
general election. Parties gradually developed nominating processes over
the 1800s and 1900s, and a few states held the first informal primaries
in 1912. The nominating process became solidified in 1972, when all of
the state parties agreed to formal, binding primary votes
(Berg-Andersson 2004; Coleman 2008; Reiter 1985, 1996; U.S. White House
2011c).
In the third era, from Franklin Roosevelt in 1932 through George W.
Bush in 2008, the rates of being nominated by one's party and
winning reelection and ratings of president quality are considerably
higher than in the second era. This change may have been driven by the
increased prevalence of primary voting and the increased ability of
sitting presidents to reach wide audiences through the media. Radio and
television are known to have played a substantive role in increasing the
American public's awareness and fostering their engagement with
national issues such as politics and the economy. While both
technologies were first invented in the 1920s, the "golden
age" of radio, when the medium reached its peak popularity, is
generally thought to be in the 1930s and 1940s, after which it was
replaced by the television (Finkelstein 2000; Hilmes 2011). The
radio's use as a political tool coincided with its rise as a medium
of entertainment, as evidenced by Franklin Roosevelt's
"fireside chats," which began in 1933 (Buhite and Levy 2010).
The division between the second and third eras is selected to coincide
with these fireside chats.
Prior to Franklin Roosevelt's four-term presidency from 1932
to 1945, the Constitution included no term limit on U.S. presidents;
however, no president before Roosevelt had served more than two terms.
Congress amended the Constitution in 1951 to limit presidents from
serving from more than two terms. A president who ascended to office due
to the death of the previous president may serve three terms, provided
that the first term lasted less than 2 years. Harry Truman, who was
president at the time, was specifically exempted from this term limit;
however, after losing the primary in New Hampshire, he chose not to
pursue a third term (David 1954; U.S. Constitution 1947). Since 1932,
only one other president, Lyndon Johnson, was eligible for reelection
and did not win the nomination; Johnson served over two terms from 1963
to 1968 but was eligible for a third. Of the 11 presidents who served
since 1932 and survived to reelection (with Johnson counted twice),
two-thirds won reelection, and the average quality rating was 0.581.
III. GDP GROWTH AND THE ELECTION CYCLE
A. A. Main Findings
This section examines the main empirical question of the
article--the extent to which real GDP growth increases during
presidential reelection years. For presidents in the modern era, from
1933 to the present, election years are found to substantially affect
growth, which is more than 1 percentage point higher in the later 2
years of a typical presidential term than in the first 2 years. Results
from quarterly data indicate that this increase in growth begins in the
second half of the third year--the half of that year that matters most
for the election cycle. In the first era from 1791 to 1836, when voters
typically did not have information on nationwide economic conditions, we
observe the opposite result, with growth rates declining substantially
over a presidential term. This negative effect also appears in the
second era from 1837 to 1932, when reelection was uncommon.
The increase in growth in election years in the modern period only
appears for presidents' first terms in office. For presidents'
second terms, depending on which dataset is used, growth is steady over
the second term or falls in the fourth year--a result that is consistent
with a decline in presidential effort near the end of the second term,
just before leaving office. For two-term presidents in the first two
eras, the pattern is the same as for the modern era, with growth rising
in the reelection year at the end of the first term and falling near the
end of the second term.
B. Detailed Evidence
Figure 1A shows how annual GDP growth varies with the election
cycle for each of the three eras described in Table 1. Percentage points
of real GDP growth are plotted along the vertical axis, and the
horizontal axis plots the remainder of the year after dividing by four,
or mod(year,4). Presidential years, which are divisible by four, appear
on the right of the graph. The averages by mod(year,4) and era are
computed over the full 221-year range of available data. Growth for the
first era is shown by the solid green line. Growth for the second era is
shown by the dashed blue line, and the solid red line plots growth for
the third era. While the GDP data from the first two eras (prior to
1932) are the best available, it should be noted that they are highly
imprecise and rely upon a combination of annual data for some sectors
(such as agricultural and industrial output) and interpolated data for
others. For the interpolated sectors, constant growth rates are assumed
between decennial censuses.
In the first era, reelection was common, but voters did not have
access to information about the state of the national economy. In this
period, we observe a negative pattern over the election cycle, with
growth rates around 5% in years 1 and 2 and around 4% in years 3 and 4.
We observe a sharper election cycle in the second era, when reelection
was rare. Growth rates rose slightly from 3.7% to 4.5% in year 3 and
dropped sharply to 2.3% in year 4. For the third era, when reelection
was common and GDP data were typically available, we observe a very
different pattern, with a sharp increase in growth from 3.0% in the
second year of a term to 4.6% and 4.7% in years 3 and 4. A similar
switch from a negative to a positive election year effect appears if the
third era is defined to begin in 1929 with Hoover, in 1945 with Truman,
or in 1953 with Eisenhower.
Figure 1B uses quarterly data to present a more detailed view of
the election cycle in GDP growth than can be seen in Figure 1A. These
data begin in 1875:Q 1, so that the first era cannot be shown; the data
continue through 2011 :Q3. As with the annual data, these quarterly data
are imprecise for the earlier dates (prior to 1947) and use a
combination of annual information on some sectors (such as agricultural
and industrial output) and interpolated data for others. For the second
era when reelection was rare, we observe a steady decline in growth from
between 1% and 2% in the first two years to between zero and slightly
negative and 1% in years 3 and 4. For the third era, we continue to
observe a positive election year effect. This increase is somewhat less
sharp than in Figure 1A, in part due to differences in the measures and
comparatively due to the addition of the below-average growth quarters
from 2011. Importantly, the rise from roughly 3% growth to slightly
below 4% growth occurs midway through the third year, when the primaries
are beginning, and it lasts through the second quarter of the fourth
year, the last quarter for which voters can observe data prior to the
election.
[FIGURE 1 OMITTED]
If the increase in years 3 and 4 in the third era truly reflects
the effect of increased effort due to the incentive to be reelected,
then it should apply for presidents in their first terms, but not in
their later terms, when reelection was either banned or unlikely. To
evaluate this hypothesis, Figure 2A shows the electoral cycle in GDP
growth for presidents in their first terms, and Figure 2B shows the
8-year cycle for presidents who completed exactly two terms.
The results from Figure 2 support the view that the effects shown
in Figure 1 are driven by reelection incentives. For first-term
presidents in Figure 2A, the decline over the election cycle for the
first era is slightly less than 1 percentage point, from 4.5% and 4.4%
in years 1 and 2 to 3.7% and 3.5% in years 3 and 4. For the second era,
when reelection was rare, we observe a general decline from 3.9% in year
1 to 2.7% in year 4, with a slight increase between years 2 and 3. In
the third era, we observe an increase that is somewhat greater than 1%,
from 2.6% and 1.5% in years 1 and 2 to 3.6% and 3.2% in years 2 and 3.
The results from Figure 2B indicate that this election cycle is
concentrated in the first term. For two-term presidents in the first two
eras, as shown by the dashed line, we observe a sharp increase from
slightly below 4% growth in the third year to above 6% in year 4,
followed by a steady decline to below 2% growth in the last two years of
the presidency. For two-term presidents in 1953 and later, as shown by
the solid red line, we observe a sharp increase from 2.8% and 0.8% in
the first two years to 4.2% and 4.1% in years 1 and 3. Part of this
growth reflects selection bias, as the presidents who are most likely to
be reelected and to enter into the sample as two-term presidents are
those who experienced high growth during election years. Growth is
unsteady but generally declines in the later term, averaging 3.2% in the
last four years. While GDP declines slightly after year 4, it remains
fairly high, and we do not observe the sort of crash that one would
expect if the election year bump was artificial and caused by a monetary
or fiscal "trick."
[FIGURE 2 OMITTED]
We develop a formal econometric model for these given graphical
results. For a given President i in year t, suppose that the growth rate
of real GDP is determined by the following model:
(1) GDP [Growth.sub.it] = [[beta].sub.0] + [[beta].sub.1] Third
[YR.sub.it] + [[beta].sub.2] Fourth [YR.sub.it] + [Year.sub.t] +
[Year.sup.2.sub.t] + [[epsilon].sub.it]
where GDP [Growth.sub.it] is the growth rate of real GDP;
[Year.sub.t] controls for time effects in each year t;
[Year.sup.2.sub.t] is the square of [Year.sub.t], and [[epsilon].sub.it]
is a white noise error term. Third [YR.sub.it] and Fourth [YR.sub.it]
are the dummy variables for the third and fourth year of President
i's term in office, respectively. The parameters [[beta].sub.1] and
[[beta].sub.2] are the coefficients of interest; they measure the full
reduced-form effects of the change in the growth rate of GDP in the
third and fourth year of President i's term in office.
Columns 1-5 in Table 2 show the results obtained from Equation (1).
Columns 6-8 use a variation of Equation (1) with quarterly data. These
regressions include control variables for [Year.sub.it] and
[Year.sup.2.sub.t], plus quarter of the year dummies to control for
seasonal effects. Columns 1 and 2 show results for the first era, from
1791 to 1836; columns 2 and 6 display results for the second era, with
data ranging from 1837 to 1932 in column 2 and from 1875:Q1 to 1933:Q1
in column 4. (4) Columns 3-5, 7, and 8 show results from the third era,
with data ranging from 1933 to 2010 in the annual data and from 1933:Q2
to the 2011 :Q3 in the quarterly data. Columns 1,4, and 7 show data for
first-term presidents, and columns 2, 5, and 8 include data for
second-term presidents. Because reelection was rare in the second era,
results are shown for first- and second-term presidents together. The
second terms of Calvin Coolidge and Lyndon Johnson are treated as first
terms, because at the end of those terms, both had been in office less
than 6 years. The second terms of Franklin Roosevelt and Harry Truman
are not included in the data in columns 5 and 8, because both were
eligible for reelection. Because of data limitations, the quarterly data
from 1948:Q4 and earlier use growth in gross national product (GNP)
rather than GDP as the dependent variable; the specification in column 7
includes a dummy for whether the GNP data are used. (5) All of the
standard errors are adjusted for clustering by president.
The results from Table 2 confirm the general findings from Figures
1 and 2. As column 1 shows, there is no effect of election years on GDP
growth for first-term presidents in the first era, when voters did not
have access to information about nationwide economic conditions. Column
2 shows large but imprecisely estimated and statistically insignificant
negative effects of -2.603 and -2.835 in the third and fourth years in
office for second-term presidents in the first era. This negative effect
could reflect a decline in investment owing to uncertainty about the
identity of the future president, or could reflect a decline in
presidential effort near the end of the second term, when departure is
imminent.
Column 3 shows a similar and also insignificant negative effect of
election years on GDP growth in the second era, when reelection was rare
and departure was also imminent. Relative to the effect in the first
era, this negative effect appears later--in the fourth year rather than
the third--and is less than half the size. In the quarterly regression
in column 6, this negative effect arises in seven of eight of the
quarters in years 3 and 4, with large and significant effects of -1.897
and -1.158 in the second and third quarters of year 4. Interestingly,
this negative effect disappears in quarter four of the fourth year,
possibly because the uncertainty about the future has been resolved by
the election.
Among first-term presidents in the third era in column 4, we
observe a statistically insignificant increase in growth of 0.959 in the
third year of the term and a large and marginally significant increase
of 2.702 in the fourth year. In column 7, we observe positive effects in
the later six of the eight quarters from years 3 and 4. We observe a
large, positive, and significant effect of 2.196 in the third quarter of
year 3 and a large, positive, and marginally significant effect of 2.605
in the fourth quarter of that year. We observe smaller and statistically
insignificant effects of 1.719, 0.934, and 0.628 in the first three
quarters of year 4 and a large, positive, and significant effect of
3.276 in quarter four of year 4. This last positive effect arises after
the election and could not reflect the effects of presidential effort.
It may arise due to a delayed effect of effort in earlier quarters or
the reduced uncertainty associated with investors knowing the identity
of the new president.
In results not shown, we examined whether the results in Table 2
were robust to including other covariates in the model such as partisan
effects and potentially endogenous fiscal measures. For partisan
effects, we created two fractionalization variables that measure the
number of seats controlled by the Democratic Party in the House and
Senate, respectively. Including the two fractionalization variables in
the estimation model, we do not observe significant changes in the sign,
magnitude, or statistical significance of the other explanatory
variables. Setting aside the concerns for endogeneity, we also examined
whether the obtained results are robust to the inclusion of a growth in
government outlays variable. We found for the 1933-2010 sub-sample that
the inclusion of the growth in government outlays variable was
statistically significant at the 1% level (as one might expect a
priori). The sign and statistical significance of the fourth-year
coefficient in column 4, however, did not change. These robustness
checks suggest that the estimated coefficients for the third sub-sample
are robust to the inclusion of a variety of other covariates.
We find that some of the estimates are sensitive to the starting
dates of the sub-samples. Using the beginning of World War I as the
starting point instead of 1933, we observe that the signs of the
coefficients in column 4 of Table 2 remain the same, but their
magnitudes are smaller than presented in this article. Also, the
statistical significance for the fourth-year coefficient in column 4
changes from marginally significant to insignificant. We find similar
results when we shift the starting point for the 1933-2010 sub-sample to
begin in 1951.
Among second-term presidents not eligible for reelection in the
third era in column 5, we observe a large but imprecisely estimated and
insignificant increase in growth in year 3 that disappears to zero in
year 4. As the quarterly results in column 8 show, the increase in year
3 is concentrated in the first two quarters of that year. In the fourth
year in the quarterly data, we observe slight positive and insignificant
effects in quarters one and two, a large and significant negative effect
of -3.849 in quarter three, and a large and insignificant effect of
-4.521 in quarter four. As before, this drop in year 4 may reflect
market uncertainty or a decline in presidential effort. The samples of
years are similar for columns 5 and 8, and the discrepancy between the
quarterly and annual data in year 4--with zero effect in the annual data
but a large negative effect in the quarterly data--probably results from
differences in the way in which the growth variable is measured.
IV. HETEROGENEITY IN THE EFFECTS OF THE ELECTION CYCLE
A. Main Findings
This section explores how the largest election-year increases and
decreases in growth vary across different types of presidents. For
presidents from the third era, from 1933 to the present, the difference
in growth between the first 3 years of the term and the election year is
positive for all of the presidents who survived until reelection except
for Jimmy Carter and Dwight Eisenhower. Large effects can be observed
for Democrats (including Franklin Roosevelt and Harry Truman) and for
Republicans (including Gerald Ford, Richard Nixon, and Ronald Reagan).
The returns to effort are probably greatest for high-quality presidents,
and growth is larger--and the positive election year effect is
larger--for presidents later rated as high quality by historians than
for those later rated as low quality. The effect of election years is
also largest for presidents whose parties initially controlled both
houses of Congress, a result that suggests the returns to presidential
effort are greatest when the president's actions are relatively
unimpeded. Finally, this election year effect is largest for those
presidents who initially commanded high shares of the electoral vote.
The negative effects of reelection years on growth in the first and
second eras are somewhat less widespread than is the positive effect in
the third era, with drops in the election year for about two-thirds of
the presidents. The drops are tremendous for Theodore Roosevelt and
Herbert Hoover and are still very large for many; however, large
positive effects are observable in a few cases as well. The negative
effect of election years is most pronounced for high-quality presidents,
for those whose parties initially commanded both houses of Congress, and
for those presidents who initially received large shares of the
electoral votes. These interactions between election years and
presidential quality, divided government, and presidential popularity
are the mirror images of the interactions observed in the third era and
are consistent with the theory that, when the chances of reelection are
relatively remote, presidential effort to increase GDP growth declines
near the end of the term.
B. Detailed Evidence
Table 3 illustrates how these election year effects vary across
presidents. Real GDP growth rates are shown from the annual data for
every president who survived to reelection from George Washington to
George W. Bush. Column 1 shows growth in the year preceding office,
column 2 shows average growth among non-election years in the first
term, column 3 shows growth in the reelection year of the first term,
and column 4 shows average annual growth in second and later terms.
Column 5 shows the difference in mean growth rates between reelection
years and non-election years in the first term. The asterisks indicate
whether the difference is significant or marginally significant when
each year is treated as an independent observation.
For the third era, the difference in growth between reelection
years and non-election years is positive for 9 of the 11 presidents (all
except Dwight Eisenhower and Jimmy Carter). None of the positive
differences are significant for specific presidents, and one (for George
W. Bush) is marginally significant. The negative difference for Carter
is large and significant. When averaged across all 11 presidents, the
difference comes out to 2.20 percentage points and is statistically
significant. When the standard errors are corrected for clustering by
president, this effect is marginally significant.
Among the 19 presidents in the second era, growth is lower in
reelection years than in other first-term years in 11 cases. One of
these differences (for James Polk) is significant, and one (for Theodore
Roosevelt) is marginally significant. When averaged across all of the
presidents from 1837 to 1932, this difference is -1.50 percentage points
and statistically insignificant. Among the seven presidents in the first
era, the difference between reelection years and other first-term years
is negative in five cases; one of these negative differences (for John
Quincy Adams) is significant. The average of these seven differences is
insignificant at -0.98 percentage points.
We obtain real GDP growth data from the Bank of England to examine
whether presidential effort is a proxy for underlying (and unobserved)
factors that are common among similar economies over time. Notably, the
United Kingdom does not have the same election cycle as the United
States Thus, as a placebo test for the U.S. results, we report data from
1830 to 2008 for the United Kingdom in column 6 of Table 3 (i.e., using
the same dates as in column 5 rather than U.K. election dates). For
similar periods in the third era, we observe that the United States
experienced positive growth in first-term reelection years in contrast
to the negative average growth rates in the United Kingdom. In the
second era, average growth for the United States during reelection years
was lower than non-election years. In comparison, for the United
Kingdom, the opposite is true. These results provide further evidence
that the prospect of being reelected provides modern-day U.S. presidents
incentives to increase GDP growth rates during reelection years.
Next, Figure 3 illustrates the degree to which the election year
effect varies systematically with characteristics of different
presidents. All four panels have the same structure as in Figure 1A, but
the graphs are shown for different types of presidents. Panels A and B
show the electoral cycle in growth for presidents rated as high and low
quality by historians. Panels C and D are restricted to data from 1857
to the present, when the party system had solidified, and they show
results for presidents whose parties did not control the House and
Senate and those whose parties did control both houses at the start of
office. Panels E and F restrict the data to presidents who were
initially elected to office (rather than through succession) and show
results for presidents whose initial elections commanded high and low
fractions of the electoral vote.
The presidential quality index that we use is computed from the
rankings of presidential quality as presented in Wikipedia (2011). The
rankings are computed by using a weighted average for 17 reputable
surveys from presidential historians and other notable scholars on the
subject. The results in these surveys reflect historians' views on
overall presidential quality and not specifically on their economic
competency. Economic policy competence is taken into consideration in
these surveys, but it is not the only factor for the overall measure.
Other measures such as competency on foreign policy, character,
integrity, and so forth are taken into consideration in the rankings as
well.
The quality measures in the literature are dominated by rankings of
overall performance. Most of the other rankings in the literature, which
are not part of the 17 weighted studies in Wikipedia (2011), are highly
correlated with the rankings that we use. For instance, rankings across
multiple decades such as those from Murray and Blessing (1983),
University of Illinois at Chicago (2000), and Faber and Faber (2012)
have correlations of 0.97, 0.96, and 0.79 with our ranking,
respectively. Some rankings such as those from Faber and Faber (2012)
have sub-rankings of various characteristics of presidents. Faber and
Faber (2012) present ratings for presidents on foreign relations,
domestic programs, administration and intergovernmental relations,
leadership and decision making, and personal qualities. These have
correlations of 0.40, 0.51, 0.79, 0.78, and 0.76 with our ranking,
respectively. Thus, the graphs as shown in Figure 3 might change
slightly if different types of sub-rankings were used instead of the
index.
[FIGURE 3 OMITTED]
The general pattern shown in Figure 1A appears in most of the
panels of Figure 3; however, the patterns are extreme for some types of
presidents, and the curves are relatively flat for others. The patterns
shown in Figures 3A and 3B provide suggestive evidence that president
quality is a substantial determinant of economic performance and that it
interacts with presidential effort. For low-quality presidents in Figure
3A, GDP growth is mostly between 2% and 4%. We observe a steady decline
in growth over the election cycle for the first era, and the patterns
are relatively flat for the remaining two eras. For the third era among
low-quality presidents, growth is highest in the first year (possibly
due to lagged effects of the previous president's policies) and
increases steadily over years 2-4. Among high-quality presidents in
Figure 3B, growth is generally higher than for low-quality presidents,
beginning around 3% to 5% for all three eras. We observe a slightly
negative relationship in the first era, a dramatic negative relationship
in the second era, with a large drop to almost zero growth in year 4,
and a dramatic positive relationship in the third era, with growth rates
beginning at 3.1% in the first year (probably low due to lagged effects
of the previous president's policies) and 3.8% in the second year
and rising sharply to 6.8% and 5.9% in years 3 and 4.
The dashed blue line in Figure 3C shows that, for presidents in the
second era whose parties did not control both houses of the legislature,
economic performance was highly variable and does not follow a clear
pattern--growth is initially low at just below 1% in year 1 of the term,
increases sharply to 6.8% and 6.6% in years 2 and 3, and drops off
somewhat to 4.4% if year 4. As the solid red line shows, the pattern is
also somewhat unclear in the third era. Growth begins at 3.7% in year 1
of the term but drops sharply to 0.9% in year 2, then increases to 3.2%
and 4.6% in years 3 and 4. For cases in Figure 3D of presidents that
controlled both houses, the pattern is clearer. Growth steadily declines
over the term in the earlier era when reelection was uncommon, as shown
in the dashed blue line, which starts at 4.0% in the first year and
drops to 1.1% by the fourth year. Having control of both houses appears
to be particularly important for growth in the third era, when
reelection was common. Growth ranged from 0.9% to 4.6%, averaging 3.1%
over the 4 years for presidents with divided governments. For presidents
with undivided governments in the third era, growth ranged from 3.2% to
5.5%, averaging 4.6% and steadily rising over the 4 years of the term.
As Figure 3E shows, for initially unpopular presidents, we observe
a generally flat but slightly negative relationship in the first era. In
the second era, we observe a steady and large increase from years 1 to 3
followed by a sharp decline in the fourth year. Finally, we observe a
large increase from 2% to slightly over 3% in the third era. For
initially popular presidents in Figure 3F, the pattern is more dramatic,
with sharp declines for presidents from the first two eras and sharp
increases for presidents from the third era.
V. SOURCES AND MECHANISMS OF GROWTH
A. Main Findings
This next section examines the ways in which the election cycle
affects growth. Results from sector-specific growth data from 1948 to
2010 indicate that the electoral cycle in growth is spread across
multiple industries. For first-term presidents, much of the rise in
growth comes from the services and government sectors. Large positive
effects can also be seen in the wholesale and retail trade, and finance,
insurance, and real estate sectors, and a moderate-sized effect appears
for construction. We observe a large negative effect on growth in the
manufacturing sector and a moderate-sized negative effect in the mining
sector. The negative effects for second-term presidents look much like
the mirror image of the positive effects for first-term presidents. The
negative effect on growth is concentrated in the wholesale and retail
trade sector, with moderate-sized effects in services, finance,
insurance, and real estate, and communication. The decline is partially
mitigated by an increase in the manufacturing sector.
It is not entirely clear how presidential effort translates into
improvements in economic performance. Because the positive effect of
election years involves a decline in manufacturing and an increase in
services, the growth appears to be associated with the modernization of
the economy and may involve day-to-day policies to establish property
rights and infrastructure for this modernization. We observe a decline
in unemployment in election years, and no clear trends in other
macroeconomic variables such as government spending or inflation. One
factor that may comove with the election cycle is the real interest rate
on federal bonds. Since 1913, interest rates have been affected by
policies made by the Federal Reserve System, which is designed to be an
independent entity but may be influenced by the president. The real
interest rate on federal bonds is lower in years 2 of 3 of the electoral
cycle for first-term presidents; this reduction in interest rates in
years 2 and 3 may have spurred investment and had a lagged effect on
growth in years 3 and 4 of the electoral cycle. Another pattern for
which there is suggestive evidence is a decline in defense-related
expenditure in election years for first-term presidents, accompanied by
increases in federal spending in the "health" and
"other" categories. There is no such change for second-term
presidents. This result is consistent with presidents focusing more on
the economy and less on national security in election years.
B. Detailed Evidence
Table 4 shows the effects of election years on growth in modern
times, separately for different sectors of the economy. Within each
panel, each column shows results from a different OLS regression.
Columns 1 to 11 show results for different sectors of the economy, and
column 12 shows results for all sectors added together. For each sector,
the dependent variable is sector-specific GDP in the current year
divided by total GDP in the previous year. The regressors of interest
are dummies for the third and fourth year in the electoral cycle. The
coefficients on these dummy variables can be interpreted as the amount
of growth in GDP caused in that sector by the third and fourth years in
the electoral cycle. The effects on total GDP growth in column 12 are
approximately equal to the sums of the effects from columns 1 to 11. The
totals are not exactly equal because the 11 sectors do not include the
"rest of the world" and "residual" sectors. Table 4A
shows results for first-term presidents, and Table 4B shows results for
second-term presidents. All of the regressions control for year and year
squared. The sector definitions change in 1981, and each regression also
controls for a dummy for 1981 or later. (6)
The results from Table 4 reveal an interesting combination of
positive and negative sector-specific effects. For first-term presidents
in Table 4A, the total 0.248% increase in GDP in year 3 involves a
decline of nearly 1% of GDP in manufacturing and compensating increases
in GDP attributable to the government, services, and finance, insurance,
and real estate sectors. We also observe a moderate decline in mining
and a moderate increase in construction in the third year. In the fourth
year among first-term presidents, we observe a slightly smaller decline
of -0.527% of GDP in manufacturing and larger increases of 0.643% of GDP
in services, 0.461% in government, 0.341% in trade, and 0.305% in
finance, insurance, and real estate, adding to a total increase of
1.299% of GDP in the fourth year. We also observe a moderate-sized
decline in mining and a moderate-sized increase in construction in the
fourth year. Notably, we find high R2s above 0.90 in columns 1, 4, 5, 6,
8, 9, and 10. This suggests that a large portion of the variation in the
GDP growth rates in these sectors is explained by including dummy
variables for the third and fourth year of President Fs term in office
and control variables including year, year squared, and a dummy for 1981
or later in the regressions.
Among second-term presidents in Table 4B, we observe a large 1.699%
increase in GDP in year 3, two-thirds of which is attributable to an
increase in production in manufacturing. We continue to observe a large
positive effect of 0.584% on GDP due to manufacturing in the fourth
year, but these effects are counterbalanced by a large decline in growth
of trade and moderate declines in growth of finance, insurance, and real
estate, services, and communications, leading to an overall negative
effect of the fourth year on GDP growth of -0.313%. Similar to the high
[R.sup.2]s found in Table 4A, we find values above 0.90 in columns 1,4,
5, 6, 8, 9, and 10 in Table 4B.
To better understand the mechanisms behind the electoral cycle in
growth, Figure 4 plots a variety of macroeconomic indicators against
mod(year,4). Figures 4A and 4B show the rates of growth in real
government spending and revenue for the three eras. Figure 4E shows the
real growth rates in three other key components of GDP--private domestic
investment, consumption, and consumption of durables--for the 1933-2010
era. Figure 4F shows monthly unemployment rates for January, 1948
through October, 2011, presented separately for first-term and
second-term presidents. Figure 4D shows inflation for the three eras,
measured based upon the GDP deflator. Figure 4C shows the real interest
rates on federal bonds for 1800 to 1899 (excluding 1833-1841, for which
no data are available) and 1934-2011. The earlier data show the average
interest rate on "selected federal bonds" by Homer and Sylla
(2005), and the later data show the return on 3-month treasury bills on
the secondary market.
As Figure 4A shows, the electoral cycle in government spending
growth varies substantially by era. In the first era, shown by the green
line, growth in spending was small for the first 3 years of a term and
increased substantially in the election year. In the second era, shown
by the dashed blue line, growth in spending was higher generally and was
greater than 40% in the second year of the term for reasons that are not
clear. Unlike these earlier eras, the red line indicates that there is
no apparent electoral cycle to spending growth in modern times, and the
curve is flat across the 4 years of the election cycle. Government
receipts grow fairly steadily over the election cycle for the era from
1791 to 1836, and they exhibit a rocky increase over the cycle for
1837-1932. The reason for these rises is unclear, as they rise at the
same time that GDP growth was declining. In modern times, as shown by
the red line, growth in government revenue is flat in the third year
despite the rise in income growth, a result that suggests a slight
reduction in tax rates in the third year of the electoral cycle.
Figure 4C provides some very suggestive evidence of electoral
cycles in interest rates; however, it is not clear from these data
whether the patterns reflect causal relationships or simply sampling
variation. For the period 1800-1899, real interest rates on federal
bonds were 0.5-1 percentage point lower in years 3 and 4 than in earlier
years--an effect that may reflect an attempt to stimulate investment in
years close to the reelection. For the modern period from 1934 to 2011,
real interest rates are 3.5% to 3.7% in August to October of the second
year and decline sharply to 2.3% in April to June of the third year. The
rate hovers between 2.3% and 2.6% through July of the third year and
then increases, ranging from 2.6% to 3.5% through the fourth year. The
decline in real interest rates in years 2 and 3 could reflect strategic
behavior, if the government wished to stimulate investment, which it
believed had a lagged effect on growth.
As with government spending, the relationship between inflation and
the electoral cycle varies by era, as shown in Figure 4D. In the first
era, inflation declines steady from the first to the fourth year in
office. In the second era, we observe three relatively low rates of
inflation and a sharp increase in the fourth year. In the modern era, we
observe an inverse u-shape, with slightly higher rates of inflation just
below 4% in years 2 and 3 of the presidential term, relative to rates
just above 3% in years 1 and 4. This rise in inflation in years 2 and 3
mirrors the decline in the real interest rate over the same period, but
this pattern does not appear to match with a clear election-based
strategy, unless it involves a strategy for reducing real interest rates
in years 2 and 3.
[FIGURE 4 OMITTED]
For the other components of GDP in the third era, as shown in
Figure 4E, we observe that real growth in private investment spikes in
year 2, when the real interest rate is lowered. In 2010, private
domestic investment is roughly 12% of GDP, and 12% growth in investment
would contribute 1.5 percentage points of growth in GDP. Investment
growth also rises in year 4 for reasons that are not entirely clear.
Consumption naturally increases with income, and as GDP growth rises
over the election cycle, both total consumption and consumption of
durables rise as well.
The monthly data from 1948 to 2011 on unemployment in Figure 4F do
appear to show an election cycle. For first-term presidents, we observe
a steady rise in the unemployment rate over the first 3 years, followed
by a sharp drop from 5.6% near the end of the third year to 5.0% at the
start of the fourth year; the rate declines further to 4.8% by the end
of the fourth year. This drop is slightly less pronounced but still
appears when 2009-2011, 3 years with particularly high unemployment, are
dropped from the sample. For second-term presidents, we observe some
evidence of an election cycle, but it is less pronounced, with a
temporary increase in unemployment from 4.5% to 5.1% in the second year
in office and a steady rise in the later part of the fourth year, from
4.4% in April to 5.2% in December.
The final set of evidence comes from Table 5, which explores
specific policies in greater detail for the period from 1933 and later.
The structure of Table 5 is the same as for Table 4, with separate
panels for first- and second-term presidents. The outcome variables
include maximum marginal income tax and corporate tax rates in columns 1
and 2, inflation in column 3, and growth in real government expenditure,
broken down by major function: defense in column 4, health (including
Medicare) in column 5, income security programs in column 6, Social
Security in column 7, and all other federal expenditures in column 8.
The results from Table 5 help to clarify some of the findings from
Figure 4. Figure 4B showed a slight increase in tax revenue in year 4 in
the later period, presumably attributable to rising income in election
years. In column 1 of Table 5A, we obtain the imprecise results that,
relative to the first 2 years of the term, income tax rates are 0.289
and 681 percentage points lower in years 3 and 4 for first-term
presidents. The effect is larger for corporate taxes, with declines of
1.710 and 1.377 percentage points in years 3 and 4. It is unlikely that
this pattern is driving the electoral cycle in growth, however, as
larger drops in taxes are observable among second-term presidents (for
whom growth decreases in election years) in Table 5B. In column 3, we
observe a slight increase of 0.662 percentage points in inflation in
year 3 of first-term presidents' terms and slight decreases in
inflation in years 3 and 4 of second-term presidents' terms. The
slight increase for first-term presidents does not continue into the
fourth year, however, and the relationship does not appear to be
sufficiently strong to support the contention by Nordhaus (1975) that
the rise in real growth is driven by inflation. We observe declines in
inflation of -1.170 and -0.943 in years 3 and 4 for second-term
presidents.
One notable, although statistically insignificant, change in fiscal
policy that can be seen in columns 4 to 8 is that growth in defense
spending declines for first-term presidents, with growth 6.437 and 1.520
percentage points lower in years 3 and 4 than in earlier years.
Corresponding increases of +6.283 and +2.394 percentage points can be
seen in health-related expenditures, and we observe a large 10.96
percentage point increase in expenditures in the "other"
category during the fourth year. This pattern does not appear for
second-term presidents, for whom we see moderate-sized declines in
election years in growth in the income security and "other"
categories.
VI. CONCLUSION
This study presents new evidence that the president plays a
substantial role in influencing GDP growth. Sitting presidents in their
first terms face strong incentives to increase growth in election years;
presidents in later terms face no such incentive. To the extent that the
president's effort influences economic activity, we should observe
increases in growth in election years for first-term presidents but not
for second-term presidents. Using data from 1933 to the present, we find
that growth increases by 1.0 to 2.7 percentage points in the third and
fourth years of first-term president's terms. There is no such
pattern for second-term presidents or for presidents in earlier eras,
when reelection was rare. Many of the specifications show a decline in
GDP growth at the end of second-term presidents' terms, an effect
that is consistent with a decline in presidential effort as departure
becomes imminent. The positive effect of reelection years on GDP is
concentrated among high-quality presidents, for whom the returns to
effort are probably largest. The effect is spread across multiple
sectors of the economy and coincides with a drop in unemployment and a
slight shift in federal spending away from national defense and toward
other sectors.
ABBREVIATIONS
GDP: Gross Domestic Product
GNP: Gross National Product
OLS: Ordinary Least Squares
doi: 10.1111/ecin.12111
Online Early publication June 25, 2014
REFERENCES
270toWin.com. 2011. Accessed May 26, 2014. http://www.
270towin.com/
Alesina, A., and H. Rosenthal. Partisan Politics, Divided
Government, and the Economy. Cambridge: Cambridge University Press,
1995.
Alesina, A., N. Roubini, and G. D. Cohen. Political Cycles and the
Macroeconomy. Cambridge, MA: MIT Press, 1997.
Berg-Andersson, R. "How Did We Get Here, Anyway? An Historical
Analysis of the Presidential Nominating Process." The Green Papers:
History, 2004. Accessed May 26, 2014. http://www.thegreenpapers.
com/Hx/NomProcess.html
Buhite, R. D., and D. W. Levy. FDR's Fireside Chats. Norman:
University of Oklahoma Press, 2010.
Coleman, K. J. "Presidential Nominating Process: Current
Issues." Congressional Research Services Report for Congress. Order
code RL34222. 2008. Accessed May 26, 2014. http://www.au.
af.mil/au/awc/awcgate/crs/rl34222.pdf
David, P. T. Presidential Nominating Politics in 1952, Vol. 1.
Baltimore: Johns Hopkins Press, 1954, 33-40.
Drazen, A. "The Political Business Cycle After 25 Years,"
National Bureau of Economic Research (NBER) Macroeconomics Annual 15,
2000, 75-117.
Faber, C. F., and R. B. Faber. American Presidents Ranked by
Performance. Jefferson. NC: McFarland and Company, 2012.
Fair, R. "Econometrics and Presidential Elections."
Journal of Economic Literature, 10(3), 1996. 89-102.
Finkelstein, N. H. Sounds in the Air: The Golden Age of Radio.
Lincoln, NE: Macmillan, 2000.
Fritze, J. 2010. "Election Years Boost Lawmaking," USA
Today, March 26.
Gordon, R. J. 1986. The American Business Cycle: Continuity and
Change. National Bureau of Economic Research Studies in Business Cycles,
Vol. 25. Chicago: University of Chicago Press. Accessed May 26, 2014.
http://www.nber.org/data/abc/
Grier, K. "U.S. Presidential Elections and Real GDP Growth,
1961-2004." Public Choice, 135(3-4), 2008, 337-52.
Hilmes, M. Only Connect: A Cultural History of Broadcasting in the
United States. Boston: Wadsworth, 2011.
Homer, S., and R. Sylla. A History of Interest Rates. New
Brunswick, NJ: Rutgers University Press, 2005.
Jones, B. F, and B. A. Olken. "Hit or miss? The effect of
assassinations on institutions and war." American Economic Journal:
Macroeconomics, 1(2), 2009, 55-87.
Levitt, S. D. "Using Electoral Cycles in Police Hiring to
Estimate the Effect of Police on Crime," American Economic Review,
87(3), 1997, 270-90.
Lowry, R. "The Case for Mitt Romney." Time Magazine,
November 1, 2012.
Measuring Worth. 2014. "What Was the U.S. GDP Then? Annual
Observations in Table and Graphical Format 1790 to the Present."
Accessed May 26, 2014. http://www.measuringworth.com/usgdp/
Murray, R. K. and T. H. Blessing. "The Presidential
Performance Study: A Progress Report." Journal of American History,
70(3), 1983.535-55.
Nordhaus, W. D. "The Political Business Cycle," Review of
Economic Studies, 42(2), 1975, 169-90.
Reiter, H. L. Selecting The President: The Nominating Process in
Transition. Philadelphia: University of Pennsylvania Press, 1985.
--. "The Evolution of the Presidential Nominating Process:
Slouching Toward a Nationwide Primary." HPol's Online Seminar:
The Presidential Nominating Process. 1996. Accessed May 26, 2014.
http://www.hnet.org/~pol/ssha/netnews/f96/reiter.htm
Schlesinger, A. M., Jr. "Ranking the Presidents: From
Washington to Clinton," Political Science Quarterly, 112(2), 1997,
179-90.
Silver, N. "Is Obama Toast? Handicapping the 2012
Election." New York Times Magazine, November 3, 2011.
Tax Policy Center. "Historical Corporate Top Tax Rate and
Bracket: 1909-2010." 201 la. Accessed May 26, 2014.
http://www.taxpolicycenter.org/taxfacts/Content/PDF/
corporate_historical_bracket.pdf
--. "Historical Top Tax Rate." 2011b. Accessed May 26,
2014. http://www.taxpolicycenter.org/taxfacts/
displayafact.cfm?Docid=213
U.S. Bureau of Economic Analysis. "Gross Domestic Product:
Percent Change from Preceding Period." 2011. Accessed May 26, 2014.
http://www.bea.gov/ national/xls/gdpchg.xls
U.S. Bureau of Labor Statistics. "Labor Force Statistics from
the Current Population Survey: series id: LNS14000000." 2011.
Accessed May 26, 2014. http://data.bls.gov/timeseries/lnsl4000000
U.S. Bureau of the Census. "Data." 2014. Accessed May 26,
2014. http://www.census.gov/
U.S. Constitution. Twenty-second Amendment. 1947.
U.S. Federal Reserve Bank of St. Louis. "3-Month Treasury
Bill: Secondary Market Rate (TB3MS)." Federal Reserve Economic Data
(FRED). 2011.
U.S. House of Representatives. "Party Divisions of the House
of Representatives (1789 to present)." Art and History Division.
2011. Accessed May 26, 2014. http://
artandhistory.house.gov/house_history/partyDiv.aspx
U.S. Senate. "Party Division in the Senate,
1789-present." Art and History Division. 2011. Accessed May 26,
2014. http://www.senate.gov/pagelayout/history/one_item_
and_teasers/partydiv.htm
U.S. White House. Economic Report of the President. Washington, DC:
U.S. Government Printing Office, 2011a.
--. "Table 3.1: Outlays by Superfunction and Function,
1940-2016." Historical Tables of the U.S. Budget. 201 lb. Accessed
May 26, 2014. http://www.white house.gov/omb/budget/Historicals
--. "The Presidents." 2011c. Accessed May 26, 2014.
http://www.whitehouse.gov/about/presidents
University of Illinois at Chicago. Office of Public Affairs.
"Scholars Rank the Best and Worst U.S. Presidents." 2000.
Accessed May 26, 2014.
http://www.uic.edu/depts/paff/opa/releases/2000/prese val_release.html
Wikipedia. "Historical Rankings of Presidents of the United
States." 2011. Accessed May 26, 2014.
http://en.wikipedia.org/wiki/Historical_rankings_of_
Presidents_of_the_United_States
(1.) We should note that Alesina and Rosenthal's work does not
differentiate between high- and low-quality presidents, and it is not
clear by how much or in what direction one might expect growth to change
after specific changes in presidents.
(2.) These negative effects provide evidence to support the view
that the correlation between growth and the election cycle is primarily
attributable to effort and not due to presidential learning. If growth
increased over the election cycle primarily due to presidential
learning, then we would expect it to rise continuously into the second
term. Growth is somewhat higher during presidents' second terms
than during their first terms, which lends suggestive support to the
view that presidential learning affects growth as well.
(3.) One variable that comoves with the election cycle is the
electoral cycle for Congress. Between the second and third years of a
president's term, the likelihood that the president's party
controls the Senate declines from 0.70 to 0.60, and the likelihood that
the president's party controls the House of Representatives
declines from 0.65 to 0.40. This drop occurs for both first- and
second-term presidents, however, and it is not clear how having a
divided government would foster immediate growth. For these reasons,
presidential effort appears like a more plausible explanation for the
patterns observed in the data. Data taken from U.S. House of
Representatives (2011) and U.S. Senate (2011).
(4.) Franklin Roosevelt's second term was the first for which
the inauguration was held on January 20. Previous inaugurations occurred
on March 4, so that the previous president remained in office for most
of the first quarter of the year after the election.
(5.) Quarterly GDP growth data are available beginning in 1947;
however, GNP data are used through 1948 so that the variable does not
switch for any president mid-term.
(6.) Sector-specific GDP data are available for both definitions of
sectors for multiple years; 1981 is selected as the cutoff year so that
these sector definitions do not change for any president mid-term.
CHRIS ROHLFS, RYAN SULLIVAN and ROBERT MCNAB *
* Thanks for helpful comments from Jeff Kubik, Casey Mulligan,
Stephanie Rohlfs, Kevin Tsui, and Patrick Warren.
Rohlfs: Vice President of Risk Management. Morgan Stanley, New
York, NY. Phone 1-212-762-0276, Fax 1-212-5074288, E-mail
carohlfs@gmail.com
Sullivan: Assistant Professor of Economics, U.S. Naval Postgraduate
School, Monterey, CA 93943. Phone 1-515-451 9407, Fax 1-831-656-2139.
E-mail rssulliv@nps.edu
McNab: Professor of Economics, U.S. Naval Postgraduate School,
Monterey, CA 93943. Phone 1-831-917-2455, Fax 1-831-656-2139, E-mail
rmmcnab@nps.edu
TABLE 1
Rates of Party Nomination and Reelection
Among First-term Presidents by Time Period
(1) (2) (3)
1789-1836 1837-1932 1933-2008
Ran in general 1.000 0.550 0.917
election 0.714 0.300 0.667
Won reelection 0.779 0.407 0.581
Index of president
quality (0-1) 7 20 12
Observations
Notes: Information on party nominations and elections taken from
270toWin.com (2011), U.S. Census Bureau (2014), and U.S. White House
(2011c). Index on quality is based upon historians' rankings (the
weighted average of rankings presented in Wikipedia 2011), which
range from 1 to 43. The index is 43 minus the ranking, all divided by
42. Calvin Coolidge's and Lyndon Johnson's second terms are
classified as first terms, because Johnson was eligible for a second
reelection and Coolidge would have been under the provisions of the
22nd Amendment. Only presidents who survived until reelection are
included in the sample.
TABLE 2
Ordinary Least Squares (OLS) Estimates of Effect of Election Cycle on
Percentage Points of Real GDP Growth
(1) (2) (3) (4) (5)
1791-1836 1933-2010
Year in First Second First Second
Election Cycle Term Term 1837-1932 Term Term
Third year -0.220 -2.603 0.967 0.959 1.647
(1.018) (1.716) (1.182) (1.069) (1.806)
Fourth year -0.147 -2.835 -1.182 2.702 0.005
(0.916) (1.669) (1.515) (1.384) * (0.902)
[R.sup.2] 0.173 0.521 0.112 0.116 0.220
Observations 26 19 96 48 18
Clusters 7 5 22 13 5
(6) (7) (8)
1933:Q2-2011:Q3
Year in Quarter in 1875:Q1- First Second
Election Cycle Election Cycle 1933:Q1 Term Term
Third year Third year Q1 -0.135 0.273 1.651
(0.724) (1.555) (3.412)
Third year Q2 -0.446 -0.364 3.181
(0.498) (1.583) (2.427)
Third year Q3 -0.369 2.196 -0.956
(0.815) (0.954) ** (2.316)
Third year Q4 -0.035 2.605 1.297
(1.264) (1.423) * (1.404)
Fourth year Fourth year Q1 -0.727 1.719 0.734
(1.235) (1.118) (4.031)
Fourth year Q2 -1.897 0.934 0.989
(0.736) ** (2.206) (1.609)
Fourth year Q3 -1.158 0.628 -3.849
(0.430) ** (0.873) (1.216) **
Fourth year Q4 0.306 3.276 -4.521
(0.623) (1.445) ** (3.195)
[R.sup.2] 0.124 0.163 0.220
Observations 171 196 70
Clusters 12 13 5
Notes: Each column shows results from a different OLS regression in
which the dependent variable is real GDP growth. Columns 1 -5 use
annual data, and columns 6 to 8 use quarterly data of annualized
growth rates. The annual regressions control for year and year
squared, and the quarterly regressions include those controls plus
seasonal dummies for quarters. Due to data limitations, GNP rather
that GDP data are used for 1875:Q1-1948:Q4 in the quarterly data; the
regression in column 7 includes a control for whether GNP data are
used. Standard errors adjust for clustering by president. Of note,
the applied standard error correction is valid asymptotically only.
Franklin Roosevelt's and Harry Truman's second terms are omitted from
the "second term" regressions because both were eligible for
reelection. Sources for growth data described in notes to Figure 1.
** and * indicate 5% and 10% significance, respectively.
Additional details in the text.
TABLE 3
Percentage Real GDP Growth in Election Years and Non-election Years
by President, 1791-2008
(1) (2) (3) (4)
Percentage Real GDP Growth in ...
Second
Year Non-election Reelection and
Preceding Year, First Year, First Later
Office Term Term Terms
Bush II 4.14 1.78 3.47 1.83
Clinton 3.39 3.15 3.74 4.44
Bush I 4.11 1.74 3.39
Reagan -0.27 1.70 7.19 3.73
Carter 5.37 4.43 -0.27
Ford -0.55 -0.21 5.37
Nixon 4.84 2.22 5.31 2.62
L. Johnson 4.37 5.15 5.31
Eisenhower 3.83 3.72 1.98 2.69
Truman 8.08 -4.32 4.32 4.95
F. Roosevelt -13.07 6.16 13.05 9.81
Average, 1933-2008 2.20 2.32 4.52 4.42
Hoover 1.15 -3.02 -13.07
Coolidge 13.17 3.28 2.11
Wilson 4.69 -0.33 6.47 1.60
Taft -10.81 3.85 4.69
T. Roosevelt 5.31 4.03 -7.18 1.78
McKinley -1.65 7.36 2.51 5.31
B. Harrison 5.75 4.59 5.10
Cleveland 1.72 5.25 2.05 -0.19
Arthur 12.50 4.04 -1.65
Hayes 4.14 6.61 8.29
Grant 3.90 3.49 8.36 3.58
A. Johnson 1.13 0 3.90
Lincoln 1.02 7.30 1.13
Buchanan 4.02 3.93 1.02
Pierce 11.55 5.27 4.02
Fillmore 1.39 6.35 11.55
Polk 5.68 7.09 3.37
Tyler 0.27 3.46 5.68
Van Buren 2.96 2.54 0.27
Average, 1837-1932 3.57 3.95 2.45 2.42
Jackson 1.35 7.09 4.86 3.26
J.Q. Adams 5.93 3.71 1.35
Monroe -0.01 2.63 4.96 5.01
Madison 0.23 5.91 1.99 2.67
Jefferson 5.69 3.24 2.05 2.58
J. Adams 3.19 4.45 5.69
Washington 5.98 5.29 7.67
Average, 1791-1836 2.73 4.72 3.74 4.24
(5) (6)
Reelection Minus UK Placebo
Non-election Test
Bush II 1.69 * 0.60
Clinton 0.59 -0.31
Bush I 1.66 -0.49
Reagan 5.48 0.82
Carter -4.71 ** -4.34 **
Ford 5.58 3.21
Nixon 3.09 1.21 **
L. Johnson 0.16 2.86 **
Eisenhower -1.75 -2.53 **
Truman 5.34 4.23 *
F. Roosevelt 1.06 -0.10
Average, 1933-2008 2.20 ** -0.17
Hoover -10.05 1.41
Coolidge -1.17 6.38
Wilson 5.40 -3.92
Taft 0.84 -1.53 *
T. Roosevelt -12.38 * 0.61
McKinley -4.33 -3.60 *
B. Harrison 0.51 -4.19 *
Cleveland -0.72 1.87
Arthur -5.68 -3.34
Hayes 1.67 8.15 *
Grant 2.82 -3.19 *
A. Johnson 3.90 1.96
Lincoln -6.16 -0.35
Buchanan -2.91 0.23
Pierce -1.25 2.64 *
Fillmore 5.20 0.61
Polk -3.72 ** 0
Tyler 2.22 5.93
Van Buren -2.27 -6.06
Average, 1837-1932 -1.50 0.45
Jackson -2.23
J.Q. Adams -2.36 **
Monroe 2.33
Madison -3.92
Jefferson -1.19
J. Adams 1.23
Washington -0.70
Average, 1791-1836 -0.98
Notes: This table lists GDP growth for each president who survived
until reelection, broken down based upon when the year fell in the
presidential election cycle. Reelection year averages growth in
reelection years for presidents prior to 1951 (22nd Amendment).
Annual GDP data from Measuring Worth (2014) are used. ** and * indicate
differences between election years and pre-election years that are
significant under the assumption of independent observations with
heteroskedasticity. No significance test is possible for Ford, who
had only one pre-election year and one election year. In addition to
the U.S. results, we also examine whether there is a common growth
pattern in the United Kingdom. We employ U.K. growth rates and U.S.
President terms for this test in column 6. For real GDP growth in the
United Kingdom, data are from the Bank of England and available from
1830 onwards. Additional details in the text.
TABLE 4
OLS Estimates of Effects of Election Cycle on Contribution to Real
GDP Growth by Sector, 1948-2010
(1) (2) (3) (4)
Panel A: First Term Presidents (41 Years, 12 Presidents)
Agriculture Mining Construction Manufacturing
Third year -0.046 -0.175 0.156 -0.863
(0.075) (0.117) (0.233) (0.864)
Fourth year 0.028 -0.132 0.168 -0.527
(0.059) (0.136) (0.231) (0.578)
[R.sup.2] 0.990 0.656 0.155 0.946
Panel B: Second Term Presidents (22 Years, 6 Presidents)
Third year -0.015 -0.001 0.167 1.140
(0.153) (0.154) (0.119) (0.722)
Fourth year -0.032 0.062 -0.036 0.584
(0.168) (0.164) (0.194) (0.401)
[R.sup.2] 0.992 0.795 0.299 0.979
(5) (6) (7) (8)
Panel A: First Term Presidents (41 Years, 12 Presidents)
Wholesale
& Retail
Transportation Communication Utilities Trade
Third year -0.032 -0.020 -0.038 0.061
(0.096) (0.081) (0.075) (0.210)
Fourth year 0.089 0.075 -0.041 0.341
(0.102) (0.073) (0.094) (0.220)
[R.sup.2] 0.941 0.971 0.691 0.953
Panel B: Second Term Presidents (22 Years, 6 Presidents)
Third year 0.065 0.067 0.116 -0.004
(0.091) (0.103) (0.063) (0.359)
Fourth year -0.080 -0.117 0.076 -0.444
(0.026) ** (0.046) ** (0.089) (0.364)
[R.sup.2] 0.972 0.994 0.786 0.932
(9) (10) (11) (12)
Panel A: First Term Presidents (41 Years, 12 Presidents)
Finance,
Insurance, & Real
Estate Services Government Total
Third year 0.261 (0.253) 0.408 0.507 (0.493) 0.248
(0.309) (1.072)
Fourth year 0.305 (0.268) 0.643 0.461 (0.516) 1.299
(0.352) * (1.078)
[R.sup.2] 0.974 0.984 0.637 0.167
Panel B: Second Term Presidents (22 Years, 6 Presidents)
Third year 0.040 (0.228) 0.080 0.087 (0.330) 1.699
(0.209) (1.388)
Fourth year -0.167 (0.296) -0.175 0.012(0.385) -0.313
(0.358) (0.857)
[R.sup.2] 0.990 0.993 0.697 0.207
Notes: Within each panel, each of columns 1-11 in this table presents
results from a separate OLS regression in which the dependent
variable is real sector-specific GDP divided by the previous year's
total GDP. The regressor of interests in each case are dummy for
whether the remainder of year when divided by four is three or zero.
The coefficients in columns 1-11 add up to roughly the coefficient in
column 12, which measures the effect of an election year on overall
real GDP growth. All 12 regressions control for year, year squared,
and a dummy for 1981 or later, when the sector definitions changed.
Standard errors are adjusted for clustering by president. Sector-
specific growth rates taken from U.S. Census Bureau (2014) and U.S.
White House (2011a). Additional details in the text.
TABLE 5
OLS Estimates of Effects of Election Cycle on Tax Rates, Inflation,
and Growth in Government Expenditures by Function
(1) (2) (3)
Panel A: First Term Presidents
Maximum
Marginal Maximum
Income Corporate
Tax Rate Tax Rate Inflation
Third year -0.289 -1.710 0.755
(1.039) (0.966) (0.784)
Fourth year -0.681 -1.377 0.209
(2.319) (1.404) (0.672)
[R.sup.2] 0.750 0.729 0.110
Years 49 48 44
Presidents 13 13 12
Panel B: Second Term Presidents
Third year -2.891 -1.222 -1.170
(3.761) (1.365) (0.836)
Fourth year -4.630 -2.356 -0.943
(6.685) (3.008) (0.937)
[R.sup.2] 0.953 0.902 0.213
Years 18 18 18
Presidents 5 5 5
(4) (5) (6) (7) (8)
Panel A: First Term Presidents
Percentage Growth in Real Federal Spending on ...
Income Social
Defense Health security Security Other
Third year -6.437 6.283 4.450 0.559 10.96
(5.994) (10.26) (5.153) (1.247) (9.433)
Fourth year -1.520 2.394 -4.531 -1.861 1.127
(2.910) (4.126) (6.753) (1.542) (3.383)
[R.sup.2] 0.397 0.129 0.038 0.651 0.119
Years 44 44 44 44 44
Presidents 12 12 12 12 12
Panel B: Second Term Presidents
Third year -1.039 1.059 -4.890 -2.117 -4.062
(1.472) (1.819) (4.170) (1.484) (10.91)
Fourth year -0.140 -2.117 -5.122 -1.204 -2.109
(3.337) (1.474) (11.52) (1.200) (1.666)
[R.sup.2] 0.169 0.700 0.210 0.870 0.080
Years 18 18 18 18 18
Presidents 5 5 5 5 5
Notes: Table is structured in the same way as Table 4. Tax rates are
taken from Tax Policy Center (2011a, 2011b). Inflation is taken from
the GDP deflator in Measuring Worth (2014). Federal spending by
category is taken from U.S. White House (2011b). Each regression
controls for year and year squared, and standard errors adjust for
clustering by president.
