You are on page 1of 53

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228425439

Sector rotation over business-cycles

Article · January 2007

CITATIONS READS

17 11,748

3 authors:

Jeffrey Scott Stangl Ben Jacobsen


Massey University TIAS
8 PUBLICATIONS   34 CITATIONS    58 PUBLICATIONS   1,404 CITATIONS   

SEE PROFILE SEE PROFILE

Nuttawat Visaltanachoti
Massey University
90 PUBLICATIONS   548 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Ben Jacobsen on 01 August 2014.

The user has requested enhancement of the downloaded file.


Sector Rotation over Business Cycles

Jeffrey Stangl∗
Ben Jacobsen
Nuttawat Visaltanachoti

Massey University - Department of Finance and Economics

First draft: July 2006

This draft: August 2009

Abstract
According to conventional market wisdom a sector rotation strategy over different stages of
the business cycles outperforms the market. We introduce a simple way to test its value. In
our test an investor anticipates business cycle stages perfectly and rotates sectors following
popular belief. Even with perfect foresight and ignoring transactions costs sector rotation
would have generated at best a 2.3% annual outperformance since 1948. In more realistic
settings outperformance quickly dissipates. We do find an alternative rotation strategy that
historically would have beaten the market by 7%. Whether by chance or due to fundamentals
time will tell.

JEL Classifications: E32, G10, G12


Keywords: stock market, sector rotation, business cycles, investment strategies

This paper has benefited from presentations at meetings of the Australian Banking and Finance Colloquium
(2007), New Zealand Finance Colloquium (2007) and the Financial Services Institute of Australia (FINSIA)
(2007). We thank Henk Berkman, Utpal Bhattacharya, and Ben Marshall for valuable comments and INFINZ
for awarding this paper the Best New Zealand Paper in Investments award in 2007.

Corresponding author: Massey University, Department of Economics and Finance, Private Bag 102904, North
Shore Mail Centre, Auckland, New Zealand 0745, E-mail: j.stangl@massey.ac.nz
Sector Rotation over Business Cycles

Abstract
According to conventional market wisdom a sector rotation strategy over different stages of
the business cycles outperforms the market. We introduce a simple way to test its value. In
our test an investor anticipates business cycle stages perfectly and rotates sectors following
popular belief. Even with perfect foresight and ignoring transactions costs sector rotation
would have generated at best a 2.3% annual outperformance since 1948. In more realistic
settings outperformance quickly dissipates. We do find an alternative rotation strategy that
historically would have beaten the market by 7%. Whether by chance or due to fundamentals
time will tell.

JEL Classifications: E32, G10, G12


Keywords: stock market, sector rotation, business cycles, investment strategies

2
1. Introduction

According to the Fidelity website technology stocks outperform the index after a trough of

the business cycle. Just after a peak investors are better off putting their money in Utilities.

Other financial websites and advisors share Fidelity’s view on when specific sectors should

perform well over the business cycle. Also Standard & Poor’s and the website

Marketoracle.co.uk recommend Technology after a recession. With the onset of an economic

slowdown Wall Street banker Goldman Sachs1 and CNN Money2 advise investors to target

Utilities. Conventional wisdom tells which cycles perform well across the business cycle and

most practitioners share the same view. Even though as some suggest, "if you are in the right

sector at the right time, you can make a lot of money very fast"3, translating popular beliefs

into an exact sector rotation strategy is not obvious. The problem is to identify exact turning

points and stages of the business cycle contemporaneously. This lack of clarity may explain

why to date academic research has not rigorously tested whether investors can profit from

sector rotation based on conventional wisdom.

Please insert Figure I around here.

While we cannot test whether actual sector rotation works, we can test the fundamental

assumptions underlying sector rotation. Do sector returns differ significantly and predictably

across the business cycle? Does rotating sectors in accordance with popular belief outperform

a simple buy-and-hold strategy? We answer these questions using a simple but new approach.

We give sector rotation the benefit of the doubt and assume an investor can perfectly predict

stages and turning points of the business cycle. As Bodie, Kane and Marcus (2009) put it: “...

1
Reuters (2008)
2
CNN Money (2006)
3
Business Week Online (2002)

3
sector rotation, like any other form of market timing, will be successful only if one anticipates

the next stage of the business cycle better than other investors,” (pg.574). If it is indeed the

business cycle driving outperformance an investor who can perfectly anticipate business cycle

stages and rotates sectors using conventional wisdom should generate the highest risk

corrected outperformance. We then consider how relaxing this perfect foresight assumption

and including transactions costs affect performance.

In our perfect scenario evidence does favor sector rotation but only marginally so. Our base

case covers the 1948-2007 period using NBER defined phases of economic expansion and

recession. We divide NBER phases into smaller sub-periods that coincide with business cycle

stages where popular belief expects optimal sector performance. With a few exceptions –

attributable to chance – sectors that should perform well in different stages show no

significant outperformance. When we combine sector returns across stages in a full-fledged

rotation strategy, we find that investors guided by conventional market wisdom and foresight

of business cycle stages realize risk corrected outperformance of 2.3% annually excluding

transactions costs. To put this number in perspective it is good to realize that for such

clairvoyant investors it is easy to design better market timing strategies.4 When we relax

assumptions and add transaction costs the outperformance quickly dissipates.

We verify robustness of our base case and consider a range of alternative tests, data sets,

performance measures, samples and approaches. We test whether results differ if investors

anticipate changes in turning points earlier or later. However, this worsens outperformance,

which drops from 2.3% to 1.9% and 1% when investors implement sector rotation one month

4
For instance, a simple strategy that invests continuously in the market except during early recession would
perform substantially better than sector rotation.

4
and two months in advance and to 2.2% and 1.8% if investors respond to turning points with a

delay. In addition to NBER business cycles, we construct business cycle stages from the

Chicago Federal Reserve National Activity Index (CFNAI). The Chicago Federal Reserve

Bank builds the CFNAI from well-known financial and economic variables that according to

the literature signal changes in the business cycle. Whether we construct stages from the

CFNAI or use different business cycle proxies suggested in the literature, like the term-spread,

the default-spread, the dividend yield, unemployment and industrial production, our main

result holds. When we divide the sample period in half and look at the two different sub-

periods, we find no improvement. When we divide the stages in half we find no significant

differences in the first half of the stage or the second half of the stage. Different performance

measures do not affect results, regardless whether we use Sharpe ratios, Jensen’s alphas or

more recent performance measures like the Goetzmann, Ingersoll, Spiegel and Welch (2007)

manipulation-proof performance measure. We consider sector and industry returns from

alternative data sources such as Fidelity Select funds, Fama and French sectors, and Standard

& Poor’ sector indices. However, different data sets still leave our main result intact. We

verify whether our results depend on the measure of relative outperformance. For instance,

Chordia and Shivakumar (2002) and Avramov and Chordia (2006) show that size, value, and

momentum factors track business cycles. However, results are similar whether we measure

outperformance using the single index model, the Fama and French three-factor model, or the

Carhart four-factor model.

To put it bluntly, no matter what we try, we find no improvement on our base case

scenario. Our results suggest the popular belief that sector rotation might work is only

marginally correct at best. Different sectors do not significantly and systematically outperform

other sectors and industries over the business cycle in the way conventional market wisdom

5
would like us to belief happens. To prevent any misunderstanding: we do not exclude the

possibility that there are practitioners who profit from sector rotation. What our result

indicates is that, if so, the outperformance of these investors has little to do with what

practitioners claim is the main driver of sector rotation outperformance: systematic variation

in sector returns across the business cycle.

We try whether we can take our results one step further and test for consistent and

significant outperformance of any sector across the business cycle, not only the sectors

suggested by conventional wisdom. We feel with respect to the latter test that there are two

different ways to interpret our evidence. If we consider the different outperformance alpha’s

(Jensen, Fama and French, or Carhart) one could argue these are well in line with the

hypothesis of no significant sector outperformance for all sectors. We find significance levels

only slightly higher to what we would expect in the absence of outperformance. However,

there seems some evidence for an alternative strategy. We would feel uncomfortable rejecting

this alternative strategy outright as a possibility. In sample this strategy generates large

outperformance of 7% annually. The strategy survives most of our robustness checks although

sometimes marginally so. When it does not one could use the often heard argument that “this

could be due to the limited number of observations, but that at least signs are consistent.”

Although with hindsight justification of any strategy is possible one could also argue that

there might be underlying fundamental reasons for the sectors we see outperforming in

different stages: for instance food and entertainment do well during recession, which might

reflect the idea of consumers offering themselves small pleasures in a recession. But contrary

to the conventional wisdom case that dictates when which sectors should perform relatively

well, we feel it is too soon to tell whether these outperformance results are due to chance or

the result of economic fundamentals.

6
Our study contributes to the literature as the first to question the underlying assumption

that the business cycle offers opportunities for profitable sector rotation; at least in the way

conventional wisdom suggests. The perfect foresight approach gives sector rotation the

benefit of the doubt and allows us to do this. Sector rotation seems popular among both

professionals and retail investors based on the number and types of websites dedicated to the

topic. “Sector rotation” returns about 95,400 hits in Google compared with 833,000 and

878,000 for more generic terms like “market timing” and “stock picking”.5 Bodie, Kane and

Marcus (2009) in their textbook state that the notion of sector rotation is “one way that many

analysts think about the relationship between industry analysis and the business cycle,” (pg.

573).6 However, even though there are a number of sector rotation funds,7 it is hard to get

exact numbers on how many investors use sector rotation. A recent study by Beber, Brandt

and Kavajecz (2009) analyzing sector order flows suggest that sector rotation is indeed an

important strategy. Jiang, Yao and Yu (2007) conclude that “industry allocation plays an

important role in mutual fund market timing activities” and that mutual funds shift from high

to low beta stocks in response to macroeconomic information. The demand for sector related

investment products may be an alternative proxy for its popularity. According to the

Investment Company Institute, new cash flows to sector funds account for 20% of new cash

flows to all equity funds.8 One family of mutual funds alone, Fidelity Select, offers investors a

choice of 42 sector funds.9 The number of sector exchange traded funds has also seen

5
July 2009
6
Other textbooks also confirm the important role of sector rotation. For instance Fabozzi (2007) on pg. 581 state,
“Sector rotation strategies have long played a key role in equity portfolio management.” Recent papers that
suggest that sector rotation plays an important role in mutual funds are for instance Elton, Gruber and Blake
(2009) and Avramov and Wermers (2006).
7
Popular funds include the Rydex/SGI All-Cap Opportunity H (RYSRX), Rydex/SGI All-Cap Opportunity A
(RYAMX), Claymore/Zacks Sector Rotation (XRO), and PowerShares Value Line Industry Rotation (PYH)
8
Investment Company Institute (2001)
9
http://personal.fidelity.com/products/funds/content/sector/products.shtml

7
incredible growth from less than 10 in 1998 to over 180 in 2008 with a total net asset value in

these ETFs over 50 billion USD.10

While we are the first to consider sector rotation according to conventional wisdom, the

interest in sector rotation and industry allocation is growing. Several recent studies consider

sector rotation and other time variation in sector and industry returns. Avramov and Wermers

(2006) suggest a link between mutual fund performance related to industry allocation and

business cycle proxies. Hou (2007) finds a significant lead/lag relation between the different

responses of sectors to new economic information. Commodity or basic material industries

respond more quickly to economic news than consumer goods industries. Hong, Torous and

Valkanov (2007) and Eleswarapu and Tiwari (1996) observe that sectors with strong

economic links such as metals, services, and petroleum sectors lead the general market by as

much as two months. Menzly and Ozbas (2004) show how sector performance relates to its

position in the production and consumption supply chain. Conover, Jensen, Johnson and

Mercer (2008) show how sector rotation using monetary conditions may generate

outperformance. Jacobsen and Visaltanachoti (2009) show how sector market timing based on

summer and winter patterns in US sectors outperforms a buy and hold portfolio. O'Neal

(2000) finds sector momentum is an indicator of future sector performance. A recent study by

Beber, Brandt and Kavajecz (2009) studies sector order flows and finds evidence that order

flows between sectors predicts future economic conditions.

10
Investment Company Institute (2009)

8
2. Business Cycles

2.1. NBER business cycle dates

Our base case ‘perfect world’ analysis covers ten business cycles from January 1948 to

December 2007. Two considerations motivate the starting point of our sample. First, we want

to eliminate the possibility of business cycle distortions caused by the Great Depression or

World War II.11 For instance, although the US economy was officially in a depression during

1945, industries still operated at full wartime production. Second, studies such as Stock and

Watson (2002) suggest business cycles durations changed after World War II. Fama (1975) in

part attributes this change to adoption of the 1951 Federal Reserve Accord that allows the

Federal Reserve Bank to moderate business cycles through interest rate adjustments.

The official U.S. government agency responsible for dating business cycles is the National

Bureau of Economic Research (NBER). While NBER cycle reference dates are widely

accepted by academics and practitioners, other measures of business cycle activity are also

available.12 The NBER provides dates for cycle peaks and troughs that define phases of

economic expansion and recession (Table I, Panel A). Panel A also reports business cycle

durations from business cycle peak to business cycle peak. While in the years since 1948

business cycles last on average 71 months, earlier cycles were much shorter than recent

cycles, particularly during phases of expansion.

Please insert Table I around here.

11
See for example Chatterjee (1999) and Cover and Pecorino (2005)
12
For a good survey of business cycle dating methodologies see Chauvet and Hamilton (2005)

9
2.2. Business cycle stages

While the NBER defines broad phases of economic expansion and recession, researchers

and investment practitioners commonly divide these phases into smaller sub-periods. A study

by DeStefano (2004) divides business cycles into two stages of expansion (early/late) and two

stages of recession (early/late). More commonly, investment professionals and practitioner

guides such as Stovall (1996) divide expansions into three stages (early/middle/late) and

recessions into two stages (early/late) to allow for the much longer duration of economic

expansions. We follow that convention.

Please insert Figure II around here.

We measure three stages of expansion (of equal length) from the first month following a

cycle trough date to subsequent cycle peak date and two equal length stages of recession from

the first month following a cycle peak date to the subsequent cycle trough date. We define our

five business cycle stages as early expansion (Stage I), middle expansion (Stage II), late

expansion (Stage III), early recession (Stage IV), and late recession (Stage V). Table I Panel B

reports the duration of expansions, recessions, and stages over the nine business cycles along

with averages. Expansions last approximately five years on average and recessions 10 months.

3. Sector performance across business cycles

3.1. Data description

For our base case scenario we use market, industry, and Treasury bill return data from the

Kenneth French website.13 Market returns represent the total value weighted returns for all

NYSE, AMEX, and NASDAQ listed stocks. The Fama and French 49 industry portfolios

13
http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html

10
follow Compustat SIC classifications. We omit the “Other” industry portfolio leaving returns

for 48 industries. The one-month Treasury bill serves as a proxy for the risk-free interest rate.

For clarity of interpretation, we report results as continuously compounded annualized returns.

Please insert Table II around here.

3.2. Popular guidance on industry performance

Table II shows the particular stage of the business cycle where conventional wisdom

suggests sectors perform best. We follow Stovall (1996) in his popular practitioner guide to

sector investing. He divides the economy into ten basic sectors and then maps the optimal

performance of these sectors to one of five different business cycle stages.14 For example, the

guide suggests that technology and transportation sectors provide the best early expansion

performance, basic materials and capital goods the best middle expansion performance, and so

forth with outperformance shifting to other sectors across the remaining business cycle stages.

We map each of the 48 Fama and French industry portfolios to its corresponding sector and

then map sectors to the business cycle stage expected to provide outperformance.15

3.3. Nominal industry performance

We start with the sector part of the strategy and consider industry performance in different

stages first. We analyze nominal returns in section 3.3 and use risk corrected measures in

section 3.4. In section 3.5 we consider the rotation part of the strategy when we look at the

joint performance across all stages of the business cycle of our base case scenario.

14
Lofthouse (2001) trace a similar approach back to Markese (1986). There are other strategies around. Salsman
(1997) describes an alternative strategy that uses not only the dividend yield (as we do) but also short-term
interest rates combined with precious metal prices.
15
The Stovall (1996) classification is consistent with the macroeconomic indicators shown in Table VII.

11
The crudest test is to make no risk correction and see whether industry returns are high

during the business cycle stage where based on conventional wisdom these industries should

outperform. We consider nominal industry performance across the different stages of the

business cycle first. Are there significant differences in sector outperformance over the course

of a business cycle and does this outperformance coincide where conventional wisdom

suggests that it should? The computer software industry, for example, should outperform the

market during the first stage of expansion and basic materials during the second stage of

expansion. We estimate nominal industry performance by business cycle stage with equation

1 and report our results along with some additional descriptive statistics in Table III.

5
ri ,t = ∑ μi , s Ds ,t + ε t (1)
s =1

We define ri,t as nominal industry returns and Ds a dummy variable that indicates one of

five business cycle stages. As an example, D1 takes the value of 1 during months of early

expansion and zero for all other months. Dummy variables D1, D2, and D3 correspond with the

three stages of economic expansion (early/middle/late) and D4 and D5 with the two stages of

economic recession (early/late). Thus the μi,s coefficients measure nominal industry returns for

each of the five stages. For brevity, Table III only reports industry results for the stage where

conventional wisdom suggests high performance. We report number of observations,

annualized returns, standard deviations, betas, and autocorrelation coefficients (all measured

during the indicated business cycle stage) along with average industry and market results

beneath each stage. For comparison, Table III also has the annualized industry returns over

the entire sample period in the last column. Last, but not least, Table III contains the p-values

of a Wald test that these returns are significantly different across stages. However, in most

cases we reject the null hypothesis of equal industry returns across the business cycle. This

12
result is comforting because if we could not reject the null hypothesis of equal returns, the

practical usefulness of sector rotation as a strategy would be questionable from the start.

Please insert Table III around here.

We compare industry returns in these stages with the market return during the same stage

of the business cycle. As an example, Table III reports that transportation industry returns are

25% on an annualized basis during the months of early expansion in comparison with a

market return of 17% for the same stage and thus provide 8% outperformance. Transportation

industries provide outperformance as expected from conventional wisdom. This realization

does not happen in all cases. Out of the 48 industries, 33 industries have raw returns higher

than the market in the stage when they should perform relatively well. Thus, two out three

industries do offer higher returns in the stage of the business cycle when expectations are that

these industries should outperform the market.

Two stages show a surprising result if we look at industry averages for those stages. The

average 14% return for industries that are supposed to perform well during early expansion

results in a 3% underperformance of the market. Similarly in middle expansion, the average

industry, which according to conventional wisdom should outperform, underperforms the

market by 1%.

Based on this simple approach popular belief does hold in the other three stages. In late

expansion, early recession, and late recession, industries on average outperform the market.

Overall, it appears that nominal sector performance coincides only partially with popular

belief. Eyeballing the risk measures suggests that these results will become stronger if we use

13
risk corrected outperformance. In early expansion and middle expansion, we observe that on

top of lower industry returns, industry risk is actually higher as measured by both beta and

standard deviation. For the other three stages, we observe more mixed results with mostly

lower betas but higher standard deviations.

It seems that even with this crude approach results are disappointing for sector rotation

investing. Historically, many sectors would have done better during early expansion and

middle expansion stages but these are not included in the popular sector allocation strategy.

The more important question for an investor, and the one that we address next, is if risk

adjusted industry outperformance differs significantly across business cycles.

Please insert Table IV around here.

3.4. Risk adjusted industry performance measures

We compare Sharpe ratio’s of industries with the Sharpe ratio of the market in different

stages, and we consider excess market returns, Jensen’s alpha and alpha’s based on the Fama

and French (1992) three-factor model and Carhart (1997) four-factor model. Table IV reports

performance measures as annualized rates with White (1980) heteroskedasticity consistent t-

statistics highlighted for statistical significance at the 10% level. We also test whether these

different alpha performance measures differ significantly over business cycle stages with a

Wald test statistic and report p-values in Table IV under a null hypothesis of equal

outperformance. For brevity, Table IV only reports results for the period of expected optimal

industry performance.

14
Table IV starts with the difference between industry and market Sharpe ratios.16 If

conventional wisdom holds we would expect a positive and statistically significant Sharpe

ratio difference. However, only one early recession industry (Gas & Electric) has Sharpe

ratios significantly higher than the market. All early expansion industries in the technology

sector have statistically significant lower Sharpe ratios than the market. A large majority of

late recession industries also have lower Sharpe ratio performance than the market although

here none of the differences are significant. The best average performance comes from late

expansion industries but none of the differences is significant. Contrary to conventional

wisdom, industries considered optimal for a particular period mostly underperform the market

on a Sharpe ratio performance basis (28 out of 48 sectors). Sharpe ratios may over penalize

for the idiosyncratic volatility inherent if we consider industries. We use alternative risk

adjustments next. Based on equation 2 we estimate excess market industry performance across

business cycles. We run a regression of the difference between industry i and market returns

(ri-rmkt) on the cycle dummy variables (Ds) described previously. The regression coefficient

αmkt is simply market outperformance performance for industry i during business cycle stage s.

5
ri ,t − rmkt ,t = ∑α mkt ,i ,s Ds ,t + ε t (2)
s =1

Additionally, we report Jensen’s alphas that we estimate for each stage of the business

cycle with a modified market model using equation 3.

5 5
ri ,t − rf t = ∑ α J ,i , s Ds ,t + ∑ β1,i , s (rmkt ,t − rf t ) Ds ,t + ε t (3)
s =1 s =1

16
We estimate Ledoit and Wolf (2008) p-values for industry and market Sharpe ratio differences corrected for
potentially non-iid returns and indicate statistically significant differences at the 10% level or higher in bold.

15
We define ri-rf as excess industry returns above the one-month Treasury bill, Ds as one of

five business cycle timing variables, and rmkt-rf the market risk premium.

To ensure our results do not depend on specific performance measures, we include Fama

and French (1992) three-factor and Carhart (1997) four-factor alphas. We estimate the Fama

and French alphas (αF) with an equation similar to (3) where we control for size and book-to-

market ratios in addition to the market. We also estimate Carhart alphas (αC) with a modified

four-factor model that adds an additional momentum factor to the Fama and French three-

factor model.

Regardless of what measure we use, we find very little evidence of significant industry

outperformance in stages when they should outperform as conventional wisdom claims. These

results strengthen our earlier findings for the nominal returns. Based on Jensen’s alpha, we

find 6 industries with significant outperformance, based on the Fama and French three-factor

model, we find only 5 industries with significant outperformance in the stage where they

should outperform and only 2 using the Carhart four-factor model. At the 10% level that is

roughly the number of industries out of 48 that one would expect to show significant

outperformance if there were no significant outperformance.17

As an additional step, we test for differences in market outperformance across the five

business cycle stages (αmkt) using a Wald test. We report p-values in Table IV under a null

hypothesis of equal performance. If industry outperformance is unequal across business cycle

stages holds, we should reject the null hypothesis. We cannot reject the null hypothesis of

17
If anything, it seems that Gas & Electric sectors behave somewhat as popular wisdom suggests. These sectors
show relative good performance during early recession. While for these sectors this performance is insignificant,
it is positive and large, and the limited number of observations might explain the lack of significance.

16
constant excess market industry returns for thirty of the forty-eight industries. This result

differs from the previous Wald test based on constant nominal returns. Sensitivity to general

market movements seems largely to explain differences in industry returns over business

cycles. Wald tests of Fama and French and Carhart alphas also indicate no statistical

difference in industry performance across business cycles. Our results suggest that after

controlling for risk using different measures industry outperformance across business cycles

does not occur – or only marginally so – when conventional wisdom tells us it should.

Finally in unreported results we count the percentage of months in the different stages

where relative industry outperformance occurs and verify whether the sectors outperform

more often in months when they should outperform based on popular belief. But we also find

no indication that this is the case.

4. Sector rotation performance

Can a strategy of sector rotation still be profitable? We now focus on the rotation part of

the strategy: the joint performance across all stages of the business cycle. In our base case

scenario we assume that investors perfectly time the business cycle based on the NBER cycle

and rotate sectors according to conventional wisdom. We assume equally weighted portfolios

over the industries in each stage. We compare these results with a simple buy-and-hold

strategy in Table V.

Please insert Table V around here.

17
The annualized outperformance of sector rotation in a perfect world amounts to 2.3%.18

This outperformance may sound large enough to be of interest to investors, but it represents a

best-case scenario only. Only an investor who had followed popular market wisdom in the last

sixty years, who had ignored transactions costs, and who had perfectly timed the business

cycles in accordance with the NBER – although the NBER did it ex post – would have

realized this 2.3% outperformance. To put this number in perspective: an investor who had –

based on the same information - just timed the early recession right and had held cash during

that period and the market during the remaining time would already have performed better

with 2.5% outperformance. This same investor would also have had a better-diversified

portfolio over time with less industry specific risk. We now consider what happens under

assumptions that are more realistic when we include transactions costs.

Transaction costs, both explicit and implicit, are difficult to estimate with any precision

and depend on the stock, where it trades and when it trades. We use a range of roundtrip

transaction costs between 0.5% and 1.5% as estimates vary considerably and given changes in

costs over the sample span.19 Estimated transaction costs include commissions, bid-ask

spread, and market impact. Sector rotation has 50 roundtrip transactions and the market timing

strategy we use for comparison has 20 roundtrip transactions over the full sample period.

Once we include transactions costs, outperformance for the sector rotation strategy diminishes

substantially and now ranges between 1.1 to 1.9% and becomes significantly indistinguishable

from zero. The alternative strategy based on market timing would be superior as it has fewer

transactions raising relative outperformance further.

18
Since we estimate Jensen’s outperformance with a constant beta over the full sample period, outperformance
does not equal the weighted average of industry outperformance by business cycle stage reported in Table IV.
19
Estimates of total trading costs vary greatly depending on the study. For instance, Lesmond, Schill and Zhou
(2004) estimate roundtrip transaction costs of between 1–2 % for most large-cap trades while Keim and
Madhavan (1998) estimate total round-trip transaction costs as low as 0.2%.

18
Our results so far indicate that the outperformance of sector rotation in line with popular

wisdom is marginal at best even if we give investors the benefit of the doubt and assume that

investors can correctly time the business cycle. Results for industries expected to perform well

in the early expansion stage and the middle expansion stage are particularly disappointing.

Still, it would be premature to conclude that sector rotation does not work. Investor may use

different industry or sector classifications, different business cycle indicators or different

business cycle stages. Alternatively, investors may anticipate business cycles, which could

generate outperformance. On the other hand, our results may be sample specific.

In the robustness tests, we consider all these possible explanations and several others. We

use a range of alternative tests, data sets, performance measures, samples and approaches. We

verify whether results improve if we assume investors would anticipate changes in turning

points earlier. In addition to NBER business cycles, we try different business cycle proxies

and business cycle stages constructed from the CFNAI. We separate our sample in two and

business cycle stages in two. We vary sectors and industries and use sector and industry

returns from alternative data sources such as Fidelity Select funds, Fama and French sectors,

and Standard & Poor’s sector indices. We try obvious explanations for our results first and

then try to relax more and more assumptions.

5. Robustness checks

5.1. Other data sets

The Fama and French industries may not be representative for sectors and industries used

by sector rotation investors. We use three additional data sets: Standard & Poor’s sector

indices, Fama and French sector portfolios, and Fidelity Sector Select Funds.

19
The Standard & Poor’s indices provide a benchmark of sector performance frequently used

by financial practitioners. There are a number of Standard & Poor’s sector indices. We use the

15 Standard & Poor’s indices that are available from Global Financial Data for the entire

1948-2007 period.

The Fama and French sectors comprise all NYSE, AMEX, and NASDAQ traded stocks

mapped to one of seventeen sector portfolios by SIC classifications. It is less likely with the

sector mapping that one or two large firms dominate portfolio returns.20 We obtain the Fama

and French sector data from the Kenneth French website.

Although only available for a shorter period, the performance of Fidelity Sector Select

funds is perhaps the best proxy for actual sector rotation strategies of investors. We source the

Fidelity Sector Select data for 42 funds from Morningstar services. The earliest start date is

August 1981 for the Energy, Health Care, and Technology funds while the most recent start

date is July 2001 for Pharmaceuticals. Due to the shorter Fidelity data series, and to make use

of all available data, we extend the sample period from December 2007 to August 2008 for

the Fidelity fund data. Even so, there are a very limited number of observations for the more

recently added Fidelity Sector Select funds. Total return observations during recessions are

further limited by the infrequent recessions occurring since 1981.

Please insert Table VI around here.

20
For instance, the Fama and French agriculture industry portfolio includes as few as four firms. Consequently,
one can argue that results shown for the agriculture industry might merely measure firm specific developments
unrelated to the business cycle.

20
Table VI provides a comparison of average statistics and performance measures using the

different data sets for the industries that should outperform based on popular wisdom during

the different stages. Here again, there appears no consistency in sector/industry performance

regardless of the data set chosen. The Fama and French sectors slightly outperform in early

expansion and early recession. Consistent with earlier results, late expansion and late

recession industries perform best and early expansion industries largely underperform

although, results are mostly insignificant. Only during late recession do the Fidelity sector

funds outperform the market based on all performance measures although not significantly so.

However, this might be due to the lack of observation. Generally speaking, alternative data

sets do not seem to offer any performance improvement.

5.2 Different ways to measure business cycle

We use the CFNAI and Conference Board Leading Indicator (CBLI) as alternatives to

NBER cycle dates. As results for these two indicators are only marginally different, we focus

on results for the CNFAI. The CFNAI comprises 85 economic and financial variables from

five broad categories: output (21), employment (24), consumption (13), manufacturing (11),

and inventories (16). CFNAI construction follows the methodology of Stock and Watson

(1989) that uses first principal components of a large number of economic variables known to

track economic activity. By construction, the CFNAI has a zero mean and unit standard

deviation where a positive (negative) index value indicates above (below) trend economic

activity. Publication of the CFNAI began in 2001 with series data available from 1967.21 In

Figure III, we overlay the CFNAI index on NBER delineated phases of economic expansion

and contraction (shaded area). We can see that the CFNAI more or less tracks NBER cycle

dates but shows a variation, which may better reflect the uncertainty investors face when they

21
More information is available at http://www.chicagofed.org/economic_research_and_data/cfnai.cfm

21
try to call the business cycle stages in real time. Based on the CFNAI data we try three

approaches. We choose levels at 0.57, 0.26, -.01, and -.045 as to divide the business cycle in

five equal stages and test for outperformance of the conventional wisdom industries using

dummy variables and regressions as before. As an alternative test we partition CFNAI values

according to the Chicago Federal reserve website that defines values above 0.20 as late

expansion and values below -0.70 as recession and subdivided further into stages based on

values we deemed representative: early recession as the range from 0.0 to -0.70 and early and

middle expansion split the 0.0 to 0.20 range.22 Thirdly, eyeballing Figure III it seems that on

average CNFAI index levels larger than zero and positive changes characterize an early

expansion, late expansion has positive index levels but mostly negative changes. Early

recession has a negative level and negative changes and in late recession levels are still low

but changes are positive. Our last test uses this characterization. We run a regression for each

sector where we test for sector outperformance based on a level and a change in the CNFAI

index. We assign the middle expansion sectors to the early or the late expansion depending on

where they perform best. Again this gives conventional wisdom the benefit of the doubt.

ri ,t − rf t = α 0 + α1CFNAI t + α 2 ΔCFNAI t + β i (rmkt ,t − rf t ) + ε t (6)

Please insert Figure III and Table VII around here.

Table VII Panel A reports a summary of average statistics and performance measures for

industries grouped according to where they should outperform based on conventional wisdom

22
We omit results based on Chicago Federal Reserve cut-off points, as they are material similar to those from
equal CFNAI partitions and provide them upon request.

22
now using the two different CFNAI partitions.23 Industry outperformance is even lower over

CFNAI business cycle stages than previously observed with NBER delineated stages. Only

for late expansion industries are mean returns larger than their overall sample mean. Risk

adjusted performance is no better. Average industry Sharpe ratios are lower than the market

for all but during late recession. All three alpha performance measures are mostly negative

and none statistically different from zero. There appears no improvement on our base case

scenario if we use CFNAI rather than NBER measured business cycles. Panel B contains the

results where we estimate the sensitivity of sectors to the level and change of the CNFAI

variable. We report bootstrapped p-values for the likelihood that level and change coefficients

jointly have the correct sign. At the 10% level we only find four sectors that perform well

when they should all in late recession (Recreation, Printing & Publishing, Apparel and

Textiles). For all the other sectors there is no significant outperformance.24 This table suggests

that only in late recession do some sectors perform significantly better than others. Also

Transportation, Electrical Equipment and Business Services seem to recover faster than other

sectors in late recession.

5.4. Timing the business cycle in advance or with a delay

Investors might profit from consistently timing the business cycle incorrectly. Suppose that

investors consistently assume turning points occur earlier than the NBER dates or with a

delay. If so, our base case scenario might underestimate actual outperformance of sector

rotation. We advance the implementation of sector rotation by one month, two months, and

three months prior to NBER business cycle turning points. Similarly we consider delays up to

three months. Table VIII contains our results excluding transactions costs.

23
For brevity, we report industry averages by business cycle stage and provide complete results upon request.
24
For Steel Works (Middle Expansion) the joint probability that significant outperformance occurs in expansion
is statistically significant.

23
Please insert Table VIII around here.

Overall performance declines monotonically and becomes insignificant as we rotate sectors

further in advance of NBER business cycle stage turning points. The sector rotation Jensen’s

alpha of 2.3% decreases to 1.9% and then 1.0% when we rotate sectors one, two, and three

months early. Similarly results decrease if we assume investors respond with a delay. These

results suggest that timing peaks and troughs precisely at business cycle turning points is

important.

5.5. Different business cycle proxies

The literature shows that several economic variables, like term-spread and default-spread,

and dividend yield, proxy business cycles. If investors use these variables to predict the

business cycle and rotate sectors accordingly we can test more directly whether a model that

predicts relative industry or sector outperformance, based on these proxies aligns with the

stages in which conventional wisdom suggest they should outperform.

We create a forecast model where include the one-month Treasury bill, term-spread,25

default-spread,26 and dividend yield as business cycle variables (BCV). Chordia and

Shivakumar (2002) among others show that these variables lagged one period are a good

predictor of momentum profits related to business cycles. Our forecast model uses monthly

25
We calculate the term-spread as the difference between the 10-year Treasury constant maturity yield and the
three-month Treasury yield. Fama and French (1989) find the term-spread closely tracks short-term business
cycles and measures the difference between long-term growth and current short-term business conditions. The
term structure is smallest (largest) at NBER defined business cycle peaks (troughs).
26
We calculate the default-spread as the difference between low-grade Baa and high-grade Aaa corporate bonds.
The default-spread measures a default premium. Expected returns are greater for risky investments during times
of economic uncertainty. As such, the default-spread should increase during periods of recession as investor
required rates of return also increase.

24
changes in the business cycle variables as a proxy for unexpected shocks as the literature

shows that shocks provide the best forecast of asset prices.27 We forecast industry

outperformance related to the business cycle with parameter estimates obtained from a

regression of excess industry returns (ri-rf) on a constant, lagged changes in the business cycle

variables (ΔBCV), and excess market returns (rmkt-rf) using equation 7.

4
ri ,t − rft = c0 + ∑ γ i ΔBCVi ,t −1 + β (rmkt ,t − rft ) + ε t (7)
i =1

Our model is the single index model with the inclusion of lagged changes in the business

cycle variables to capture the relation between business cycle determinants and industry

performance. We use parameter estimates obtained from changes in the business cycle

variables to forecast a period ahead Jensen’s alpha where we decompose Jensen’s alpha to

allow for the contribution of business cycle determinants to industry outperformance. Similar

to Chordia and Shivakumar (2002), we use a sixty-month rolling window to estimate the (γi)

forecast parameters. The rolling window moves forward each month to obtain γi estimates

from the most recent sixty-month window. We then use the parameter estimates to forecast

industry outperformance for the following month (αˆ J ,t +1 ) measured with equation 8 as the sum

of the gamma estimates times changes in current business cycle variable values from the

4
proceeding period (∑ λˆi ΔBCVt ) . We include all industries with forecast positive
i =1

outperformance in the period-ahead sector rotation portfolio for a one-month holding period.

The following month we repeat the same process once again and continue this repetition over

the entire sample period.

27
See for example studies by Chen, Roll and Ross (1986) and Keim and Stambaugh (1986) among others.

25
4
αˆ J ,t +1 = ∑ λˆi ΔBCVt (8)
i =1

To clarify with an example, in month 61 we first estimate the γi parameters with month 1 to

month 60 data. We next multiply the γI parameters estimates times ΔBCVi,61 measured as the

(BCVi,61-BCVi,60) difference to obtain a Jensen’s alpha attributable to business cycle

4
determinants. Lastly, we select all industries where ∑ λˆ ΔBCV > 0
i =1
i t for inclusion in a sector

rotation portfolio for a one-month holding period. The following month we move the rolling

window forward one month and repeat the entire process.

Please insert Table IX around here.

Table IX Panel A and Panel B overlays results from the forecast model on sector

performance with NBER delineated business cycle stages. We want to observe if forecast

industry outperformance coincides with the popular belief of sector rotation investors on

industry performance.28 Panel A reports the average number of industries the forecast model

selects for inclusion in the sector rotation portfolio during each business cycle stage. On

average the forecast model selects approximately half of all industries for inclusion in the

sector rotation portfolio during any given business cycle stage.

28
We also overlay the forecast model results on NBER business cycle sub-stages where we divide each stage
into an early and late stage and additionally sub-periods where we divide the sample. There is no change in our
basic results for both sub-stage and sub-period.

26
Panel B reports the percentage of time a particular industry is included in the sector

rotation portfolio for the full period and for each business cycle stage. We would expect if the

business cycle variables were able to forecast industry outperformance related to the business

cycle, and if industry performance aligns with popular belief, that the model would select an

industry for inclusion in the portfolio during the period of expected optimal performance a

high percentage of the time. However, the forecast model selects industries for inclusion

evenly across the different business cycle and independent of business cycle stages. (Values in

bold indicate percentages that are significantly different from 50 percent at the 10 percent

level.) Also using this method there appears no evidence that sectors perform well when

conventional wisdom suggests they should.

5.6. Description of other robustness tests29

5.6.1. Business cycle proxies: extended analysis

We not only use relative sector outperformance forecasts based on business cycle proxies

but also try to see whether there are alternative links between these economic variables and

sector performance. If conventional belief claims a sector should outperform during part of an

expansion and we know that an economic variable is relatively high during expansion than we

would expect to find a strong and positive link between outperformance of that sector when

that economic variable is at a high level. We test this relation using the most common

business cycle proxies: term-spread, default-spread, dividend yield, unemployment, and

industrial production.30

29
All tables related to these results are available on request from the authors.
30
See for instance, Campbell (1987), Chen (1991), Chen, Roll and Ross (1986), Fama and French (1989),
Jensen, Mercer and Johnson (1996), Keim and Stambaugh (1986), Lewellen (2004), and Petkova (2006).

27
We first establish how these economic variables behave across the business cycle. For

instance, term-spread, default-spread, and dividend yield should be smallest near economic

peaks and largest near economic troughs (Fama and French, 1989). Stock and Watson (1998)

and Hamilton and Lin (1996) show how industrial production growth peaks and

unemployment rates bottom out around business cycle peaks. Boyd, Hu and Jagannathan

(2005) also look at the impact on stocks of changes in unemployment across periods of

economic expansion and recession. We find that changes in the business cycle proxies across

successive stages mostly have the proper sign and are significant. For instance, changes in

unemployment rates from one business cycle stage to the next are all significantly negative

across stages of economic expansion and significantly positive across stages of economic

contraction. Similarly, changes in default-spread are negative during early and middle

expansion and positive during early and late recession. Results tend to be less strong and

insignificant for dividend yields.

Next, we investigate the connection between industry outperformance and these same

variables over the business cycle. We test for a link between outperformance and the business

cycle proxies using equation 9.

ri , p ,t − rf p ,t = α 0 + β BCPj , p BCPj , p ,t + β mkt , p (rmkt , p ,t − rf p ,t ) + ε t (9)

We regress excess returns (ri-rf) for industry i during business cycle phase p (where phase

is expansion or recession) at time t on a constant, business cycle proxy j (BCPj), and a

correction for excess market returns (rmkt-rf). The estimate βBCP times the proxy value captures

the contribution of the business cycle proxy to overall industry outperformance. To make our

results independent of stages we use full NBER expansion and recession periods rather than

28
stages. For instance, term-spread becomes smaller across expansions and larger across

recessions. Therefore, industries that should outperform during periods of expansion

(recession) should have negative (positive) βterm-spread coefficients.

Then we show that while the proxies do track the business cycle a link between industry

outperformance and the business cycle proxies is almost absent. This general result holds

regardless of how we partition the business cycle or whether we look at levels, one-month

lags, or changes in the business cycle proxies.31

5.6.2. CNFAI forecasts of relative sector performance

Similar to our forecast model based on the economic we verify whether using of forecasts

based on the CNFAI indicator would fare better. Using changes in the CNFAI indicator rather

than the different proxies does not change our result in favor of conventional wisdom either.

5.6.3 Sequencing the industries

As the Standard and Poor’s graph in Figure I show outperformance of Technology follows

outperformance of Utilities, and precedes outperformance of the Financial sector. The other

figures suggest similar sequential patterns. We try a number of tests where we ignore the

business cycle completely and verify whether outperformance of one sector predicts future

performance of other sectors at some lag. We try lags up to 24 months for returns and

Jensen’s alphas. We find no results that conventional wisdom sequence holds or in any way

provides a better predictor of outperformance than any other sequence we can think of. We

31
The correspondence between industry outperformance relative to the market and business cycle proxies
measured across business cycle stages is materially similar, if not somewhat weaker, than across phases of
expansion and recession. Similarly, the results hold regardless of whether we use level, one-month lags, or
changes in the business cycle variables. For brevity, we limit our discussion to the link between industry
performance across phases of economic expansion and recession using business cycle proxy levels and provide
results of the additional tests upon request.

29
find some one-month lead lag relations between industries. However, beyond one-month lags

significant results occur seemingly random.

5.6.4. Varying stages

We try a number of variations of the stages that could improve our base case scenario.

Outperformance might only occur at the beginning or at the end of stages. We divide all

stages in two and re-run our main tests again. We find no major difference between the first

half and the second half of the returns during the different stages. We consider shorter periods

where we test for significant outperformance for two, four and six months around turning

points only. The idea being that investors might anticipate different stages and react in shorter

around business cycle turning points rather than over the full length of a stage. Again we can

find no significant outperformance.

5.6.5. Sub-samples

Significant events over a full 60-year sample period, like the 1970’s bear market and the

1990’s dot com market could overly influence our results. We compare 1948-1977 and 1978-

2007 sub-periods measures for the average of the industries in each stage with the full sample

period measures. 32 The performance measures appear relatively constant across all periods

and business cycle stages. Consistent with our previous analysis, early expansion and middle

expansion industries provide inferior outperformance across both sub-periods based on all

measures. Overall, our results do not seem sample specific.

32
The complete results for individual industries are available upon request.

30
5.6.6. Alternative Performance Measures

We evaluate two alternative performance measures to compare the sector rotation, market-

timing, and buy-and-hold strategies. But result are similar if we use the Goetzmann, Ingersoll,

Spiegel and Welch (2007) manipulation-proof performance measure (MPPM) or the Barrett

and Donald (2003) simulated tests of second-order stochastic dominance.

6. Alternative sector rotation strategy

So far we find little evidence in favor of sector rotation based on conventional wisdom.

The obvious next question is if there are any sectors that consistently and significantly

outperform the market and if so whether a rotation strategy based on these alternative sectors

might perform well. We try whether we can take our results one step further and test for

consistent and significant outperformance of any sector across the business cycle, not only

those sectors suggested by conventional wisdom. As a first step we considered the

performance measures for all sectors in all stages. Under the null hypothesis of no significant

outperformance, we would expect to find the different alphas almost normally distributed

around zero. In Figure IV we plot the expected distribution under the null and the actual

distribution of Jensen’s alpha t-statistics (all the other measures show similar patterns). At

first sight both plots seem similar. However, we do find slightly more significant

outperforming sectors than we would expect under the null (20 versus 12). This number might

be close enough to the null for some. Others might argue it is almost double the number of

outperforming sectors one would expect under the null. As we would prefer to err on the side

of caution we take a closer look whether we can find a group of sectors not based on

conventional wisdom but that would otherwise survive all our tests. We do find sectors with

jointly significant Jensen’s, Fama and French, and Carhart alphas during particular stages of

the business cycle. There are no sectors in early expansion; Candy & Soda, Pharmaceutical in

31
middle expansion; Mining, Tobacco Products in late expansion; Shipping Containers, Food

products, Utilities, Entertainment in early recession; and Personal services, Food products and

once more Tobacco products in late recession.

Please insert Figure IV and Table X around here.

Historically a rotation strategy based on these sectors (which holds the general market in

early expansion) would have done well; an average market outperformance of 7.3% a year

(6.1% assuming 1.5% round trip transactions costs as in Table V). These alternative sectors

perform well in the months of the stages where they were supposed to perform well about 60-

70% of the time. If implemented in advance 1, 2, or 3 months strategy returns would have

reduced to 6.9%, 6.1% and 4.9% and if implemented with a delay of 1, 2, or 3 months would

have given returns of 7.2%, 6.5%, and 5.6%. All these sectors also outperformed in both sub-

periods although not always significantly so. One could argue that this means there was no

outperformance. Alternatively, one might attribute this result to a lack of observations. Similar

sectors and industries in other datasets also show outperformance but again not in all cases

significantly so.

7. Conclusion

No matter how hard we try, we find little support for conventional wisdom that sector

rotation across business cycles outperforms the general market. Even if we give sector

rotation the benefit of the doubt and assume that investors can perfectly time business cycles

returns are marginally higher than the market. Our evidence even allows the conclusion that

this result holds more generally and that no sector performs consistently and significantly

better in different business cycle stages. We find one possible exception. A strategy that holds

32
the market in early expansion, Candy & Soda, Pharmaceutical in middle expansion; Mining,

Tobacco Products in late expansion; Shipping Containers, Food products, Utilities,

Entertainment in early recession and Personal services, Food products and once more Tobacco

products in late recession, would have beaten the market in our sample with around 7%

annually. This strategy also seems to survive our robustness checks although marginally so.

Whether the outperformance of this strategy is a result of data mining or a result of underlying

fundamental reason we cannot tell. However, based on our results it seems a more promising

sector rotation strategy and safer bet than the traditional rotation strategy based on popular

wisdom if investors feel that the business cycle contains information about sector

performance.

33
REFERENCES

Avramov, D., and Chordia, T., 2006, Asset pricing models and financial market anomalies,
Review of Economic Studies 19, 1001-1040.
Avramov, D., and Wermers, R., 2006, Investing in mutual funds when returns are predictable,
Journal of Financial Economics 81, 339-377.
Barrett, G., and Donald, S., 2003, Consistent tests for stochastic dominance, Econometrica 71,
71-104.
Beber, A., Brandt, M., and Kavajecz, K., 2009, What does equity sector orderflow tell us
about the economy?, Working Paper Series (SSRN).
Bodie, Z., Kane, A., and Marcus, A., 2009. Investments (McGraw Hill Irwin).
Boyd, J., Hu, J., and Jagannathan, R., 2005, The stock market's reaction to unemployment
news: Why bad news is usually good for stocks, Journal of Finance 60, 649-672.
Business Week Online, 2002, Get in, get out, and move on, (McGraw-Hill Companies, Inc.).
Campbell, J., 1987, Stock returns and the term structure, Journal of Financial Economics 18,
373-399.
Carhart, M., 1997, On persistence in mutual fund performance, Journal of Finance 52, 57-82.
Chatterjee, S., 1999, Real business cycles: A legacy of countercyclical policies?, Business
Review 17-27.
Chauvet, M., and Hamilton, J., 2005, Dating business cycle turning points, NBER Working
Paper.
Chen, N., 1991, Financial investment opportunities and the macroeconomy, Journal of
Finance 46, 529-554.
Chen, N., Roll, R., and Ross, S. A., 1986, Economic forces and the sock market, The Journal
of Business 59, 383-403.
Chordia, T., and Shivakumar, L., 2002, Momentum, business cycle, and time-varying
expected returns, Journal of Finance 57, 985-1019.
CNN Money, 2006, Stocks that can ride a pause, (Cable News Network. A Time Warner
Company).
Conover, M., Jensen, G., Johnson, R., and Mercer, J., 2008, Sector rotation and monetary
conditions, Journal of Investing 34-46.
Cover, J., and Pecorino, P., 2005, The length of US business expansions: When did the break
in the data occur?, Journal of Macroeconomics 27, 452-471.
DeStefano, M., 2004, Stock returns and the business cycle, Financial Review 39, 527-547.
Eleswarapu, V., and Tiwari, A., 1996, Business cycles and stock market returns: Evidence
using industry-based portfolios, Journal of Financial Research 19, 121-134.
Elton, E., Gruber, M., and Blake, C., 2009, An examination of mutual fund timing ability
using monthly holding data, Working Paper Series (SSRN).
Fabozzi, F., 2007. Fixed income analysis (Wiley).
Fama, E., 1975, Short-term interest rates as predictors of inflation, American Economic
Review 65, 269-282.
Fama, E., and French, K., 1989, Business conditions and expected returns on stocks and
bonds, Journal of Financial Economics 25, 23-49.
Fama, E., and French, K., 1992, The cross-section of expected stock returns, Journal of
finance 47, 427-465.
Goetzmann, W., Ingersoll, J., Spiegel, M., and Welch, I., 2007, Portfolio performance
manipulation and manipulation-proof performance measures, Review of Financial
Studies 20, 1503-1546.
Hamilton, J., and Lin, G., 1996, Stock market volatility and the business cycle, Journal of
Applied Econometrics 11, 573-593.

34
Hong, H., Torous, W., and Valkanov, R., 2007, Do industries lead stock markets?, Journal of
Financial Economics 83, 367-396.
Hou, K., 2007, Industry Information diffusion and the lead-lag effect in stock returns, Review
of Financial Studies 20, 1113-1138.
Investment Company Institute, 2001. Investment Company fact book (Investment Company
Institute).
Investment Company Institute, 2009. Investment Company fact book (Investment Company
Institute).
Jacobsen, B., and Visaltanachoti, N., 2009, The Halloween effect in US sectors, Financial
Review 44, 437-459.
Jensen, G., Mercer, J., and Johnson, R., 1996, Business conditions, monetary policy, and
expected security returns, Journal of Financial Economics 40, 213-237.
Jiang, G., Yao, T., and Yu, T., 2007, Do mutual funds time the market? Evidence from
portfolio holdings, Journal of Financial Economics 86, 724-758.
Keim, D., and Madhavan, A., 1998, The cost of institutional equity trades, Financial Analysts
Journal 54, 50-69.
Keim, D., and Stambaugh, R., 1986, Predicting returns in the stock and bond markets, Journal
of Financial Economics 17, 357-390.
Ledoit, O., and Wolf, M., 2008, Robust performance hypothesis testing with the Sharpe ratio,
Journal of Empirical Finance 15, 850-859.
Lesmond, D., Schill, M., and Zhou, C., 2004, The illusory nature of momentum profits,
Journal of Financial Economics 71, 349-380.
Lewellen, J., 2004, Predicting returns with financial ratios, Journal of Financial Economics
74, 209-235.
Lofthouse, S., 2001. Investment management (Wiley).
Markese, J., 1986, The stock market and business cycles, AAII Journal 8, 30-32.
Menzly, L., and Ozbas, O., 2004, Cross-industry momentum, Working Paper Series (SSRN).
O'neal, E., 2000, Industry momentum and sector mutual funds, Financial Analysts Journal 56,
37-49.
Petkova, R., 2006, Do the Fama-French factors proxy for innovations in predictive variables,
The Journal of Finance 61, 581-612.
Reuters, 2008, Goldman Sachs sees recession in 2008, (Thomson Reuters ).
Salsman, R., 1997, Using market prices to guide sector rotation, (CFA Institute).
Stock, J., and Watson, M., 1989, New indexes of coincident and leading economic indicators,
NBER macroeconomics annual 351-394.
Stock, J., and Watson, M., 1998, Business cycle fluctuations in US macroeconomic time
series, NBER working paper.
Stock, J., and Watson, M., 2002, Has the business cycle changed and why?, NBER
macroeconomics annual 159-218.
Stovall, S., 1996. Standard & Poor's guide to sector investing (McGraw-Hill).
White, H., 1980, A heteroskedasticity-consistent covariance matrix estimator and a direct test
for heteroskedasticity, Econometrica 48, 817-838.

35
Table I. NBER reference business cycle dates and stage partitions

Notes: Panel A shows NBER published business cycle peak and trough reference dates from January
1948 to December 2007. We count periods of recession as the first month following a cycle peak to the
subsequent trough and periods of expansion as the first month following a cycle trough to subsequent
peak. The last column shows the total months in a business cycle from peak to peak. The last recorded
NBER business cycle date is December 2007. Panel B shows the total duration in months for
recessions and expansions based on the NBER turning points shown in Panel A. We partition NBER
defined periods of expansion into three equal stages (early, middle, and late) and NBER defined
periods of recessions into two equal stages (early and late). The bottom of Panel B shows the average
duration of each stage.

Panel A: NBER business-cycle dates (Jan 1948 - Dec 2007)


Cycle
Peak Trough Peak Length
11/48 10/49 07/53 56
07/53 05/54 08/57 49
08/57 04/58 04/60 32
04/60 02/61 12/69 116
12/69 11/70 11/73 47
11/73 03/75 01/80 74
01/80 07/80 07/81 18
07/81 11/82 07/90 108
07/90 03/91 03/01 128
03/01 11/01 12/07 81

Panel B: Number of months in NBER delineated business cycle stages


Period of Expansion Period of Recession
Early Stage Middle Stage Late Stage Total Early Stage Late Stage Total
15 15 15 45 6 5 11
13 13 13 39 5 5 10
8 8 8 24 4 4 8
35 35 36 106 5 5 10
12 12 12 36 6 5 11
19 19 20 58 8 8 16
4 4 4 12 3 3 6
30 31 31 92 8 8 16
40 40 40 120 4 4 8
25 25 23 73 4 4 8
average: 20 20 20 60 5 5 10

36
Table II. List of expected best performing industries across business cycle stages

Notes: Table reports stages of business cycle where, based on the Stovall (1996) classification and
popular investment websites such as those shown in Figure I, sectors provide the best performance.
We partition periods of expansion into three equal stages (early, middle, and late) and periods of
recession into two equal stages (early and late).We then map each of the 48 Fama and French
industries to its appropriate sector.

Period of Expansion Period of Recession


Early Expansion - Stage I Middle Expansion - Stage II Late Expansion - Stage III Early Recession - Stage IV Late Recession - Stage V
Technology: Basic Materials: Consumer Staples: Utilities: Consumer Cyclical:
Computer Software Precious Metals Agriculture Gas & Electrical Utilities Apparel
Measuring & Control Equip. Chemicals Beer & Liquor Telecom Automobiles & Trucks
Computers Steel Works Etc Candy & Soda Business Supplies
Electronic Equipment Non-Metallic & Metal Minin Food Products Construction
Transportation: Capital Goods: Healthcare Construction Materials
General Transportation Fabricated Products Medical Equipment Consumer Goods
Shipping Containers Defense Pharmaceutical Products Entertainment
Machinery Tobacco Products Printing & Publishing
Ships & Railroad Equip. Energy: Recreation
Aircraft Coal Restaraunts, Hotels, Motels
Electrical Equipment Petroleum & Natural Gas Retail
Services: Rubber & Plastic Products
Business Services Textiles
Personal Services Wholesale
Financial:
Banking
Insurance
Real Estate
Trading

37
Table III. Descriptive industry statistics by NBER delineated business cycle stages

Notes: Table reports nominal industry returns and standard deviations for the business cycle stage
considered optimal by conventional wisdom as annualized rates. We estimate nominal industry returns
for each business cycle stage with equation 1 where we regress industry returns on business cycle
dummy variables (Ds) that take a value of one or zero depending on the business cycle stage. The beta
estimate is from a standard single index model and rho (1) is the first order serial correlation
coefficient with statistical significance at 10% highlighted. We also report Wald test results for
differences in industry returns across the five business cycle stages and report p-values under a null
hypothesis of equal industry returns across the business cycle. For comparative purposes, we provide
annualized industry returns for the full sample period in the far right column and also industry
averages and market results beneath each business cycle stage. Column 2 also reports the number of
industry return observations (obs.) included in a business cycle stage.

5
ri ,t = ∑ μi ,s Ds ,t + ε t (1)
s =1

Sample Period 1948:01 - 2007:12


stage sample
Sector/Industry obs. mean std. dev. beta rho(1) p-value mean
Early Expansion - Stage I:
Computers 201 0.13 0.22 1.38 -0.02 0.00 0.13
Computer Software 130 0.00 0.34 1.71 0.05 0.15 0.02
Electronic Equipment 201 0.17 0.25 1.50 0.04 0.00 0.11
Measuring & Control 201 0.10 0.22 1.36 0.09 0.00 0.12
Shipping Containers 201 0.18 0.17 0.96 -0.01 0.00 0.12
Transportation 201 0.25 0.17 1.02 0.05 0.00 0.10
Industry Averages 0.14 0.23 1.32 0.03 0.03 0.10
Market 0.17 0.13 1.00 0.03 0.00 0.12

Middle Expansion - Stage II:


Chemicals 202 0.12 0.17 1.12 0.02 0.00 0.11
Steel Works 202 0.13 0.22 1.23 -0.03 0.00 0.10
Precious Metals 166 0.08 0.32 0.76 -0.05 0.58 0.08
Mining 202 0.12 0.23 1.19 -0.05 0.00 0.12
Fabricated Products 166 0.12 0.20 1.00 -0.03 0.01 0.05
Machinery 202 0.17 0.18 1.22 0.03 0.00 0.11
Electrical Equipment 202 0.19 0.19 1.26 -0.03 0.00 0.14
Aircraft 202 0.19 0.21 1.15 0.10 0.00 0.14
Shipbuilding & Railroad 202 0.08 0.19 1.15 0.00 0.01 0.10
Defense 166 0.15 0.21 1.07 -0.01 0.02 0.12
Personal Services 202 0.12 0.22 1.17 0.10 0.00 0.09
Business Services 202 0.13 0.16 1.04 0.11 0.00 0.10
Industry Averages 0.13 0.21 1.11 0.01 0.05 0.10
Market 0.14 0.13 1.00 0.03 0.00 0.12

38
Table III. Continued

Sample Period 1948:01 - 2007:12


stage sample
Sector/Industry obs. mean std. dev. beta rho(1) p-value mean
Late Expansion - Stage III:
Agriculture 213 0.11 0.22 0.81 -0.06 0.00 0.10
Food Products 213 0.07 0.15 0.61 0.03 0.00 0.13
Candy & Soda 166 0.05 0.24 0.74 0.01 0.01 0.12
Beer & Liquor 213 0.10 0.19 0.81 0.00 0.00 0.13
Tobacco Products 213 0.15 0.20 0.40 0.11 0.04 0.15
Healthcare 136 0.09 0.33 1.16 0.09 0.02 0.09
Medical Equipment 213 0.12 0.17 0.86 -0.02 0.01 0.14
Pharmaceutical 213 0.10 0.16 0.70 -0.05 0.01 0.13
Coal 213 0.21 0.33 1.02 0.08 0.00 0.14
Petroleum & Natural 213 0.11 0.18 0.74 -0.04 0.00 0.14
Industry Averages 0.11 0.22 0.79 0.01 0.01 0.13
Market 0.07 0.15 1.00 -0.02 0.00 0.12

Early Recession - Stage IV:


Utilities 53 0.00 0.16 0.76 0.11 0.05 0.11
Communication 53 -0.04 0.14 0.63 0.06 0.07 0.10
Industry Averages -0.02 0.15 0.70 0.09 0.06 0.11
Market -0.16 0.16 1.00 -0.05 0.00 0.12

Late Recession - Stage V:


Recreation 51 0.64 0.31 1.22 -0.13 0.00 0.09
Entertainment 51 0.50 0.31 1.28 0.19 0.10 0.14
Printing & Publishing 51 0.62 0.23 1.03 0.29 0.00 0.12
Consumer Goods 51 0.49 0.21 1.01 0.12 0.00 0.12
Apparel 51 0.63 0.27 1.09 0.17 0.00 0.10
Rubber & Plastic 51 0.42 0.22 0.90 0.07 0.00 0.12
Textiles 51 0.47 0.25 1.09 0.05 0.00 0.10
Construction Material 51 0.51 0.23 1.17 0.03 0.00 0.11
Construction 51 0.63 0.33 1.51 -0.01 0.00 0.12
Automobiles & Truck 51 0.38 0.25 1.06 0.20 0.00 0.11
Business Supplies 51 0.44 0.24 1.19 -0.07 0.00 0.11
Wholesale 51 0.43 0.23 1.06 0.13 0.00 0.11
Retail 51 0.55 0.24 1.11 0.27 0.00 0.12
Restaurants & Hotels 51 0.52 0.28 1.27 0.04 0.00 0.12
Banking 51 0.48 0.23 1.14 0.11 0.02 0.12
Insurance 51 0.44 0.20 0.86 0.14 0.01 0.12
Real Estate 51 0.56 0.31 1.21 0.06 0.00 0.07
Trading 51 0.53 0.23 1.16 0.12 0.00 0.14
Industry Averages 0.51 0.25 1.13 0.10 0.01 0.11
Market 0.40 0.18 1.00 0.11 0.00 0.12

39
Table IV. Industry performance measures by NBER delineated business cycle stages

Notes: Table reports differences between industry and market Sharpe ratios, excess market returns
(αmkt), Jensen’s alphas (αJ), Fama and French (1992) three-factor alphas (αF), and Carhart (1997) four-
factor alphas (αC) for the business cycle stage considered optimal by conventional wisdom. We report
annualized alpha returns with White (1980) heteroskedasticity consistent t-statistics highlighted for
statistical significance at 10%. To calculate Sharpe ratios, we divide returns in excess of the one-month
Treasury bill by the standard deviation of returns. We estimate Ledoit and Wolf (2008) p-values for
industry and market Sharpe ratio differences corrected for potentially non-iid returns and indicate
statistically significant differences at the 10% level or higher in bold. We estimate excess market
returns, Jensen’s alphas, Fama and French alphas, and Carhart alphas by business cycle stage with
equations 2-5 respectively using business cycle stage dummy variables (Ds) previously described. We
also report Wald test results for differences in performance measures across the five business cycle
stages and report p-values under a null hypothesis of constant industry performance. Table also reports
industry averages beneath each business cycle stage.

5
ri ,t − rmkt ,t = ∑α mkt ,i ,s Ds ,t + ε t (2)
s =1
5 5
ri ,t − rft = ∑α J ,i ,s Ds ,t + ∑ β1,i ,s (rmkt ,t − rft ) Ds,t + ε t (3)
s =1 s =1
5 5
ri ,t − rft = ∑α F ,i ,s Ds ,t + ∑ ⎡⎣ β1,i ,s (rmkt ,t − rft ) + β2,i ,s SMBt + β3,i ,s HMLt ⎤⎦ Ds ,t + ε t (4)
s =1 s =1

5 5
ri ,t − rft = ∑αC ,i ,s Ds ,t + ∑ ⎡⎣ β1,i ,s (rmkt ,t − rft ) + β2,i ,s SMBt + β3,i ,s HMLt + β4,i ,s MOM t ⎤⎦ Ds ,t + ε t (5)
s =1 s =1

Excess Market Jensens alpha Fama-French alpha Carhart alpha


Sharpe Ratio Wald Wald Wald Wald
Sectors/Industries Difference α mkt p-value αJ p-value αF p-value αC p-value
Early Expansion - Stage I:
Computers -0.16 -0.03 0.48 -0.08 0.32 -0.05 0.92 -0.01 0.90
Computer Software -0.21 -0.12 0.55 -0.17 0.88 -0.18 0.85 -0.19 0.83
Electronic Equip. -0.13 0.00 0.29 -0.06 0.94 -0.04 0.94 -0.01 0.90
Measuring & Control -0.19 -0.06 0.06 -0.10 0.05 -0.07 0.22 -0.06 0.22
Shipping Containers -0.05 0.01 0.10 0.01 0.12 -0.01 0.15 -0.01 0.15
Transportation 0.06 0.07 0.00 0.07 0.00 0.03 0.08 0.02 0.10
average: -0.11 -0.02 -0.05 -0.05 -0.04

40
Table IV. Continued

Excess Market Jensens alpha Fama-French alpha Carhart alpha


Sharpe Ratio Wald Wald Wald Wald
Sectors/Industries Difference α mkt p-value αJ p-value αF p-value αC p-value
Middle Expansion - Stage II:
Chemicals -0.07 -0.02 0.05 -0.03 0.04 -0.03 0.16 -0.02 0.13
Steel Works -0.08 -0.01 0.53 -0.03 0.60 -0.06 0.32 -0.05 0.25
Precious Metals -0.14 -0.05 0.97 -0.03 0.97 -0.07 0.97 -0.07 0.96
Mining -0.10 -0.01 0.05 -0.03 0.05 -0.06 0.01 -0.06 0.01
Fabricated Products -0.07 -0.01 0.16 -0.01 0.24 -0.04 0.16 -0.03 0.16
Machinery -0.01 0.03 0.02 0.01 0.19 0.00 0.26 0.00 0.22
Electrical Equip. 0.01 0.05 0.36 0.02 0.33 0.02 0.39 0.01 0.41
Aircraft 0.00 0.05 0.66 0.03 0.84 0.01 0.96 0.00 0.96
Shipbuilding/Railroad -0.14 -0.05 0.62 -0.07 0.62 -0.08 0.58 -0.08 0.59
Defense -0.04 0.01 0.67 0.01 0.48 -0.02 0.52 -0.03 0.48
Personal Services -0.10 -0.02 0.17 -0.03 0.29 -0.04 0.48 -0.03 0.50
Business Services -0.04 0.00 0.09 -0.01 0.13 -0.01 0.16 -0.01 0.17
average: -0.07 0.00 -0.01 -0.03 -0.03
Late Expansion - Stage III:
Agriculture 0.05 0.04 0.24 0.04 0.20 0.04 0.15 0.04 0.14
Food Products 0.00 0.00 0.00 0.01 0.01 0.00 0.01 0.01 0.01
Candy & Soda -0.06 -0.03 0.32 -0.03 0.20 -0.04 0.18 -0.03 0.18
Beer & Liquor 0.04 0.03 0.37 0.03 0.65 0.03 0.37 0.04 0.37
Tobacco Products 0.11 0.08 0.01 0.09 0.09 0.08 0.05 0.10 0.05
Healthcare 0.00 0.01 0.23 0.01 0.30 0.01 0.24 0.02 0.24
Medical Equipment 0.08 0.05 0.04 0.05 0.05 0.05 0.10 0.03 0.09
Pharmaceutical 0.05 0.03 0.00 0.04 0.01 0.04 0.04 0.01 0.04
Coal 0.10 0.14 0.06 0.14 0.09 0.13 0.07 0.06 0.06
Petroleum & Natural 0.07 0.05 0.50 0.05 0.42 0.04 0.36 0.03 0.35
average: 0.04 0.04 0.04 0.04 0.03
Early Recession - Stage IV:
Gas & Electric 0.33 0.20 0.01 0.13 0.74 0.12 0.49 0.12 0.50
Communication 0.23 0.15 0.02 0.05 0.65 0.04 0.74 0.04 0.64
average: 0.28 0.17 0.09 0.08 0.08
Late Recession - Stage V:
Recreation -0.04 0.18 0.47 0.10 0.66 0.06 0.56 0.00 0.59
Entertainment -0.13 0.07 0.76 -0.01 0.18 -0.05 0.06 -0.04 0.05
Printing & Publishing 0.10 0.16 0.06 0.15 0.07 0.12 0.07 0.08 0.07
Consumer Goods 0.03 0.07 0.37 0.06 0.28 0.07 0.24 0.00 0.22
Apparel 0.02 0.17 0.27 0.14 0.34 0.07 0.27 0.08 0.39
Rubber & Plastic -0.07 0.01 0.45 0.04 0.56 0.01 0.47 -0.03 0.47
Textiles -0.07 0.05 0.34 0.03 0.34 -0.02 0.81 0.01 0.89
Construction Material 0.00 0.08 0.09 0.03 0.33 0.00 0.29 -0.03 0.31
Construction -0.07 0.17 0.00 0.01 0.03 -0.01 0.01 0.03 0.01
Automobiles & Truck -0.14 -0.01 0.07 -0.03 0.08 -0.07 0.39 -0.05 0.34
Business Supplies -0.08 0.03 0.22 -0.03 0.26 -0.04 0.65 -0.04 0.68
Wholesale -0.08 0.02 0.56 0.00 0.75 -0.01 0.72 -0.07 0.72
Retail 0.02 0.11 0.10 0.08 0.10 0.04 0.09 0.01 0.10
Restaraunts & Hotels -0.07 0.09 0.37 0.01 0.57 -0.02 0.60 -0.04 0.66
Banking -0.04 0.06 0.43 0.01 0.48 0.00 0.34 -0.05 0.38
Insurance -0.01 0.03 0.92 0.07 0.92 0.09 0.82 0.04 0.84
Real Estate -0.09 0.11 0.08 0.05 0.14 -0.03 0.10 -0.12 0.11
Trading 0.02 0.10 0.02 0.05 0.60 0.03 0.77 0.05 0.82
average: -0.04 0.08 0.04 0.01 -0.01

41
Table V. Comparison of market, sector rotation, and market timing performance

Notes: Table compares Jensen’s alpha and Sharpe Ratio performance measures for the market, sector
rotation, and market timing after allowing for a range of transaction costs. The market strategy invests
in the market portfolio for the entire period. Sector rotation holds equal weights in sectors/industries
based on conventional wisdom during a particular business cycle stage. Market timing holds the
market portfolio for all business cycle stages except for cash during early recession. We report
Jensen’s alphas as annualized rates with White (1980) heteroskedasticity consistent t-statistics.

Full Period 1948:01 - 2007:12

Strategy Jensen's alpha t-statistic Sharpe ratio


Market - - 0.13

0% round-trip transaction costs


Sector rotation 2.3% 1.94 0.15
Market-timing 2.5% 3.57 0.17

0.5% round-trip transaction costs


Sector rotation 1.9% 1.59 0.14
Market-timing 2.3% 3.32 0.17

1.0% round-trip transaction costs


Sector rotation 1.5% 1.24 0.14
Market-timing 2.1% 3.06 0.16

1.5% round-trip transaction cost


Sector rotation 1.1% 0.89 0.13
Market-timing 1.9% 2.78 0.16
*sector rotation and market-timing have respectively 50 and 20 round-trip transactions

42
Table VI. Average statistics and performance comparison using different data sets by NBER
business cycle stages

Notes: Table reports the average beta, standard deviation, stage mean return, full period mean
return, excess market return (αmkt), Jensen’s alpha (αJ), Fama and French three-factor alpha
(αF), and Carhart four-factor alpha (αC), and the difference between industry and market
Sharpe ratios for the business cycle stage considered optimal by conventional wisdom. We report
annualized standard deviations, means, and alpha performance measures.

Performance Measures
stage period Sharpe Ratio
Industries Period beta std. dev. mean mean α mkt αJ αF α C Difference
Early Expansion Industries - Stage I:
Fama & French 48 Industries 1948:01-2007:12 1.32 0.23 0.14 0.10 -0.02 -0.05 -0.05 -0.04 -0.11
Standard & Poors 15 Sectors 1948:01-2007:12 1.12 0.21 0.15 0.09 0.01 -0.03 -0.03 -0.02 -0.12
Fama & French 16 Sectors 1948:01-2007:12 1.00 0.16 0.24 0.11 0.09 0.06 0.03 0.01 0.06
Fidelity Select 42 Sectors 1981:08-2008:08 1.45 0.26 0.07 0.07 0.05 0.00 -0.02 -0.02 0.02

Middle Expansion Industries - Stage II:


Fama & French 48 Industries 1948:01-2007:12 1.11 0.21 0.13 0.10 0.00 -0.01 -0.03 -0.03 -0.07
Standard & Poors 15 Sectors 1948:01-2007:12 1.06 0.25 0.08 0.08 -0.02 -0.05 -0.07 -0.06 -0.14
Fama & French 16 Sectors 1948:01-2007:12 1.13 0.19 0.13 0.11 0.00 -0.02 -0.03 -0.02 -0.06
Fidelity Select 42 Sectors 1981:08-2008:08 1.10 0.22 0.17 0.11 -0.03 -0.01 0.00 0.00 -0.07

Late Expansion Industries - Stage III:


Fama & French 48 Industries 1948:01-2007:12 0.79 0.22 0.11 0.13 0.04 0.04 0.04 0.03 0.04
Standard & Poors 15 Sectors 1948:01-2007:12 0.60 0.16 0.07 0.09 0.03 0.01 0.00 0.00 0.00
Fama & French 16 Sectors 1948:01-2007:12 0.67 0.16 0.11 0.13 0.03 0.04 0.04 0.03 0.06
Fidelity Select 42 Sectors 1981:08-2008:08 0.73 0.21 0.14 0.11 0.03 0.05 0.05 0.03 0.03

Early Recession Industries - Stage IV:


Fama & French 48 Industries 1948:01-2007:12 0.70 0.15 -0.02 0.11 0.17 0.09 0.08 0.08 0.28
Standard & Poors 15 Sectors 1948:01-2007:12 0.70 0.18 -0.09 0.05 0.16 0.01 -0.01 -0.01 0.18
Fama & French 16 Sectors 1948:01-2007:12 0.76 0.16 0.00 0.11 0.22 0.13 0.12 0.12 0.33
Fidelity Select 42 Sectors 1981:08-2008:08 1.05 0.21 -0.15 0.07 -0.03 -0.05 -0.02 -0.02 0.00

Late Recession Industries - Stage V:


Fama & French 48 Industries 1948:01-2007:12 1.13 0.25 0.51 0.11 0.08 0.04 0.01 -0.01 -0.04
Standard & Poors 15 Sectors 1948:01-2007:12 0.90 0.21 0.30 0.07 -0.03 -0.05 -0.07 -0.08 -0.16
Fama & French 16 Sectors 1948:01-2007:12 1.13 0.24 0.50 0.11 0.09 0.03 0.00 -0.01 -0.03
Fidelity Select 42 Sectors 1981:08-2008:08 1.18 0.26 0.35 0.11 0.14 0.12 0.11 0.09 0.12

43
Table VII. Industry measures based on the CFNAI over the 1968:01-2007:12 period

Notes: Panel A reports average industry statistics and performance measures by CFNAI
delineated business cycle stages. We divide the range of CFNAI values into five equal stages to
construct business cycles stages of 96 observations each. We report the average single index
model beta, standard deviation, stage mean return, full period mean return, excess market
return (αmkt), Jensen’s alpha (αJ), Fama and French three-factor alpha (αF), and Carhart four-
factor alpha (αC), and industry- market Sharpe ratio difference for all industries that based on
conventional wisdom provide optimal performance during a particular business cycle stage.
We report annualized standard deviations, means, and performance measures. Panel B reports
regression coefficients from equation (6) and bootstrapped p-values for the likelihood that
level and change CFNAI coefficients jointly have the correct sign.

ri ,t − rf t = α 0 + α1CFNAI t + α 2 ΔCFNAI t + β i (rmkt ,t − rf t ) + ε t (6)

Panel A:
Performance Measures
stage sample Sharpe ratio
Sector/Industry beta std. dev. mean mean α mkt αJ αF αC Difference

Early Expansion Industries - Stage I: 1.26 0.28 0.06 0.07 -0.02 -0.02 -0.01 0.01 -0.02

Middle Expansion Industries - Stage II: 1.06 0.20 0.04 0.09 -0.01 -0.01 -0.03 -0.03 -0.02

Late Expansion Industries - Stage III: 0.88 0.19 0.17 0.12 0.01 0.02 0.01 0.00 -0.06

Early Recession Industries - Stage IV: 0.89 0.18 0.10 0.11 -0.03 -0.01 -0.03 -0.02 -0.08

Late Recession Industries - Stage V: 1.08 0.27 0.11 0.10 0.02 0.02 0.00 0.01 0.01

Panel B:

Expected Sign of CFNAI regression Coefficient


Industries CFNAI ΔCFNAI Industries CFNAI ΔCFNAI
Early Expansion pos pos Early Recession neg neg
Middle Expansion pos - Late Recession neg pos
Late Expansion pos neg
joint probability of regression coefficient signs
CFNAI ≥ 0 CFNAI ≥ 0 CFNAI < 0 CFNAI < 0
(Mkt-TBL) CFNAI ΔCFNAI and and and and
Stage Industry constant coefficient coefficient coefficient ΔCFNAI ≥ 0 ΔCFNAI < 0 ΔCFNAI < 0 ΔCFNAI ≥ 0
Early Expansion Computers -0.003 1.26 0.004 0.003 0.61 0.32 0.02 0.06
Computer Software -0.009 1.77 -0.007 0.035 0.11 0.00 0.01 0.87
Electronic Equipment -0.003 1.48 -0.001 0.012 0.36 0.01 0.01 0.62
Measuring & Control -0.003 1.39 -0.001 0.010 0.36 0.03 0.02 0.59
Shipping Containers 0.001 0.94 -0.001 0.003 0.21 0.11 0.19 0.49
Transportation -0.001 1.08 -0.003 0.012 0.08 0.00 0.01 0.90

44
Table VII. Continued

joint probability of regression coefficient signs


CFNAI ≥ 0 CFNAI ≥ 0 CFNAI < 0 CFNAI < 0
(Mkt-TBL) CFNAI ΔCFNAI and and and and
Stage Industry constant coefficient coefficient coefficient ΔCFNAI ≥ 0 ΔCFNAI < 0 ΔCFNAI < 0 ΔCFNAI ≥ 0
Middle Expansion Chemicals 0.000 0.97 -0.002 0.010 0.14 0.01 0.01 0.85
Steel Works -0.003 1.23 0.004 0.006 0.77 0.19 0.00 0.05
Precious Metals -0.001 0.71 -0.004 0.014 0.19 0.04 0.10 0.66
Mining 0.000 1.02 0.001 0.015 0.64 0.01 0.00 0.35
Fabricated Products -0.006 1.09 0.001 0.009 0.60 0.05 0.02 0.33
Machinery -0.001 1.18 0.002 0.008 0.88 0.01 0.00 0.11
Electrical Equipment 0.002 1.16 -0.002 0.010 0.08 0.00 0.01 0.90
Aircraft 0.000 1.14 0.000 0.018 0.43 0.01 0.00 0.56
Shipbuilding & Railroad -0.001 1.00 0.002 0.008 0.62 0.10 0.03 0.24
Defense 0.003 0.82 -0.002 0.005 0.19 0.08 0.19 0.54
Personal Services -0.004 1.15 -0.006 0.010 0.05 0.01 0.09 0.86
Business Services -0.001 1.20 -0.003 0.006 0.04 0.01 0.04 0.91
Late Expansion Agriculture 0.001 0.89 0.000 0.002 0.28 0.20 0.19 0.33
Food Products 0.003 0.70 -0.005 -0.008 0.00 0.00 0.96 0.03
Candy & Soda 0.001 0.84 -0.001 -0.015 0.01 0.41 0.57 0.01
Beer & Liquor 0.003 0.80 -0.005 -0.007 0.00 0.04 0.86 0.10
Tobacco Products 0.006 0.65 -0.004 -0.025 0.00 0.10 0.90 0.00
Healthcare -0.001 1.12 -0.009 0.002 0.01 0.05 0.38 0.57
Medical Equipment 0.002 0.89 -0.004 -0.003 0.01 0.03 0.68 0.28
Pharmaceutical 0.002 0.82 0.000 -0.013 0.00 0.50 0.49 0.00
Coal 0.002 1.10 -0.005 -0.006 0.02 0.16 0.56 0.26
Petroleum & Natural 0.003 0.78 0.005 0.000 0.51 0.46 0.01 0.02
Early Recession Utilities 0.002 0.52 -0.002 -0.005 0.01 0.10 0.74 0.14
Communication 0.000 0.75 0.002 -0.010 0.03 0.87 0.10 0.00
Late Recession Recreation -0.003 1.18 -0.007 0.017 0.00 0.00 0.01 0.98
Entertainment 0.002 1.30 -0.004 -0.005 0.03 0.18 0.56 0.23
Printing & Publishing -0.001 1.02 -0.005 0.012 0.02 0.00 0.01 0.97
Consumer Goods 0.000 0.83 -0.004 0.003 0.02 0.01 0.22 0.76
Apparel 0.000 1.11 -0.010 0.018 0.00 0.00 0.00 1.00
Rubber & Plastic 0.000 1.05 -0.002 0.009 0.15 0.01 0.03 0.82
Textiles 0.000 0.98 -0.008 0.018 0.00 0.00 0.00 1.00
Construction Material 0.000 1.10 -0.003 0.003 0.02 0.01 0.24 0.72
Construction -0.001 1.30 -0.004 0.022 0.12 0.00 0.00 0.88
Automobiles & Truck -0.002 1.02 -0.004 0.007 0.06 0.01 0.13 0.80
Business Supplies 0.000 0.95 -0.001 0.007 0.34 0.06 0.04 0.56
Wholesale 0.000 1.09 -0.002 0.012 0.12 0.00 0.00 0.88
Retail 0.001 1.03 -0.008 0.004 0.00 0.00 0.20 0.80
Restaraunts & Hotels 0.001 1.13 -0.005 0.008 0.06 0.02 0.11 0.81
Banking 0.001 1.01 -0.001 -0.010 0.00 0.25 0.72 0.03
Insurance 0.002 0.90 -0.002 0.000 0.09 0.14 0.37 0.40
Real Estate -0.005 1.09 -0.002 0.014 0.28 0.01 0.02 0.70
Trading 0.002 1.23 0.002 -0.002 0.28 0.61 0.05 0.07

45
Table VIII. Comparison of strategy performance with changes in timing the business cycle

Notes: Table reports the performance of sector rotation and market timing when we advance or delay
strategy implementation from the base case by the number of months shown. The table reports
Jensen’s alphas (αJ) as annualized rates with White (1980) heteroskedasticity consistent t-statistics and
Sharpe ratio performance measures. Table includes market results at the bottom for comparison. The
performance results shown are before transaction costs.

Full Period 1948:01 - 2007:12

Strategy implementation Jensen's alpha t-statistic Sharpe ratio


Sector Rotation:
+ 3 month 1.0% 0.86 0.13
+ 2 month 1.0% 0.85 0.13
+ 1 month 1.9% 1.63 0.14
at turning point 2.3% 1.94 0.15
- 1 month 2.2% 1.81 0.15
- 2 month 1.8% 1.53 0.14
- 3 month 1.5% 1.27 0.14

Market-timing:
+ 3 month 1.7% 2.44 0.16
+ 2 month 2.6% 3.54 0.18
+ 1 month 2.9% 3.75 0.18
at turning point 2.5% 3.57 0.17
- 1 month 1.2% 3.10 0.15
- 2 month 0.9% 2.25 0.15
- 3 month 0.3% 0.63 0.13

Market - - 0.13

46
Table IX. Construction of sector rotation portfolios based on a one-period-ahead forecast
model

Notes: Table reports the composition of sector rotation portfolios constructed with a forecast
model using business cycle variables (BCV) that the literature shows forecast stock returns
over the course of business cycles. The business cycle variables comprise lagged changes in
the one-month Treasury bill, term-structure, default-spread, and dividend yield. We forecast
industry with parameters estimated with a regression of excess industry returns (ri,t-rf ) on a
constant, lagged change in the business cycle variables (ΔBCVi), and excess market returns
(rmkt-rf) using equation 9. We estimate the (γi) forecast parameters with a sixty-month rolling
window that moves forward each month and use these parameter estimates to obtain period
ahead forecasts of industry outperformance calculated as the sum of the gamma estimates
times current period changes in business cycle variables from the proceeding period. We
include all industries with positive forecast outperformance in the period-ahead sector rotation
portfolio. Panel A reports the average number of industries selected for inclusion during each
business cycle stage. Panel B reports the percentage of time an industry has positive forecast
outperformance and thus selected for inclusion during a particular business cycle stage. We
also test for any difference between the percentage of time the model selects an industry for
inclusion in the portfolio and a random 50/50 probability of inclusion with 10% statistical
significance indicated in bold. The shaded area in Panel B represents the business cycle stage
that conventional wisdom considers optimal.

ri , p ,t − rf p ,t = α 0 + β BCPj , p BCPj , p ,t + β mkt , p (rmkt , p ,t − rf p ,t ) + ε t


(9)

Panel A: Average number of industries selected for inclusion by model


Full Early Middle Late Early Late
Period Expansion Expansion Expansion Recession Recession
23 23 23 23 22 23

Panel B: Percentage of time model forecasts excess industry returns and includes in sector rotation portfolio
Full Early Middle Late Early Late
Period Sector/Industries Period Expansion Expansion Expansion Recession Recession
Early Expansion - Stage I Computers 46 48 48 43 40 48
Computer Software 38 38 37 40 38 33
Electronic Equipment 47 47 47 47 49 50
Measuring & Control 50 50 52 47 43 54
Shipping Containers 48 48 42 54 55 46
Transportation 49 56 46 44 51 57
Middle Expansion - Stage II Chemicals 50 53 49 46 55 46
Steel Works 54 57 56 48 43 67
Precious Metals 40 38 42 40 36 43
Mining 51 53 52 48 51 52
Fabricated Products 40 34 45 44 30 39
Machinery 48 56 45 47 45 39
Electrical Equipment 47 47 47 45 62 43
Aircraft 48 53 48 47 36 50
Shipbuilding & Railroad 50 48 50 54 36 54
Defense 40 39 41 41 36 39
Personal Services 48 46 52 47 53 43
Business Services 52 54 55 49 36 57

47
Table IX. Continued
Panel B: Continued
Full Early Middle Late Early Late
Period Sector/Industries Period Expansion Expansion Expansion Recession Recession
Late Expansion - Stage III Agriculture 53 53 57 51 53 52
Food Products 49 48 46 52 49 52
Candy & Soda 39 33 38 43 53 33
Beer & Liquor 49 49 45 50 60 50
Tobacco Products 49 47 47 54 49 52
Healthcare 32 31 37 34 21 22
Medical Equipment 48 40 49 57 53 37
Pharmaceutical 51 45 49 57 57 48
Coal 49 47 49 49 64 41
Petroleum & Natural 52 54 51 51 51 59
Early Recession - Stage IV Utilities 47 39 48 52 55 48
Communication 51 45 57 52 55 52
Late Recession - Stage V Recreation 48 55 43 47 36 52
Entertainment 46 47 48 42 43 57
Printing & Publishing 48 51 48 48 36 43
Consumer Goods 48 48 47 45 51 54
Apparel 52 55 51 52 45 50
Rubber & Plastic 50 60 48 48 40 39
Textiles 49 57 47 46 40 43
Construction Material 48 49 42 49 47 54
Construction 49 51 49 50 40 48
Automobiles & Truck 52 56 49 53 43 54
Business Supplies 50 55 51 45 49 46
Wholesale 49 50 52 45 51 48
Retail 48 45 48 48 57 48
Restaraunts & Hotels 50 52 53 44 49 57
Banking 51 44 52 57 53 46
Insurance 50 41 55 56 55 37
Real Estate 51 56 52 49 36 54
Trading 53 51 58 51 49 54

Table X. Alternative sector rotation strategy

Notes: Table reports the performance of industries that over the 1948-2007 period provided
statistically significant outperformance for the indicated business cycle stage. Column 3
reports the percentage of time (%) that that an industry actually realized statistically
significant Jensen’s alpha outperformance. We test the percentage of time an industry actually
provides outperformance against a random 50/50 chance with 10% statistical significance
indicated in bold. Column 4 reports annualized Jensen’s alpha estimates while the last column
reports corresponding White (1980) heteroskedasticity consistent t-statistics highlighted for
statistical significance at 10%.

Period Sector/Industries % αJ t-statistic


Middle Expansion - Stage II Candy & Soda 62 0.10 2.51
Pharmaceutical 55 0.07 2.35
Late Expansion - Stage III Mining 53 0.08 2.14
Tobacco Products 59 0.09 1.83
Early Recession - Stage IV Shipping Containers 62 0.10 2.42
Food Products 72 0.13 2.12
Utilities 66 0.13 2.45
Entertainment 66 0.20 2.71
Late Recession - Stage V Personal Services 71 0.15 2.06
Food Products 69 0.16 2.78
Tobacco Products 61 0.17 1.75

48
Figure I. Conventional Wisdom: Sector Rotation across the Business Cycle

Source: http://personal.fidelity.com/products/funds/content/sector/cycle.shtml

Source: http://www.marketoracle.co.uk/Article3618.html

Source: http://www2.standardandpoors.com/spf/pdf/index/Global_Sector_Investing.pdf

49
Figure II. Stylized business cycles with stage partitions

Notes: Figure illustrates a stylized economic business cycle. The official government agency
responsible for dating U.S. business cycles is the National Bureau of Economic Research (NBER). The
NBER publishes dates for business cycle peaks and troughs. We measure phases of expansion from
trough to peak and recession from peak to trough. Similar to Stovall (1996), we divide expansions into
three equal stages (early/middle/late) and recessions into two stages (early/late).

NBER peak

Expansion Recession

Stage I Stage II Stage III Stage IV StageV

NBER trough NBER trough

Stages of Expansion Stages of Recession


Early Expansion - Stage I Early Recession - Stage IV
Middle Expansion - Stage II Late Recession - Stage V
Late Expansion - Stage III

50
Figure III. CFNAI business cycle stages

Notes: Figure illustrates the CFNAI economic indicator over the period 1968-2007. Shaded areas
indicate NBER defined periods of economic contraction. The range of CFNAI values covering the full
sample are partitioned into 5 equal periods of economic activity that can be thought of as
corresponding to periods of early expansion (SI), middle expansion (SII), late expansion (SIII), early
recession (SIV), and late recession (SV). The partitions between adjoining stages are shown with
delineations at CFNAI values of 0.57, 0.26, -.01, and -.045 between periods SI|SII, SII|SIII, SIII|SIV, and
SIV|SV respectively.

2.5

1.5

0.5

1968 1974 1979 1985 1990 1996 2001 2007


-0.5

-1.5

-2.5

-3.5

-4.5

NBER Recessions S1|S2 S2|S3 S3|S4 S4|S5 CFNAI

51
Figure IV. Distribution of Jensen’s alphas of sectors in different stages

Note: Chart illustrates the actual percentage of time that industry Jensen’s alpha t-statistics
fall within the indicated range and compares with the expected distribution of t-statistics under
a normal distribution. We calculate Jensen’s alphas for each industry during each business
cycle for a total 240 corresponding t-statistics.

20%
18%
16%
14%
Frequency

12%
10%
8%
6%
4%
2%
0%
< ‐3.29

‐3.29 to ‐2.58

‐2.58 to ‐1.96

‐1.96 to ‐1.65

‐1.65 to ‐1

‐1 to ‐0.5

‐0.5 to 0

0 to 0.5

0.5 to 1

1 to 1.65

1.65 to 1.96

1.96 to 2.58

2.58 to 3.29

> 3.29
T‐Statistic Range

Actual distribution of Jensen's alpha t‐statistics Expected distribution of Jensen's alpha t‐statistics  assuming  a normal distribution

52

View publication stats

You might also like