SEASONALITY IN momentum profitability
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
SEASONALITY IN
momentum profitability
ANQI (ANGEL) ZHONG, PhD student, Department of Banking and Finance, Monash University
MANAPON LIMKRIANGKRAI, Lecturer in Finance, Department of Banking and Finance, Monash University
PHILIP GRAY, Professor of Finance, Department of Banking and Finance, Monash University
In Australia, and around the world, momentum trading generates
economically and statistically significant profits. This paper documents
seasonalities in momentum profitability rather than examining
returns averaged across all months. We report a strong reversal
around the financial year end and apparent quarter-end seasonality
in momentum profits. Preliminary tests support the hypothesis that
seasonality in quarterly equity returns is driven by window dressing by
institutional investors.
Momentum trading refers to a strategy that enters averages momentum profits across all calendar
long positions in recent winners and short positions months. While findings like this are common in the
in recent losers. Academic research documents momentum literature, it is rare to see the profitability
economically and statistically significant momentum broken down to a month-by-month analysis. This
returns in many global equity markets. As such, the is surprising, especially given that seasonalities in
momentum phenomenon has had a profound impact equity returns are well documented.
on both academic research and investment practice.
This paper, therefore, examines the seasonality of
While academics continue to search for a better
momentum profits. Rather than reporting an overall
understanding of the anomaly, momentum filters are
average monthly momentum profit, we tabulate
an important component of many fund managers’
momentum profits separately for each calendar
stock selection process.
month. There are several attractive reasons to
Australia is no exception when it comes to interest study momentum seasonality. First, in the event
in momentum profitability, with numerous studies that distinct seasonalities exist, there are obvious
existing on this topic.1 While, at face value, some implications for professional traders. Second, from
findings appear to be conflicting, the different an academic perspective, documenting seasonalities
conclusions reached are largely attributable to may guide efforts to better understand what drives
the ways in which stocks are weighted within the momentum phenomenon.
portfolios. In general, the evidence in favour of
momentum profitability is overwhelming when Data and methodology
stocks are weighted in portfolios according to their Our data are sourced from the Share Price and Price
market capitalisations (i.e. value weighted), and Relative (SPPR) file maintained by Securities Industry
less convincing when stocks are equally weighted. 2 Research Centre of Asia-Pacific (SIRCA). Monthly
Naturally, this suggests that momentum is stronger return and market capitalisation data is obtained
among large-cap stocks, which carry more influence for all ASX-listed ordinary stocks over the 1974–
on portfolio returns under value weighting. 3 2012 period.
In documenting momentum profitability, it is Previous Australian studies report strong momentum
common to report the average monthly return profits over medium-term horizons of six to 12
differential between portfolios taking long positions months (see, for example, Demir et al. 2004;
in past winners and short positions in past losers. For Brailsford and O’Brien 2008). For consistency,
example, over the 1974–2012 period, this paper finds we adopt a six-month period over which past
that a long–short momentum strategy on Australian performance is ranked, and a six-month period over
equities generates an average monthly return of 1.72 which momentum is tested. Starting in June 1974,
per cent (t=3.69). Note that this result effectively past momentum is calculated for all stocks that have
a valid trade price at both 31 December 1973 and
6 JASSA The Finsia Journal of Applied Finance ISSUE 2 201430 June 1974 (dividends paid over this period are the winners (2.85 per cent), thereby driving the
also included in the return). Stocks are sorted into 10 momentum loss in July (-2.39 per cent). This July
portfolios on the basis of this past momentum, with ‘contrarian’ (or negative momentum) effect has also
winners (losers) comprising the 10 per cent of stocks been documented by Durand et al. (2006).
with the best (worst) past six-month performance.
The second notable aspect of Figure 1 is a quarter-
The weight given to each stock in a portfolio reflects
end momentum effect. This is more readily apparent
its market capitalisation (i.e. they are value-weighted
when concentrating on statistical significance
portfolios). Once established, the portfolio is held
(denoted by asterisks). While momentum profits are
without rebalancing from July to December 1974.
positive in every month except July, they are only
The return to each portfolio is calculated in each of
statistically significant in March, June, September,
the six holding period months. In December 1975,
November and December. Sias (2007) reports a
the ranking process is repeated and portfolios are
similar pattern in US momentum profits, which he
reformed for another six months.
attributes to ‘window dressing’. He conjectures that
This portfolio formation procedure is repeated institutional investors may abandon losers to avoid
every six months through to (and including) June reporting embarrassing stocks in their holdings.
2012.4 This generates a time series of 462 monthly Similarly, they may make last-minute purchases
returns to the 10 momentum portfolios spanning of winners to give the appearance that they held
July 1974 through to (and including) December 2012. respectable stocks throughout the period (Sias 2007,
The momentum profits reported in this paper are p. 48). Such behaviour is likely to cluster in quarter-
calculated as the difference between the returns to ending months and/or at the end of the financial year,
the winner and loser portfolios. With nearly 40 years thereby contributing to the momentum seasonality.
of data, it is unlikely that the results will be unduly In Australia, at the end of the financial and calendar
affected by one or two unusual months. years, it is common to see media coverage on the
best- and worst-performing stocks. Consistent with
Results this, momentum profits in June and December are
As noted above, the overall return to momentum among the strongest.
trading during our sample is 1.72 per cent per month
(i.e. averaged over all months). Figure 1 breaks this
down by calendar month. Two particular facets In Australia, at the end of the financial and
warrant discussion. First, the most distinctive feature calendar years, it is common to see media
is the large positive return in June (3.23 per cent)
coverage on the best- and worst-performing
followed immediately by a large negative return
stocks. Consistent with this, momentum
in July (-2.39 per cent). Given that these returns
profits in June and December are among
straddle the financial year end, it is natural to think
that tax incentives may drive this seasonality. Turn-
the strongest.
of-the-year seasonal returns are often attributed to
tax-loss selling, whereby investors dump holdings of
stocks that have performed poorly in order to realise We explore the window dressing story as follows.
tax deductible losses prior to the year end. While this If quarterly seasonality in momentum profitability is
argument is often made to explain seasonality in the an outcome of window dressing, then the quarterly
market as whole, it is acutely relevant in a study of effect should become stronger over time as the
winners and losers. prominence of institutional investors has increased.
The introduction of the Australian Superannuation
We explore the tax-loss selling argument further Guarantee Scheme (SGS) in 1992 provides an
by considering the unique contribution of winners ideal setting in which to test this hypothesis. The
and losers to the overall momentum profit in June. SGS requires employers to make compulsory
Although not explicitly shown in Figure 1, the short superannuation contributions on behalf of their
positions in losers (3.92 per cent) account for most employees. Originally set at 3 per cent of the
of the momentum profits in June (3.23 per cent). employee’s income, the rate is currently 9.25 per cent
This finding is consistent with the conjecture that and will progressively increase to 12 per cent by 2019.
investors who inadvertently find themselves taking As a direct consequence of the SGS, Australia now
long positions in these loser stocks as the financial has one of the largest managed fund industries in the
year end approaches may sell out for tax reasons. world, with approximately $1.8 trillion in funds under
The downward selling pressure manifests itself in management. Superannuation funds have clearly
large negative returns to these losers in June, which been the key driver in the significant growth in the
benefits momentum traders holding short positions Australian managed fund industry, accounting for
in losers. 5 Curiously, in July, the losers (5.24 per over 80 per cent of total market share.6
cent) rebound with such strength that they dwarf
JASSA The Finsia Journal of Applied Finance ISSUE 2 2014 7Accordingly, we analyse seasonality in momentum
profitability over two different periods: the pre-SGS
We document a distinct turn-of-the-year effect
period covering 1974 to (July) 1992, and (ii) the post-
SGS period covering (July) 1992 to 2012. Figure 2
and strong quarterly effects. There is anecdotal
depicts the average monthly momentum profits for evidence that tax-loss selling and window
the pre- and post-SGS periods. Separate results are dressing by institutional investors contribute
shown for averages over all months, quarter-end to these patterns. Further, these patterns only
months, non-quarter-end months and non-quarter- exist during the past 20 years. The key findings
end months (ex. July).7 appear to be driven exclusively by the period
In the pre-SGS period, there is little discernible
that followed the introduction of the SGS
difference in momentum profits during quarter-end in 1992. In fact, momentum trading appears
months (0.82 per cent), non-quarter-end months unprofitable prior to 1992.
(0.51 per cent) and non-quarter-end months (ex.
July) (0.68 per cent). Statistical tests (not explicitly
reported) cannot detect a difference in momentum We document a distinct turn-of-the-year effect
profits between quarter-end and non-quarter- and strong quarterly effects. There is anecdotal
end months in the pre-SGS period. In fact, overall evidence that tax-loss selling and window dressing
momentum profitability in this period (0.61 per cent) by institutional investors contribute to these
is statistically insignificant. patterns. Further, these patterns only exist during
the past 20 years. The key findings appear to be
After the introduction of the SGS in 1992, the
driven exclusively by the period that followed the
profitability of momentum trading clearly increases.
introduction of the SGS in 1992. In fact, momentum
The average monthly return rises from 0.61 per
trading appears unprofitable prior to 1992.
cent to 2.70 per cent (highly significant). Further,
While it is difficult to attribute this entirely to the
post-SGS, quarter-end months generate momentum
introduction of the SGS and the resulting boom in
profits of 4.58 per cent, which are significantly higher
the managed funds industry, further research may
than non-quarter-ending months (1.76 per cent). Even
shed light on how the increasing prominence of
excluding July, the quarter-end return was still 2 per
institutional investors contributes to the momentum
cent per month higher.
phenomenon.
The dramatic difference pre- and post-SGS raises
FIGURE 1: Momentum profitability by month,
the question of whether the seasonality depicted 1974 to 2012
in Figure 1 manifests across the entire sample, or
whether it may be driven in the post-SGS period. This figure shows the average momentum
Figure 3, which presents the average momentum profitability in each calendar month. Momentum
profits on a month-by-month basis separately for profit is the difference in return on a portfolio long
the pre- and post-SGS periods, suggests the latter. past winners and short past losers. The sample spans
The top diagram (pre-SGS period) shows little 1974 to 2012. Quarter-end months are differentiated
consistency in momentum profits, little evidence of with darker shading. ***, ** and * signal that the
statistical differences from zero, no turn-of-the-year reported return is statistically different from zero at
effect, and no evidence of quarter-end seasonality. In the 1%, 5% and 10% confidence levels, respectively.
contrast, the bottom diagram (post-SGS) shows that 4%
3.23** 3.33** 3.22***
momentum trading has been consistently profitable
3%
(both economically and statistically) since 1992. The 2.49
2.30*
2.52*
turn-of-the-year effect is prominent. Quarter-end 1.77 1.80
2%
months (shown in teal) are highly profitable and 1.18
1.09
statistically significant. 1%
0.20
Conclusions 0%
While most studies focus on the average profitability
of momentum trading (i.e. when averaged across -1%
all months), this study examines seasonality in
-2%
momentum profits.
-2.39*
-3%
-4%
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Return
8 JASSA The Finsia Journal of Applied Finance ISSUE 2 2014FIGURE 2: Comparison of momentum profitability 6%
6
in pre- and post-SGS periods
5%
This figure compares momentum profitability before
4% 3.71
and after the introduction of the Superannuation
Guarantee Scheme in June 1992. Momentum profit 3%
2.40
2.83
2.21
is the difference in return on a portfolio long past
2%
winners and short past losers. For each subperiod,
we report the average monthly momentum profits 1% 0.47 0.47
0.33 0.24
for (i) all months, (ii) non-quarter-ending months, (iii)
0%
non-quarter-ending months (excluding July) and (iv) -0.13
-0.62 -0.71
quarter-ending months. -1%
5.0% -2%
4.58
4.5% -3%
4.0% -4% -3.86**
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
3.5%
Return
3.0% 2.70 5.85**
2.59 6% 5.77***
5.30*
2.5%
5%
4.49**
2.0% 1.76
4% 3.69*
1.5% 3.00 3.00**
3%
0.82 2.25
1.0% 0.68
0.61 2% 1.79
0.51 1.38
0.5%
1%
0.0%
Pre-SGS period Post-SGS period 0%
-0.09
All months Non-quarter-end months -1%
Non-quarter-end months ex. July Quarter-end months
-2%
FIGURE 3: Seasonality in momentum profits pre- -3%
SGS and post-SGS
-4% -3.84*
This figure shows the average momentum profitability Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
in each calendar month. Momentum profit is the Return
difference in return on a portfolio long past winners
and short past losers. The top (bottom) panel shows
seasonality on momentum profits before (after) the
introduction of the Superannuation Guarantee Scheme
in June 1992. Quarter-end months are differentiated
with darker shading. ***, ** and * signal that the
reported return is statistically different from zero at
the 1%, 5% and 10% confidence levels, respectively.
JASSA The Finsia Journal of Applied Finance ISSUE 2 2014 9Acknowledgement References
The authors gratefully acknowledge the comments Bettman, JL, Maher, TRB and Sault, SJ 2009, ‘Momentum profits
in the Australian equity market: A matched firm approach’,
and suggestions of Kevin Davis (Managing Editor),
Pacific-Basin Finance Journal, vol. 17, no. 5, pp. 565–79.
Peter Brooke and the anonymous referee.
Brailsford, T and O'Brien, MA 2008, ‘Disentangling size from
momentum in Australian stock returns’, Australian Journal of
Endnotes Management, vol. 32, no. 3, pp. 463–84.
1 V
anstone, Hahn and Finnie (2013) provide a succinct summary
Demir, I, Muthuswamy, J and Walter, T 2004, ‘Momentum returns
of the Australian momentum literature, as well providing
in Australian equitites: The influences of size, risk, liquidity and
further evidence of momentum profitability among the S&P/
return computation’, Pacific-Basin Finance Journal, vol. 12, no. 2,
ASX 100 stocks.
pp. 143–58.
2 This
point is borne out in studies by Brailsford and O’Brien
Dou, P, Gallagher, DR and Schneider, D 2013, ‘Dissecting
(2008), O’Brien et al. (2010), Dou et al. (2013) and Gray
anomalies in the Australian stock market’, Australian Journal
(2014), all of which present results under both value- and
of Management, vol. 38, no. 2, pp. 353–73.
equal-weights. Gray (2014) presents a separate analysis of
momentum within the top 500 stocks and shows it is stronger Durand, RB, Limkriangkrai, M and Smith, G 2006, ‘Momentum
than in the population. in Australia: A note’, Australian Journal of Management, vol. 31,
no. 2, pp. 355–64.
3 E
ven when studies weight stocks equally, the strength of
momentum profits is apparent when analysis is restricted Gray, P 2014, ‘Stock weighting and non-trading bias in estimated
to large caps (see Hurn and Pavlov 2003; Demir et al. 2004; portfolio returns’, Accounting and Finance, 54, pp.467–503.
Bettman et al. 2009).
Hurn, S and Pavlov, V 2003, ‘Momentum in Australian stock
4 T
his is a ‘non-overlapping’ approach to implementing returns’, Australian Journal of Management, vol. 28, no. 2,
momentum trading. Many academic papers form momentum pp. 141–55.
portfolios monthly, which implies that a series of overlapping
portfolios are held at any given point in time. From a practical O'Brien, MA, Brailsford, T and Gaunt, C 2010, ‘Interaction of size,
perspective, the transaction costs involved in maintaining book-to-market and momentum effects in Australia’, Accounting
overlapping portfolios would be non-trivial. Nonetheless, and Finance, vol. 50, no. 1, pp. 197–219.
the findings of the paper are virtually unchanged when
Sias, R 2007, ‘Causes and seasonality of momentum profits’,
overlapping portfolios are used. Similarly, robustness analysis
Financial Analysts Journal, vol. 63, pp. 48–54.
shows that the findings are not sensitive to the choice of June
and December as the portfolio formation points, or skipping Vanstone, B, Hahn, T and Finnie, G 2012, ‘Momentum returns to
one month between identification of winners/losers and S&P/ASX100 constituents’, JASSA, issue 3, pp. 15–18.
entering long/short positions. Full results for this robustness
analysis are available on request.
5 T
he pattern of momentum profits around the financial year
end shown in Figure 1 is remarkably similar to US findings
reported by Sias (2007). Noting that the US has a December
financial year end, Sias (2007) documents large positive
momentum profits in December on average, and large
negative profits in January.
6 A
usTrade (10 March 2011) Data Alert: www.austrade.gov.au/
Invest/Investor-Updates/2011/Data-Alert-110311-Australias-
Managed-Funds-Assets-now-at-1-8-trillion
(accessed 12 July 2013).
7 The large negative July return shown in Figure 1 has
the potential to confound results for non-quarter-end
months. Hence, we also consider non-quarter-end months
excluding July.
10 JASSA The Finsia Journal of Applied Finance ISSUE 2 2014You can also read
