Professional Documents
Culture Documents
Seasonality in Government Bond Returns
Seasonality in Government Bond Returns
Seasonality in Government Bond Returns
Adam Zaremba
Tomasz Schabek
University of Lodz
Author’s Note
Department of Investment and Capital Markets, Poznan University of Economics and Business,
Abstract
The study investigated both the January effect and the "sell-in-May-and-go-away" anomaly in
government bond returns. It also tested whether the two seasonal patterns impact the
performance of fixed-income factor strategies related to volatility, credit risk, value, and
momentum premia. Our examination of government bond markets in 25 countries for years
1992-2016 proved that both the bond returns and factor premia had remained unaffected by the
January and "sell-in-May" effects. These seasonal patterns in government bond markets appear
Keywords: seasonal anomalies, calendar anomalies, January effect, sell in May and go
commonalities in cross-sectional return patterns across various asset classes. The value effect -
the tendency of low-priced assets to outperform high-priced securities - which first originated
from equity markets, have now been found in: commodities, fixed income, and bonds (Asness,
Moskowitz, & Pedersen, 2013). The low-volatility effect - the tendency of risky assets to
underperform - has been proven in: equities, fixed income, and commodities (Frazzini and
Pedersen, 2014), and another widely respected cross-sectional effect: momentum, - the
tendency of assets successful in the past to continue to outperform in the future - has also been
found across all major asset classes (Asness, Moskowitz, & Pedersen, 2013). Finally, in 2015
Koijen, Moskowitz, Pedersen, and Vrugt demonstrated that the carry strategy - assuming
underpinning all these studies probes the integration of global financial markets and the
In comparison to the common strategies based on value, momentum, carry, and low-
risk, the seasonal patterns have been most comprehensively studied in equities whereas other
asset classes have yet to be covered by such extensive research, with government bonds in
particular appearing to have slipped the attention of the academic community. Thus, the primary
goal of this study was to bridge this gap by conducting a comprehensive examination of
We focused on two well-documented seasonal anomalies: the January effect and the
companies to display particularly high returns in January. The evidence thereof dates back to
the studies of Wachtel (1942), Rozeff and Kinney (1976), and has been reconfirmed in more
recent and long-term samples (Haug & Hirschey, 2006). The "sell-in-May-and-go-away-effect"
(in short: "sell-in-May", also referred to as the Halloween indicator) is reflected in the
outperformance of stocks within the period from November to April relative to the remainder
of the year. The phenomenon, originally identified by Bauman and Jacobsen (2002) across 36
equity markets, has been since confirmed in broader and longer samples by Castro and Schabek
(2014) or Jacobsen and Zhang (2014). Here we extended the research on these seasonal
explanation of the January effect offered by Ritter (1988), according to which the returns in
January can be driven by the inflow of money from the bond markets where the capital was
temporarily "parked". Thus, the abnormal positive returns in January should coincide with the
capital outflow from the bond markets, effectively implying abnormal negative returns on
government bonds in January. In this study, we extended this explanation to include the "sell-
in-May" anomaly, checking whether the positive abnormal equity returns within the period
While the January and "sell-in-May" effects may impact the government bond returns,
they may not exert similar influence on all market segments. For example, Clayton, Delozier,
and Ehrhardt (1989) suggested that the magnitude of the January effect might depend on the
bond maturity. Thus, if the January effect impacts only some segments of fixed-income
consequence, we were also interested whether both the January and "sell-in-May" anomalies
be reflected not only in raw returns, but also in government bond factor premia. In equities, the
January effect has been investigated in relation to: the size effect (e.g., Horowitz, Loughran, &
Savin, 2000), value effect (e.g., Davis, 1994; Loughran, 1997), and momentum (e.g., Jegadeesh
& Titman, 2001; Yao, 2012, Zaremba, 2015). Here, we considered four grant government bond
return factors: volatility, credit risk, value, and momentum and tested their performance in the
This study aims to contribute by providing new insights into asset pricing of
international government bonds. To the best of our knowledge, it is the first paper to examine
addition, no study to date has investigated the "sell-in-May" effect in government bonds or
researched the seasonal patterns in government bond factor premia. The results bear
implications for both academic and practical purposes. First, they allow to understand the
behaviour of government bond returns, and could also be viewed as a test of informational
efficiency in international bond markets in terms of seasonal anomalies. Second, they verify
the implications of the "parking-the-proceeds" hypothesis by Ritter (1988). Third, the results
might help to design optimal asset pricing factors for multi-factor models to explain the cross-
section of government bond returns. 1 Fourth, the outcomes can be utilised by market
commonalities in asset-pricing patterns across different asset classes (Asness, Moskowitz, &
Pedersen, 2013; Frazzini and Pedersen, 2014; Koijen, Moskowitz, Pedersen, and Vrugt, 2015;
Zaremba, 2016), (2) regarding seasonal patterns in anomalies and factor premia (Davis, 1994;
Loughran, 1997; Horowitz, Loughran, & Savin, 2000; Jegadeesh & Titman, 2001; Yao, 2012;
Zaremba, 2015), and - most importantly - (3) regarding the seasonal patterns in government
bond behaviour (Schneeweis & Woolridge, 1979; Smirlock, 1985; Chang & Pinegar, 1986;
1
For example, the momentum factor in equities is frequently based on the trailing return for the 12-month period
in order to disentangle the influence of seasonal patterns in cross section of returns. The question whether
Clayton, Delozier, & Ehrhardt, 1989; Clare & Thomas, 1992; Chan & Wu, 1993, 1995; de
Vassal, 1998; Lavin, 2000; Chieffe, Cromwell, & Yoder, 2000; Smith, 2000, 2006). Generally,
the existing evidence concerning seasonal patterns in government bonds remains inconclusive.
The single-country studies of Schneeweis & Woolridge (1979), Smirlock (1985), Chang and
Pinegar (1986), Clayton, Delozier, and Erhardt (1989), and Lavin (2000) have found significant
evidence for seasonality, yet the broader international investigations, including Clare and
Summarising the principal findings of this study, we identified the January effect in the
returns on Italian government bonds which outperformed in a given month although the
empirical data failed to provide consistent evidence of this patterns in the 24 out of 25 countries
we studied. The "sell-in-May" anomaly was reflected in abnormally low returns within the
period from November to April in Canada, the United States, and partly Australia, yet it failed
to appear in the 22 of 25 markets. Summing up, the aggregate tests proved that both the January
and "sell-in-May" patterns played a minor role in international government bond returns.
Furthermore, also the factor premia appeared unaffected by the seasonal patterns. The data
revealed no abnormal returns on volatility, credit risk, value, or momentum strategies related
The remainder of the study is divided into Section II, presenting our data sample and
research methods employed, Section III, discussing the results, and Section IV, which
As the study aimed to investigate seasonal patterns in government bond returns and
factor premia, we examined the returns on bond buckets using a regression analysis, formed
synthetic volatility, credit risk, value, and momentum strategies, and evaluated their
performance with analogous procedures. In this section, firstly, we outline our data sample.
Next, we discuss the design of factor portfolios, and we describe the regression-based tests.
This research was based on Bloomberg/EFFAS Total Return Bond Indices for 25
countries for the period January 1992 - June 2016. The sample encompassed all of the countries
and the entire period covered by Bloomberg/EFFAS. The indices were calculated separately
for five different maturity buckets: 1-3 years, 3-5 years, 5-7 years, 7-10 years, and over 10
years, in total, investigating 125 international government bond buckets. This made this sample
considerably broader then the former studies of return patterns in international government
bonds, which include Asness, Moskowitz, and Pedersen (2013), who researched 10 countries,
Frazzini and Pedersen (2014), covering 9 countries, and Beekhuizena, Duyvesteyna, Martens,
and Zomerdijk (2016), targeting 10 countries. To its further advantage, our sample included a
unique default event: Greece, and the necessary additional characteristics of the government
bond indices: e.g. the average duration or aggregate market value found in the Bloomberg
database.
The calculations were based on monthly returns. This offers a compromise between
ensuring a considerable number of observations for statistical interfering and avoiding the
impact of microstructure issues, which can in turn influence daily intervals. In order to use a
consistent currency approach across multiple markets and at the same time to disentangle the
currency and bond returns, we utilised returns hedged against US dollar.2 We collected data in
local currencies, and then adjusted them for hedging costs based on the 1-month forward points
2
For robustness, we also examined unhedged returns and we found no qualitative differences in results.
Note. The table presents the research sample of international government bonds buckets. R is the mean monthly arithmetic return; Vol the standard deviation of monthly returns,
and N a number of monthly observations (returns). The columns' headlines indicate maturities of bonds within the buckets stated in years (Y): 1-3, 3-5, 5-7, 7-10, and over 10
years.
All other data adopted in this study, e.g. sovereign ratings, or macroeconomic data, are
also sourced from Bloomberg. For cash rates, we closely followed the argumentation of
Beekhuizena, Duyvesteyna, Martens, and Zomerdijk (2016), using the 1-month Eurocurrency
rate. The rates seem the most relevant to bond investors, as only governments can attain
financing at the T-bill rate. Furthermore, the implied interest rate differentials in FX forwards,
which are used to hedge currency risk, approximate differences between Eurocurrency rates.
The final benefit of the Eurocurrency rates is its better data coverage compared to the
government bond returns: volatility, credit risk, value effect, and momentum premium. In the
case of each factor, we observed a uniform portfolio formation procedure which was consistent
across all of the returns patterns. Thus, each month we ranked all bond buckets based on a
sorting variable related to the return determinant. Next, we formed zero-investment portfolios
that were long (short) in the quantile of buckets with the highest (lowest) variable. To assure
the robustness of the results, we used three different types of quantiles: tertiles, quartiles, and
quintiles. Furthermore, we utilised two distinct weighting methods: equal and based on value.
variables: (1) adjusted duration, (2) beta estimated against the value-weighted portfolio of all
of 125 bond buckets, (3) idiosyncratic volatility from the regression on the value-weighted
portfolio of all 125 bond buckets, (4) standard deviation of monthly returns, (5) empirical value
at risk with a 5% threshold, (6) downside deviation of monthly returns, and (7) duration-times-
yield measure, i.e. the portfolio duration multiplied by its yield to maturity, following de
Carvalho, Dugnolle, Lu, and Moulin (2014). The variables (2)-(6) were estimated based on the
The second category of factors - credit risk - encompassed four alternative measures.
net debt-to-GDP ratios (8). Second, we used the sovereign risk assessments provided by the
Economist Intelligence Unit (9). Third, we employed an average sovereign bond rating from
Moody's, S&P and Fitch (10).3 Finally, we also considered the total market value of all the
bonds within a given bucket (11). Naturally, this measure is not strictly related to credit risk,
but as the developed markets have generally more sovereign debt outstanding, we decided to
The measurement of the value effect in government bonds, which is covered by the third
category of our factors, proves elusive and inconsistent across academic papers. For example,
Asness, Moskowitz, and Pedersen (2013) adopted the 60-month trailing return arguing this
measure to be strongly correlated with book-to-market ratio in the case of individual stocks. On
the other hand, Asness, Ilmanen, Israel, and Moskowitz (2015) advocated the yield to maturity
adjusted for expected inflation. This is also closely related to the concept of bond carry, usually
calculated as the difference between bond yield to maturity and some short-term cash rate
(Koijen, Moskowitz, Pedersen, & Vrugt, 2015). Finally, stock market practitioners usually
employ some relative value techniques which relate the expected yield to various underling risk
factors. Thus, to assure the robustness of our tests we employed a range of approaches. The first
two variables were: yield-to-maturity (12) and term premium (13), i.e. difference between the
yield-to-maturity and 1-month Eurocurrency rate in the analysed country. We also considered
three relative value approaches, under which the markets were sorted on the residuals of the
3
Precisely, we ranked all the ratings from the three rating agencies from 1 to 24 and then used an averaged ranking
value.
term premium (see [13]) regressed on another risk factor. Thus, we used relative value ([14],
residuals from the regression on adjusted duration as in [1]), credit relative value (([15],
residuals from the regression on averaged credit rating as in [10]), and term-and-credit relative
value ([16], residuals from a multiple regression on adjusted duration and averaged credit
rating). Finally, we also adopted long-term 48-month yield change (17), as advocated by
The fourth category of variables is momentum, i.e. the tendency of assets that performed
well (poor) in the past to continue to outperform (underperform) in the future. The phenomenon
was identified in government bonds by Luu and Yo in 2012 and Duyvesteyn and Martens
(2014) and others. For momentum we adopted 6 alternative definitions: 6-month and 12-month
price change (18, 19), i.e. the percentage change in price over past 6 and 12 trailing months, 6-
month and 12-month yield change (20, 21), i.e. the nominal change in yield to maturity over
past 6 and 12 trailing months, and the relation of current price to its 6-month (22) and 12-month
(23), i.e. the moving average estimated from the monthly data.
investment portfolios considered in this study. The volatility was clearly positively related to
future returns and the zero-cost portfolios displayed mean positive and significant mean
monthly returns ranging from 0.24% to 0.49%, depended on a given portfolio construction
approach. The credit premium was much less consistent and mostly insignificant. Interestingly,
the profitability of this factor had been doubted in many earlier studies (see Asvanunt and
Richardson [2016] for discussion). The value strategies proved profitable and highly
significant., The momentum strategies displayed positive and significant returns in nearly all
alternative definitions and across multiple portfolio construction methods while the worst
patterns, including Bouman and Jacobsen (2002) or Castro and Schabek (2014). Using the
ordinary least squares method, we estimated the parameters of the following regression
equations:
, (1)
, (2)
where rit is the log-return on i (i.e. the asset in month t), εit is the standard error, and αiJAN, βiJAN,
αiMAY, and βiMAY are the regression parameters. JANt is a dummy variable of 1 when month t is
November, December, January, March, or April, or 0 otherwise. Thus, the βiJAN and βiMAY
parameters could be interpreted as the average abnormal returns in January and correspondingly
within the period November-April. Finally, for each asset we tested the null hypotheses, where
the βiJAN and βiMAY parameters equaled 0 and the alternative hypothesis assuming the contrary.
In all cases, the parameters with the corresponding t-statistics were based on the Newey-West estimator
Finally, as we were also interested whether the January or "sell-in-May" returns would
significantly depart from zero on average within a group of countries or bond buckets, we
developed a bootstrap test (abbreviated BT) to investigate this issue. The BT test was based on
a bootstrap approach in which new samples were randomly drawn with replacement from the
original sample of the monthly portfolio log-returns. Where {rit, t=1, ..., T; i=1, ,2, ..., N) be the
original set of log-returns recorded for N assets over T time periods. Using the stationary
bootstrap of Politis and Romano (1994) we randomly draw (with replacement) a new sample
of log-returns {(b)ir(t), τ(1), ..., τ(T), i=1, 2, ..., N}, where τ(t) was the new time index, a random
draw from the original set {1, ..., T}. In other words, τ(t) was a randomised index common
was the indicator for the bootstrap number which ran from b=1 to B. We used 10,000 bootstrap
Next, for each draw b we conducted N regression to obtain βiJAN or βiMAY coefficients
(dependant on the seasonality thus researched) for each of the N assets. Subsequently, we
calculated a simple arithmetic mean of the N regression coefficients and repeated the procedure
B=10,000 times. As the theories investigated in this study imply that the returns in the examined
periods could be abnormally low, our null hypothesis in the BT tests assumed that the average
regression coefficient βiJAN/MAY be higher or equal to zero with the alternative assuming the
contrary:
To obtain the p-value for the test we counted a number of cases for which βiJAN/MAY < 0
Results
Table 3 reports the βiJAN regression coefficients across the 25 countries and 5 maturity
buckets. Clearly, the results indicate no January effect in government bond returns. The βiJAN
usually insignificantly departed from zero while their absolute value was relatively low. As this
observation holds true across all the maturity buckets, our results contradict the observations of
Clayton, Delozier, and Ehrhardt (1989), who found the January returns to be abnormally high
Note. Table reports the results of the examination of the January effect in international government bond returns.
The columns' headlines indicate maturities of bonds within given buckets expressed in years (Y): 1-3, 3-5, 5-7, 7-
10, and beyond 10 years. βiJAN is the regression coefficient from the equation (1) and t-stat is the corresponding
Newey-West (1987) adjusted t-statistics. Asterisks *, **, and *** indicate values significantly different from zero
at the 10%, 5%, and 1% level, respectively.
displayed abnormal returns. For example, in Italy the January payoffs were abnormally high
across all the maturities, outperforming the other months by 0.39-1.29%. The exceptionally
high returns at the beginning of the year were also observable in some maturities in: Belgium,
Ireland, New Zealand, Norway, and Spain. However, the return patterns in these cases were less
consistent than in Italy. Interestingly, while the abnormal returns in these instances were indeed
significant, they were also positive: in other words, these patterns were analogous to the January
effect in equity markets and contradicted the hypotheses in this study. Importantly, the few cases
of abnormal returns across 125 bond buckets can constitute no basis for reliable and robust
statistical inferences, so the results presented in Table 3 fail to support the "parking-the-
proceeds" hypothesis.
anomaly. Differently than in Table 3, many of the coefficients are negative. The majority,
Note. Table reports the results of the examination of the January effect in international government bond returns.
The columns' headlines indicate maturities of bonds within given buckets expressed in years (Y): 1-3, 3-5, 5-7, 7-
10, and over 10 years. βiMAY is the regression coefficient from the equation (2) and t-stat is the corresponding
Newey-West (1987) adjusted t-statistics. Asterisks *, **, and *** indicate values significantly different from zero
at the 10%, 5%, and 1% level, respectively.
Interestingly, we identified notable exceptions: Canada and the United States. In the two
North American Countries the bond markets significantly underperformed within the periods
from November to April. The negative returns were generally higher in absolute terms for long-
term bonds than for short term bonds and ranged from -0.12% (1Y-3Y) to -0.53% (10Y+) in
Canada, and from -0.13% (1Y-3Y) to -0.53% (10Y+) in the United States. The returns in
Australia proved also negative in four out of five maturity baskets ranging from -0.21% to -
0.47%. In the three countries, the outcomes were consistent with the implications of the
"parking-the-proceeds" hypothesis' implications for the "sell-in-May" anomaly. Still, these are
only 3 out of 25 markets; in the overhelming majority of the examined markets we detected no
reliable and consistent pattern of underperformance. Thus, the empirical data fail to support the
"parking-the-proceeds" hypothesis.
Table 5. Summary Statistics for the Seasonal Effects in Government Bond Returns
The table reports summary results of the examinations of the January effect (panel A) and "sell-in-May" effect
(panel B) in international government bond returns. The columns' headlines indicate maturities of the bonds within
given buckets expressed in years (Y): 1-3, 3-5, 5-7, 7-10, and over 10 years. All is a pooled sample of all of the
maturity buckets. Mean βiJAN and Mean βiMAY are mean regression coefficients from equations (1) and (2) averaged
across the 25 examined countries, and mean t-stat are means of corresponding Newey-West (1987) adjusted t-
statistics. BT are p-values from the bootstrap test for the multinational samples as described in the Methods section;
BT are expressed as percentages.
Table 5 synthesises the outcomes of Tables 3 and 4. The results of aggregate tests
confirm the initial impressions from individual countries: the empirical data fail to confirm any
regularities in government bonds. The mean βiJAN across all of the examined countries proved
positive and the BT tests' hypotheses were not rejected. Furthermore, while the average βiMAY
coefficients were negative, still all the BT tests identified no seasonal patterns within the
November-April period. To sum up, our examinations within the countries proved no
significant seasonal return pattern related to either the January effect or the "sell-in-May" effect.
continued with the investigations of the analogous phenomena in the returns on factor
portfolios. Table 6 presents the results of the examination of the January effect in the returns
on government bond strategies. Clearly, this seasonality plays an insignificant role in the
government bond factor premia. Nearly all βiJAN regression coefficients were insignificant
within all of four categories of return patterns tested: volatility, credit risk, value, and
momentum. Only a handful of individual coefficients proved significant, for example positive
coefficients in the yield-to-maturity (12) and term premium (13) patterns when using the
quartile or quintile value-weighted portfolios.. Even in these exceptional cases, however, the
seasonal pattern failed to surface under other weighting schemes. Summing up, the January
seasonality exerts no influence on the performance of the volatility, credit risk, value, and
Finally, Table 7 presents results analogous to Table 6 but considering the "sell-in-May-
April period. Indeed, we saw a few significant negative coefficients among some of the value-
weighted portfolios: beta, duration-times-yield, and 6-month moving average, but none proved
reliable and consistent across many specifications. Summing up, also the "sell-in-May"
anomaly appears to exert no influence on the performance of government bond factor premia.
Concluding Remarks
The study aimed to investigate the January and "sell-in-May" seasonal patterns in government
bond returns and factor premia. The comprehensive examination conducted within 25 country bond
markets for years 1992-2016 proved these return patterns play an insignificant role in government
bond returns. The test results proved the effects appear only in a handful of individual countries and
fail to surface in an aggregated sample. Also, neither the January nor "sell-in-May" effects was
reflected in the performance of the factor premia and the volatility, credit risk, value, and momentum
The results have implications for both academic and practitioners' purposes. Our outcomes
nonexistent in government bonds, and the January effect seems to be merely a statistical artefact. The
seasonal anomalies should neither be regarded in designing asset pricing factors for government bond
markets nor constitute the basis for market timing strategies designed for international government
bonds. The results support, however, the efficient market hypothesis postulated by Fama (1970) on
the bond market in the majority of the analysed countries. As such they reinforce the rationale for
using traditional neoclassical finance tools, at least when considering seasonal anomalies.
Further studies in could be pursued in two major directions. First, by extending the research
sample to some related asset classes, like municipal bonds and second by broadening the catalogue
of the tested seasonal anomalies to some short-term effects, like turn-of-the-month, days of the week,
or holidays anomalies.4
4
See, e.g., Ziemba (2012) or Patell and Sewell (2015) for review.
References
Asness, C.S., Ilmanen, A., Israel, R., & Moskowitz, T.J. (2015). Investing with a style. Journal of
Asness, C.S., Moskowitz, T.J., & Pedersen, L.H. (2013). Value and momentum everywhere. Journal
Asvanunt, A., & Richardson, S.A. (2016). The credit risk premium. Available at SSRN:
http://ssrn.com/abstract=2563482 or http://dx.doi.org/10.2139/ssrn.2563482.
Beekhuizen, P., Duyvesteyn, J.G., Martens, M., & Zomerdijk, C. (2016). Carry investing on the yield
http://dx.doi.org/10.2139/ssrn.2808327.
Bouman, S., & Jacobsen, B. (2002). The Halloween indicator, "Sell in May and go away": another
http://dx.doi.org/10.1257/0002828027620246
Castro, F.H., & Schabek, T. (2014). "Sell not only in May". Seasonal Effects in emerging and
http://dx.doi.org/10.2139/ssrn.2458515.
Chan, K.C., & Wu, H.K. (1993). Bond market seasonality and business cycles. International Review
Chan, K.C., & Wu, H.K. (1995). Another look on bond market seasonality: A note. Journal of
Chang, E.C., Pinegar, J.M. (1995). Return seasonality and tax-loss selling in the market for long-term
http://dx.doi.org/10.1016/0304-405x(86)90071-1
Chieffe, N., Cromwell, N.O., & Yoder, J.A. (2000). Fixed-income securities and the January effect.
Clare, A.D., & Thomas, S.H. (1992). International evidence for the predictability of bond and stock
Clayton, R., Delozier, J., & Ehrhardt, M.C. (1989). A note on January returns in the U.S. government
bond market: The term effect. Journal of Financial Services Research, 2(4), 307-318.
http://dx.doi.org/10.1007/bf00114410
Davis, J. (1994). The cross-section of realized stock returns: The pre-Compustat evidence. Journal
de Carvalho, R.L., Dugnolle, P., Lu, X., & Moulin, P. (2014). Low-risk anomalies in global fixed
income: Evidence from major broad markets. Journal of Fixed Income, 23(4), 51-70.
http://dx.doi.org/10.3905/jfi.2014.23.4.051
de Vassal, V. (1998). Time and seasonal patterns in the fixed-income markets. Journal of Fixed
Duyvesteyn, J., & Martens, M. (2014). Emerging government bond market timing. Journal of Fixed
Fama, E.F. (1970). Efficient capital markets: a review of theory and empirical work. Journal of
Frazzini, A., & Pedersen, L.H. (2014). Betting against beta. Journal of Financial Economics, 111(1),
1-25. http://dx.doi.org/10.1016/j.jfineco.2013.10.005
Haug, M., & Hirschey, M. (2006). The January effect. Financial Analyst Journal, 62(5), 78-88.
http://dx.doi.org/10.2469/faj.v62.n5.4284
Horowitz, J.L., Loughran, T., Savin, N., (2000). Three analyses of the firm size premium. Journal of
Jacobsen, B., & Zhang, C.Y. (2014). The Halloween indicator, 'Sell in May and go away': An even
http://dx.doi.org/10.2139/ssrn.2154873
Koijen, R.S.J., Moskowitz, T.J., Pedersen, L.H., & Vrugt, E.B. (2015). Carry. Fama-Miller Working
http://dx.doi.org/10.2139/ssrn.2298565.
Lavin, A.M. (2000). An empirical investigation of the persistence of stock and bond return
http://dx.doi.org/10.19030/jabr.v16i2.2037
Loughran, T. (1997). Book-to-market across firm size, exchange, and seasonality: Is there an effect?
http://dx.doi.org/10.2307/2331199
Luu, B.V., & Yu, P. (2012). Momentum in government-bond markets. Journal of Fixed Income,
Newey, W.K., West, K.D. (1987). A simple positive-definite heteroskedasticity and autocorrelation
Patel, N., & Sewell, M. (2015). Calendar anomalies: A survey of the literature. Journal of
Politis, D.N., & Romano, J.P. (1994). Large sample confidence regions based on subsamples under
http://dx.doi.org/10.1214/aos/1176325770
Ritter, J.R. (1988). The buying and selling behavior of individual investors at the turn of the year.
Rozeff, M.S., & Kinney, W.R. (1976). Capital market seasonality: The case of stock returns. Journal
Schneeweis, T., & Woolridge, J.R. (1979). Capital market seasonality: The case of bond returns.
http://dx.doi.org/10.2307/2330299
Smirlock, M. (1985). Seasonality in bond market returns. Journal of Portfolio Management, 11, 42-
44. http://dx.doi.org/10.3905/jpm.1985.409009
Smith, C..E. (2006). Seasonality in Canadian bond returns: The role of international factors.
http://dx.doi.org/10.1111/j.1936-4490.2006.tb00612.x
Smith, K.L. (2002). Government bond market seasonality, diversification, and cointegration:
6803.t01-1-00004
Wachtel, S.B. (1942). Certain observations on seasonal movements in stock prices. The Journal of
Yao, Y. (2012). Momentum, contrarian, and the January seasonality. Journal of Banking and
Zaremba, A. (2015). The January seasonality and the performance of country-level value and
http://dx.doi.org/10.12775/CJFA.2015.024
Zaremba, A. (2016). Investor sentiment, limits on arbitrage, and the performance of cross-country
http://dx.doi.org/10.1016/j.jbef.2015.11.007
Ziemba, W.T. (2012). Calendar anomalies and arbitrage. World Scientific Series in Finance.
DOI:10.1142/8467