Session 5, notes on papers:

A practitioner’s guide to arbitrage pricing theory:

- Firm attributes can be related to measure of risk, however they present 3 problems
-they are based on accounting data, this data can be generated by different rules by
companies, this is inconsistent
-even if all firms use the same rules, the reporting dates may differ, so making a time-
synchronized comparison hard
-there is no rigorous theory on how firm characteristics relate to risk

- currently 2 theories try to describe the relationship between risk and returns. The CAPM
(capital asset pricing model) and the APT (arbitrage pricing theory).
-CAPM: only one type of non-diversifiable risk influences expected returns, this risk is
market risk
-APT: more general than CAPM, accepts more risk sources

- APT and CAPM agree that a well-diversified portfolio will reduce total risk (principle of
diversification)
-Even with diversification there are still common market forces which affects the
entirety of the portfolio, these are known as systematic or pervasive risks.
-In contrary to the CAPM, APT says that systematic risk does not need to be measured
in one way (market exposure risk). There are several sources of risk which consistently impact
returns (change in confidence, interest rates, inflation, real business activity and market
index). Every stock and portfolio has exposures (betas) to these types of risks.
-the pattern of betas is called the risk exposure profile, they determine the volatility
and performance of a well-diversified portfolio

-APT has 2 postulates: (maybe question on that!)

-Excess return for any asset is calculated by: the sum over all risk factors, of the risk
exposure multiplied by the realization for that factor (actual end of period value!), plus an
error term.
-pure arbitrage profits are impossible. Calculation: sum of all betas time the price of
risk

-ways to estimate the model

-using statistical techniques (factor analysis or principal components) – good for
determining the number of factors (around 5), however to read is bad since it is a non-unique
linear combination of economical factors
- X different and well diversified portfolios can replace the factors – good for insights
and strategies
-economic theory and knowledge of financial markets –intuitive, uses economic
information (if measure with errors is shit) as stock returns

-the 5 most used factors

-Confidence risk: unanticipated changes in investors willingness to undertake
relatively risky investments
 -Time horizon risk: unanticipated changes in investors desired time to payouts
-Inflation risk
-Business cycle risk
-Market-timing risk

Questions?
-agency costs in multimanager fund performance?
-postulate 1 and 2, what is the main difference of Pk and fi(t)? Is the latter already known
while the other is estimated?

Momentum Investing and Business Cycle Risk: Evidence from Pole to Pole:

-Possible reasons for momentum:

-Data mining for momentum seems a bad explanation
-Jegadeesh and Titman used ‘out-of-sample’ simulations to evidence
momentum. Present globally and still relevant after the 1993 study.
-Behavioral patterns
-maybe related to imperfect information and the revision of expectations
when new information is found
-profits from momentum are not associated with market risk
-momentum quick dissipates after the investment period
-one-step-ahead forecasts by projecting momentum profits with lagged
macroeconomic variables, can conclude that actual profits are explained by this model.

-Study wants to analyse the relation between momentum returns and macroeconomics risk
and if the dissipation of this profit is consistent with risk or behavior based models.

Short conclusions:
-If momentum is relying on macroeconomic risk it should be largely country specific, on this
paper no relation was found between momentum and macroeconomic factors.
-Momentum does not seems to be a reward for business cycle risk
-international evidence for rapid reversal for momentum profits (maybe strategy?)

-Momentum Within and Between countries:

-Data and metholodology
-From US, shares of NYSE and AMEX available from CRSP. For non US,
Datastream for countries that have more than 50 stocks. In total 40 countries. For market
indices, value weighted market index is used.
-Time coverage changes accordingly to country. Firms per country also change
-Rank stocks based on previous returns and for the test, short sell losers and
long winners in the investment period. Periods concerned take 6 months each, each stock has
the same weight.
-Since there is not enough data rankings are only in winners and losers. To
avoid statistical problems there is a one month holding period from the ranking process and
the investment.
-Momentum profits by country and region
-Winner minus losers portfolios are highly profitable around the world (Asian markets
present the weakest profits)

-Macroeconomic Risk models and Momentum Profits

-If global momentum strategies work due to their exposure to macroeconomic risk
variables, the country-specific macroeconomic factors should also have an effect.
-2 approaches used conditional and unconditional

-Unconditional tests:
-Chen the first to propose macroeconomic factors with might affect stock
returns (with the use of FF methodology). These are:
-Unexpected inflation (UI)
-Changes in expected inflation (DEI)
-Term spread (UTS)
-Changes in industrial production (MP)
-Default risk premium (URP)
-While this model has been (superado) by FF 3factor model, it is a good starting
point.
-In this paper they make a regression for the momentum profits with UI, DEI,
UTS and MP. If the factors actually explain the momentum profits the difference between the
actual and expected should be 0.
-They cannot be explained by those factors 

-Risk and behavioral explanations of momentum

-Most models account for an initial under reaction followed by return reversals.
Models show the price correction stage where winners become losers and losers winners.
(behavioral)
-stocks high past realized returns have high unconditional expected returns. Since the
nature of these expected returns do not change over time, momentum should persist. Titman
and Jegadeesh finds evidences of reversals in US data. Chorida and Shivakumar check that
profits can dissipate but not reverse.
-momentum profits exist due to persistent systematic risk in a firm’s portfolio of
projects, as those assets depreciate momentum will decrease
-momentum can occur due to growth rate shocks. High past realized return firm is
more likely to have a high growth rate.

Conclusion:
-momentum profits are large and have a weak co-movement between countries. If
momentum is driven by risk it is country specific.
-macroeconomic factors do not exhibit significance
-stocks with high predicted returns do not actually earn higher return than lower
predicted stocks. But winner stocks gain more than losers (?)
-momentum does not depend on the macroeconomic state
-momentum reverses after the investment period and become negative
The risk and return from factors:
-asset returns are driven by a small set of variables or factors. These factors are proxys
to underlying forces that affect asset prices (priced risk). Factor models are often used to
predict future returns and the cost of capital.
-risk management uses these factor models. It is hard to consistently outperform the
market, so we should have at least a portfolio with desired risk characteristics.
-this paper construct empirical proxies for the factors underlying stock returns.
Develop a set of observable variable that capture the systematic movements in returns.
-paper will selected factors by evaluating them separately, factors may appear
unimportant by themselves but can have a joint significance (treat those with suspicion and
test further). Data snooping maybe be present due to the fact that the factors are drawn from
a list, so they will also be tested in different markets. Measure characteristics in a period of
time and test whether they are relevant in a period x.
-the paper also follow FF methodology and go long on high level proxies and short on
low, to mimic the behavior of the factors.
-There is a small set of factor-mimicking portfolios that succeed in capturing the
covariation in returns. Market factor plays a big role, but so does size, past return, book-to-
market and dividend yield. For macro economical variables default and term premium explain
covariation well enough…
-Used CRSP on NYSE and AMEX, from 1968 till 1993
-Candidates for factors are factors used in previous studies. The factor are based on
accounting characteristics, past return, macroeconomical, principal component analysis and
the return on the market index
-mimicking portfolios
-sort all stocks by their attributes and assign them to a portfolio based on their
ranks.
-if the returns of several factors are highly correlated they might be picking the
same influences
-fundamental attributes tend to be correlated
-technical, macroeconomic attributes are not very correlated
-fundamental attributes also capture the covariation of returns (how return behaves
on each group)
-for momentum, stocks that have experienced certain levels of return in the past tend
to behave similarly
-for macro it seems to be bad (with the exceptions of default and term premium)
-statistical factors seem to be efficient but with moderation
-paper also used a multivariate approach to see the importance of many factors

Commonality in the determinant of expected results:

-Relative stocks can possibly be predictable with factors that are not consistent with
traditional financial theory. Behavioral and size and accounting numbers.
-some believe that is due to bias. Survivorship bias can alter the predictive power and
data snooping may arise as well.
-others believe that while claims might be exaggerated they hold some truth
-differences are based in risk
-differences are due to bias in pricing