Topic 6: Methods for Alts Flashcards
What are the 4 key components of risk-neutral modelling?
- INFINITE sets of values (P measure prob, risk premium, recovery etc) consistent with particular derivative value
- ONE set of Q measures with inputs readily observable and easy to apply. In risk-neutral world, risk premiums = 0
- Values obtained from Q identical to no-arbitrage values in risk averse using P
- Q used in conditions in which actual Deriv. prices must match risk-neutral model prices
3 Fallacies due to Averaging Compounded Returns
Non-0 NPVs CANNOT be generated when assets are efficiently priced: e.g. leverage or inverse ETF
Rebalancing does not generate better performance due to better diversification, for efficient assets (when 0 NPV)
Which of the following most closely represents risk premiums of risk-neutral investors?
- 0%
- 100%
- -100%
0%. Risk-neutral investors do not require compensation for bearing risk. Therefore, they have zero risk premiums.
4 steps of fundamental strategy
- idea generation
- idea expression
- sizing
- trade execution
Define sentiment
Beliefs about future cash flows and risks NOT justified by objective analysis of facts.
According to Baker and Wurglur, what are the six sentiment indicators?
- discount on closed-end funds
- turnover of NYSE shares
- # of IPOs
- avg first day IPO returns
- equity shares in new issues
- dividend payment
The Golden Future Hedge Fund has conducted an internal study on the trading records of its senior traders. The trading record of Kano Gunma shows that he realized profits too quickly and held losing positions for a long time. To which of the following biases is Kano most likely subject?
Disposition effect: causes investors to realize profits too quickly and not cut losses in a timely fashion
What is Principal Component Analysis?
Linear statistical technique which identifies orthogonal (uncorrelated) statistical factors or principal components “PC” that maximise percentage of explained variation
What is factor analysis?
Identifies factors and their coefficients by optimising a model with statistical assumptions
Difference between Factor Analysis and Principal Component Analysis?
- FA makes assumptions about returns process; PCA does not require a model (simply maxes explained variance)
- FA generates different factor scores when model has different #s of factors; PCA loadings do not change when # of components change
- FA seeks factors that drive at least 2 securities; PCA can identify a factor driven almost entirely by one security
Challenge with Multiple Regression Models
Realised return may be correlated with factors
Factors may not have risk premium
CHALLENGE: selecting appropriate indepedent variable
THUS omitted factors falsely attribute to alpha = captured at intercept = inflated alpha
Add factors = higher r-square
Three factors for Fama French Model
- Market risk premium
- Size premium
- Value premium
Name the ways that data set could work against multiple regression models
- outliers
- autocorrelation
- heteroskedasticity
- multicollinearity = indep variables correlated with one another
What are the adverse effects of multicollinearity?
- slope estimates may be inaccurate
- standard error of the coefficient estimates (B) can be inflated
SOLUTION: create a new variable from 2 correlated explanatory variable
What is stepwise regression?
Finding an independent variables by keeping variables with greatest t-statistics and deleting with insignificant t-stats
What is stepwise regression used for?
Stepwise regression is used to determine the independent (explanatory) variables that should be included in a regression model. It involves adding or removing variables from a model based on their statistical significance (i.e., their t-statistics).
What is Positive conditional correlation
Positive conditional correlation refers an environment in which correlation in up markets exceeds correlation in down markets. Negative conditional correlation refers to correlations in down markets exceeding correlations in up markets.
Investors typically prefer positive conditional correlation, since this environment provides investors with greater participation in profitable opportunities in markets that rose and less participation in losses in markets that declined.
Laszlo Paul has developed a statistical method for hedge fund replication based on specialized market factors. Which of the following characteristics is this method most likely to have?
identifies factors based on how well they explain OVERALL MARKET or PARTICULAR HEDGE FUND’S return in order to replicate the hedge fund’s return based on these factors
identifies factors based on how well they explain particular hedge fund’s return in order to replicate the hedge fund’s return based on these factors
Dummy variable regression analysis is used to test
Market timing ability
Which of the following represents a way to estimate a first-order partial autocorrelation?
A.as the first factor in a principal component analysis
B.as the intercept of a multi-factor model
C.as the first variable selected in a stepwise regression
D.as the beta coefficient of the first regression factor
Partial autocorrelations (PAs) may be found as the beta coefficients of the factors in a regression model. The first-order PA corresponds to the beta of the first factor, the 2nd-order PA corresponds to the beta of the second factor, etc.
An individual hedge fund’s returns are generally best explained by which of the following?
Hedge fund’s trading style OR returns of funds using similar strategies
Trading style
Which of the following best represent characteristics of principal component analysis performed on a set of stock returns for several companies?
I. Resulting principal components are orthogonal.
II. Resulting factor loadings represent each stock’s return variance.
III. Resulting factors represents the smallest percentage of explained stock return variance.
I only
- A factor loading represents a stock’s RESPONSIVENESS to a particular factor (or principal component). Thus, factor loadings are similar to betas in regression models.
- Factors represents the greatest percentage of explained stock return variance.
Which of the following is considered an advantage of principal component analysis?
A.visualization
B.dimensionality reduction
C.factors with largest t-statistics
An advantage of principal component analysis is dimensionality reduction: it significantly reduces the dimensions of the original data set (from perhaps thousands of variables to a few), which results in a relatively small number of factors.
What is a typical crack spread
A crack spread is the spread created in commodity markets by purchasing oil futures and offsetting the position by selling gasoline and heating oil futures.
