All Other Paper Reviews

<-My Homepage

 ---------GENERAL FINANCE----------


**Rubble Logic: What Did we Learn from the Great Stock Market Bubble?

Author: Clifford S. Asness

Published: Financial Analysts Journal Nov/Dec 2005

Keywords: Bubble, dividends as corporate governance, dividend payout and earnings growth, Fed Model 

Keywords: Tech Bubble, dividends as corporate governance, high P/E ratios

Summary: The author lists several lessons learned from the bubble, about value investing, dividends, and P/E ratios. This is written in a very entertaining and non-academic style and is fun to read. This has many short lessons, with historical summary-type data to back them up. Since I'm lazy, I'll just re-write most of the headings in the paper.

Higher prices today mean lower expected returns tomorrow: Higher marker P/E's have historically (1927-2004) been followed by decades of lower return. Higher prices (P/E) today might mean permanently lower future returns (since part of the past strong return was an increasing P/E) or a short period of bad returns followed by historically normal ones.

Dividends are good for some surprising reasons: even though high dividend payouts should lead to lower earnings growth, historically companies with higher dividends payouts have had higher earnings growth, so dividends have been doubly good for investors. Earnings don't grow at a real 10% per year.

The Fed Model must be fought: the fed model says the market is cheap if its earnings yield (E/P) is higher than bond yields. This is wrong because bond yields are a nominal quantity which E/P is real, AND because historically, while past stock returns have correlated with low interest rates (as the fed model predicts), forward stock returns have roughly negatively correlated with low interest rates (that is, historically periods of low interest rate have been preceded by strong market returns and followed by weak market returns).

Value wins in the long run (despite a dip in 1996-1999, HML long-short value performance is on track). Wall Street and the Media aren't looking out for you ("stocks will return an expected ~6% over the long haul and we and everyone else have no idea what they will do in the next weeks or months" doesn't make good TV). Pro-forma accounting statements aren't reliable, the public is full of bored innumerate gamblers, and do-it-yourself trading is a bad idea. International diversification is not a waste of time. Long-term Investors shouldn't be 100% in stocks (a levered position in a stock and bond is better).  


**A Very Long-Term Buy-and-Hold Portfolio

Authors: Sudhir Nanda and Donald J. Peter.

Published: Journal of Portfolio Management 2006

The authors argue that buy-and-hold's tax advantages are hard to beat, and consider a buy-and-hold (BH) portfolio from Jan. 1961- Dec.2004 (originally with 1116 stocks, down to 209).  This has slightly higher return (10.8%) than the CRSP mkt. wt. benchmark (10.7%). An equal-weight portfolio slightly underperform the CRSP mkt. wt. benchmark.  They do various robustness checks, finding that BH has a 4.6% tracking error versus the S&P, looking at sensitivity to period and sector allocations, and find buy-and-hold generally does as well as the market benchmark.  Unfortunately they do not compare volatility or model tax loss from either portfolio.


**The Currency Hedging Decision: A Search for Synthesis in Asset Allocation

Author: Gary L. Gastineau

Published: Financial Analysts Journal 1995

Keywords: Currency hedging, forward rate bias, benefits of diversification on compound geometric return

The author begins by stating two extreme views on currency hedging in asset allocation.  The first, by Perold and Shulman, is that long-run expected currency return is zero, so any currency exposure (un-hedged position) should be an active decision. The other extreme, argue by Froot, is that currency returns mean-revert over an ~8 year period, so in the long-run hedging is uneeded and adds to cost.  The middle ground is stated in Black (1989), who derives from the global CAPM that there is a single universal currency hedge ratio, and he estimates this as between 30% and 70% currency exposure.

 The author argues that a 50/50 hedged/unhedged provides a good benchmark allocation. He argues the currency decision should be integrated into asset allocation, not let to the end.  He points out that the risk-reduction benefits of diversification can increase geometric mean return even without increasing arithmetic mean return.  He suggests that a forward rate bias exists, where high interest rate currencies don't depreciate as much as "they should", allowing the holder of high interest-rate currencies to earn excess return and allowing active currency management.  Lastly he argues, I believe, that breaking foreign and domestic (equity) returns into money-market and risk-premium returns can improve return attribution.


**Is Fixed-Weight Asset Allocation Really Better?

Authors: Bala Arshanapalli, T> Daniel Coggin, and William Nelson

Published: Journal of Portfolio Management 2001

The authors compare a fixed-weight asset allocation from studies by Brinson, Singer, and Beebower (1991) and Blake, Lehman, and Timmerman (1999) with various other strategies.  Considering only cash, bonds, and stocks as assets, they use Markowitz mean-variance optimal portfolios (using past five-year returns), the Kidder-Peabody model (which values the stocks and bonds markets using a dividends discount model and rates their relative attractiveness), and the recommended allocation of eight wall street brokers. The Markowitz and Kidder-Peabody models have rapidly shifting allocations over time.  The Markowitz model has the lowest returns (not surprising since past five-year returns aren't a good proxy for future returns) and Kidder-Peabody easily beats the fixed-weight schemes. Roughly half of the Wall St. Brokers beat the fixed blend, by returns or Sharpe ratio.


**Information Horizon, Portfolio Turnover, and Optimal Alpha Models

Authors: Edward Qian, Eric H Sorenson, and Ronald Hua

Published: The Journal of Portfolio Management 2007

Keywords: portfolio optimization, turnover, factor laoding autocorrelation, information horizon

The authors explore the idea of information horizon: that some factors predictive power that diminishes over time quickly, while others dimish slowly.  Their example is a momentum factor, with an info. ratio that declines quickly, and a value factor, which is more slowly declining.  The authors then spend a lot of time deriving an estimate of turnover based on how quickly factor loadings change (factor autocorrelation) and look at optimizations that include an estimate of trading costs based on this forecast turnover.  The paper is well written and interesting, but I wonder if forecasting factor autocorrelation and then using an equation to forecast turnover is much more accurate than simply backtesting to see how empirical turnover varies.


**Demographics and Capital Market Returns

Author: Robert Arnott and Anne Casscells

Published: Financial Analysts Journal 2003

This looks at how rising numbers of retirees may affect market returns, and considers possible outcomes such as later retirement, higher savings or taxes, or imigration/emigration.  The paper is quite interesting and notes that financial solutions (more savings or taxes) are unlikely to work unless real variables (i.e. the number of workers in industries serving retirees) change, since macroeconomically the relative wealth of retirees vs workers is irrelavant: what matters is how goods / services are allocated among the groups.  i.e. in aggegate (oversimplifying), doubling retirement savings could simply double prices for retirement goods, leaving real allocations unchanged.


**Portfolio Optimization: Part 1 - Unconstrained Portfolios

Author: John Norstad

Online: As a PDF

Keywords: portfolio optimization, lagrange multipliers, CAPM, mean-variance optimization.

This unpublished paper has a nice introductory-level derivation of mean-variance portfolio optimization, including Lagrange multipliers.  It includes a derivation of the two-fund theorem and the capital asset pricing model (CAPM).  It's very easy to inderstand and has many examples.


**Parameter Optimization Techniques, Optimization Frequency, and Portfolio Return Enhancement

Author: Glen. A. Larsen and Bruce G. Resnick.

Published: Journal of Portfolio Management 2001

Keywords: Shrinkage Estimators, Bayes-Stein method

The authors look at different methods of estimating the expected return of a portfolio for input to portfolio optimization.  They consider (1) Estimating all portfolios to have the same expected return...then the portfolio output will be the min. variance portfolio (2) Estimating a portfolio's expected return to be its past return (3) Using a Bayes-Stein estimator that each portfolios return is a linear. combo of its past return and the mean portfolio return, where a Bayesian method determines the weight. They also consider 3, 6, and 12-month rebalancing periods.  As input, they use 10 CRSP market-cap-sorted portfolios.  There is surprisingly little difference in returns and Sharpe ratios for the three methods, but methods (2) and (3), which use past return information, tend to have slightly higher returns at roughly the same Sharpe ratio.


**Asset Allocation without Unobservable Parameters

 Author: Michael Stutzer

Published: Financial Analysts Journal Sep/Oct 2004

Keywords: portfolio optimization, asset allocation, rebalancing, mean-variance

References: Continually rebalanced investment strategies, JPM 1991 Mark Rubinstein

The author describes the quantitative route to investor specific advice as (1) Choose criteria to optimize (2) estiamte market parameter (3) Get investor-specific input (ex: risk tolerance) and optimize.  For example, in theory we commonly optimize: E(r) - k*Var(r). But the author argues convincingly that investor input on risk tolerance is very hard to obtain, but the portfolio results are very dependent on the risk tolerance.  He gives an example with CRRA: constant relative risk aversion, and also argues that investors may not even exhibit CRRA.  He argues instead for minimizing long-run probability of falling below some specific return target, and cites a reference that this is equivelent to an optimization where risk-aversion is allowed to vary with portfolio return.  He also shows that many traditional arguments against optimizing short-fall probability are overstated or don't apply to this situation.


**Wilmott-Delta Hedging

SlideShow titled "Which Free Lunch would you like today, sir"

Key Words: Options, Delta hedging, volatility

Online: (PDF, may need login)

Folder: General Finance

Covers delta hedging using deta calculated from either implied or "true" (forecasted correct) volatility, and difference in payoffs. Shows that Delta hedging using true volatility gives high constant path-independent payoff but with mark-to-market fluctuations. However, delta hedging using implied volatility gives path dependent variable payoff but no mark-to-market losses. SlideShow is easy to follow, despite many long formulas which are derived in the appendix.


**The Relentless Rules of Humble Arithmatic

Author: John C. Bogle (Vanguard Founder)

Published: Financial Analysts Journal 2005

Keywords: Investment Costs, Cost Matter Hypothesis

Bogle argues that active investment costs are too high, leading money managers to take more of the gains than their investors.  He argues against large financial conglomerates, showing that companies which manage less than 15 funds perform better and privately-held companies perform better


**Do Hedge Funds Hedge

Authors: Clifford Asness, Robert Krail, and John Liew

Published: Financial Analysts Journal 2001

Keywords: Hedge Fund beta, up-market down-market beta, lagged betas, stale prices, hedge Sharpe Ratio

The authors use the CSFB Hedge fund index (net of fees) from 1994-2000 to analyze betas of arious HF style indexes.  They find that simple regression estimates imply low betas, but argue this may be incorrect since many HF hold illiquid securities with prices that may no be accurately market to market.  They Adjust these betas by computing betas bersus the market return, and values of the market return lagged by one, two, and three months: summing this normal and the three lagged betas presumably gives a better measure of true market exposure.  This increases the estimated index beta from ~0.4 to ~0.8, and the betas on lagged market returns tend to be statistically significant. They find that this increase in betas tends to move alphas from stat. significantly positive to slightly negative. It also makes Hedges Sharpe ratios negative, where Hedged Sharpe ratio is the Sharpe ratio of a portfolio long an asset and short beta units of the market index: this is the same as the information ratio. They also find higher HF betas in down-market (both traditional and lagged betas) than up markets. They conclude that HF seem to price their securities at a lag, and when accounting for this many HF style produced performance nowhere near as strong as it seems with un-adjusted betas.


**Consistent Alpha Generation through Structure

Author: William H Gross (PIMCO chief investment officer)

Published: Financial Analysts Journal 2005

Argues that alpha can be generated (at least in fixed income) through the "structural" portfolio composition: that is, the long-term makeup of the portfolio (ignoring short-term 3-5 year forecasts).  As examples of ways to do this, he argues that it's profitable to over-weight mortgages (since homeowners over-value their pre-payment options by paying higher interest rates) and investing in the 12-18 month part of the yield curve (since short-term investors' demand for over-night liquidity drives down short-end yields). He also argues for selling puts and calls on treasury futures and swaps (without leverage). Structural positions like these, he claims, each add like 10 to 20 basis points annually over a long time period.


**Optimal Asset Location and Allocation with Taxable and Tax-Deferred Investing

Authors: Robert M Dammon, Chester S Spatt, and Harold H Zhang

Published: The Journal of FInance 2004

Keywords: Taxes and asset allocation

Summary: This paper models the asset allocation decision where investors choose between equity and bonds, and between a taxable and tax-deferred account.  It concludes that in almost all cases taxable bonds should be in a tax-deferred account and equity in the taxable account, and points out that this is NOT how actual investors allocate their wealth.

The authors use a simple model with different tax rates on dividends/coupons versus capital gains, and look at optimal investor asset allocation decisions at different ages.  Their general result is that the highest yield assets (yield is how much as asset pays out, so it's some sum of coupons, dividends, and cap. gains distributions) should be in the tax-deferred account, and lower-yield assets in the taxable account, and then borrowing in the taxable account can be used to strike the optimal risk-return tradeoff. Even active mutual funds have much lower yields than bonds, and so should be held in taxable accounts more so than bonds. Since gains and losses on equity can be realized or deferred to help the investor avoid taxes, equity is especially useful in taxable accounts. They show that, even in the face of consumption shocks and a penalty for withdrawing from tax-deferred accounts, it is seldom optimal to hold bonds in the taxable account.


**Risk Management for Hedge Funds: Introduction and Overview

Author: Andrew W Lo

Published: Financial Analysts Journal 2001

Keywords: Hedge Funds, risk management, assymetric beta (up beta, down beta)

References: credit / liquidity risk modelling (Bookstabber 1999, Kao 2000)

Lo begins by noting that hedge funds (HF) and institutional investors have very different goals: HF often focus more on secrecy and absolute return, and less on risk. He has a one-page table showing how decreasing risk (i.e. truncating the bottom of a log-normal distribution) can drastically increase expected return, though of course this assumes this risk reduction is free (e.g. In a distribution with E(ret)=10%, vol=50%, preventing returns below -10% increases E(ret) to 18%).

With respect to risk, he argues that VAR (value at risk, like the most you might lose with 5% probability) is extremely difficult to estimate for HF because of such a large variety of return sources (commodity and stat. arb. are very different). Lo suggests that conditional VAR (i.e. VAR when mkt drops 10%, VAR when MKT flat, ...) is more useful.  he mentions large survivorship bias in HF databases.  He argues for dynamic risk exposure measures, saying that simple measures like beta, return variance, and average return are insufficient, and creates an example (using 1992-1998 data) where simply writing out-of-the-money puts generates seemingly great returns, with risk that isn't captured by standard measures.  He notes that these puts could be written synthetically (through dynamic trading) so such risk exposures may not be obvious even if we know the HF holdings (which we don't).

Next Lo discusses how any risk modelling that uses correlations among asset classes should include "phase-locking" behavior, where with some small probability all correlations become very close to one, and has some detail on a simple model of this sort.  He also finds that HF of almost all types have much higher up-market beta than down-market beta (i.e. they go up weakly with the MKT when the MKT is up but fall strongly with the MKT when the MKT falls). e.g. Emering market equity HF have an up-market beta 0f 0.16 and a down market beta of 1.49.

Lastly he says that liquidity and credit modelling are important for HF, which may be leveraged.  He uses the absolute value of the time-series autocorrelation of returns as a proxy for liquidity, saying that securities which with highly (positively or negatively) autocorrelated returns must be illiquid or else their autocorrelation would be arbitrages away.  Doing this he finds that HF tend to have much higher autocorrelation (and hence, perhaps, hold less liquid instruments) than mutual funds.  He concludes by saying that skewed HF compensation schemes can create imperfect risk incentives for managers, and he calls for "a handful of risk analytics that could provide investors with a meaningful snapshot of a hedge fund's risk exposures without compromising ... proprietary information".


Surprise! Higher Dividends = Higher Earnings Growth

Authors: Robert D. Arnott and Clifford S. Asness

Published: Financial Analysts Journal 2003

Keywords: Dividends, earnings growth, retention rate

References: Several sources of long-term historical earnings and price information

Summary: The authors challenge the conventional idea that, for the market as a whole, a high earnings retention rate (1-Div/Earnings) leads to higher earnings growth through investment: they find the opposite. They find, for the market as a whole, D/E is very significantly positively related to future 10-year real earnings growth.

(Notation: D=dividends per share, E=earnings per share, P=price, g=earnings growth rate)

The authors decompose market returns as R=(D/E)*(E/P)+g, noting that high P/E lowers return if g is unchanged. 

They find that, using data from as far back as the late 1800s, for the market as a whole, high payout ratio (D/E) is associated with higher subsequent 10-year REAL earnings growth.  This is a very strong positive relationship, and they propose that it is either because paying out dividends prevents wasteful empire building or because dividends are sticky (managers won't cut them) so dividends will only be high if managers expect (using inside info) future earnings growth.

Finds that regressions of future 10-year real earnings growth on payout ratio (D/E) are highly significant.  Adding past 10-year real earnings growth (which appears with a negative loading...mena reversion) as an additional covariate doesn't reduce the significance of the coefficient on payout ratio much, nor does another proxy for mean reversion n real earnings growth. Also, adding a covariate relating to yield-curve slope (to explain fut. real earnings growth) does have predictive power but is subsumed by D/E.

Finally, the authors note that a large increase in share buybacks (another form of dividend) since the 1980's make recent payout ratios appear lower than they really are. They find lower earnings yield (E/P) has a significant positive relationship with future real earnigns growth (if it didn't, low E/P would directly lead to lower return by the equation at the top) but it is less of a significant predictor than D/E


Zembia's Wilmott Articles on Hedge Funds

**Hedge Fund Concepts and a Typical Trade, Gambling and Investment Hedge Fund concepts II, and Hedge Fund Risk, Disasters, and their prevention

File: Hedge Fund Concepts and a Typical Trade.pdf, Gambling and Invetment - Hedge Fund Concepts II.pdf, Hedge Fund Risk, Disasters, and their Prevention.pdf

Author: Bill Ziemba

Keywords: Hedge Funds, Trading strategies

Folder: General Finance

References: Kelly Criterion, some Stochastic programming

This series of Wilmott magazine articles details several hedge fund concepts.  The first contains an example of a hedge fund convergence trade based on mispriced options on the Nikkei index, and explains many typical hedge fund strategies and has references to several books and papers on hedge fund return distributions.  It also desribes the author's work with hedge fund managers and some of the managers' traits.  The next two articles discuss various hedge fund disasters, scenario-based risk management (and the Kelly Criterion).  Though not presenting any novel ideas, the explanations are very well written and informative

Score: 7/10

Online: (requires login BEFORE going to the URL):


**Information Ratios and batting Averages

Authors: Neil Constable and Jeremy Armitage, CFA

Published: Financial Analysts Journal 2006

Folder: General Finance

Keywords: Information Ratios, Skewness

The authors discuss information ratios (excess return, say over a benchmark, divided by excess standard deviation) and its relation to a game where a fund manager picks IID bets that win with probability p and lose with (1-p).  They use this model to argue that, for a given p, going to bat more often improves the information ratio greatly.  They also show that if two manages have the same information ratio, if one manager has a lower p but "goes to bat" more often and the other has a higher p but bats less often, the one with the lower p who bats more often has a less skewed return distribution.  The article is simple and short but somewhat interesting.

Score: 6/10


**Are Active Management Fees To High?

Author: Richard M Ennis.  Published Financial Analysts Journal 2005.

The author notes that, because of the rise of hedge funds, active management fees have been rising, despite many studies citing lack of return persistence in actively managed funds (though these studies often don't cover hedge funds). The author also argues that the probability of a decent return after fees becomes very low, even with high manager skill, at today's fee levels.


**Forecastign Exchange Rates using Cointegration Models and Intra-day Data

Adrian Trapletti, Alois Geyer, Friedrich Leisch.

Journal of Forecasting 2002

Folder: General Finance

Keywords: High-frequency time series, FX forecasting, currency forecasting.

The authors transform the underlying timescale of intra-day high frequency data to take more samples at volatile times and less at non-volatile times (and omit weekends). They claim this removes seasonality. They then find a cointegrating relationship among the serie and use this forecasts. They find that the transformed time series gives better forecasts out of sample, which are somewhat statistically significant. They show that a naive trading strategy based on these forecasts will do poorly due to transaction costs, but one using a clever stop-loss rule may have cumulative positive net returns.

Score: 6/10


**Speed of Adjsutment in U.S. Financial Markets

T. Daniel Coggin and Bala Arshanapalli

Published: The Journal of Portfolio Management 2006

Keywords: Cointegration

The authors search of a cointegration relationship in the log price series of S&P500 vs 1-mo. T-bills, CRSP1-5 vs. CRSP9-10 size deciles, FF large value vs. FF large growth, S&P500 vs. L.T. bonds in Jan. 1947 to Dec. 2003.  They use a general model of cointegration that allows for structural breaks in the level and slopes of the cointegrated series.  They find that, though the series have unit roots, the evidence for cointegration is slim.  They use this to argue that asset-class timing schemes will be difficult.

In truth, their paper only argues that using pure timeseries information and cointegration for asset-class timing is difficult, and it's a long way from closing the door on asset timing.


**Evidence of Long Memory in Short-term Interest Rates

File: Evidence of Long Memory in Short-term Interest Rates (2003).pdf

Nigel Meade and Margaret R Maier

Journal of Forecasting 2003

Keywords: Mean reversion, time series, GARCH

Folder: General Finance

Uses various alterations of general time series techniques (GARCH, ARMA) to model real interest rates in 10 countries. One model is a fractionally-integrated autoregressive moving average, another is an assymetric fractionally-integrated GARCH model (where innovations are student's T, giving kurtosis, and the backup operator is applied with a fractional power d (1-B)^d. Some interpretation is given for the meaning of this parameter d when it falls in various ranges relating to how fast shocks die out, mean reversion, etc. The conclusion was that 'The evidence for mean reversion was weak', but fractional integration led to better forecasts than a martingale (where we forecast tomorrow's I.R. to be the same as today's). I didn't think higher model complexity paid off in giving much intuition behind what's going on, and author's could have just found properties that relate to their sample.


**Revisiting Mean-Variance Optimization

Authors: Enis Uysal, Francis H Trainer Jr, and Jonathan Reiss

Published: Journal of Portfolio Management 2001

The authors look at asset allocation in bond portfolios.  Defining assets to be corporates, treasuries, and mortgages/agency securities, they find that optimizers will typically put 0% in treasuries under any reasonable assumptions about spreads and risk aversion.  They suggest that this is why bond managers often use scenario analysis (ex: rising spreads, falling spreads, ...) instead of optimization.  They suggest combining scenario analysis with optimization by 1) Creating a fixed number of scenarios (returns and variance-covariance to each asset class), each with a known probability 2) Using these to create a joint probability distribution (which will be non-normal) and optimizing (I think maximizing expected utility) over this distribution.  The paper skips the optimization details, but the focus on the core idea is quite interesting.


**Needles, Haystacks, and Hidden Factors

Author: Guy Miller

Published: Journal of Portfolio Management 2006

Keywords: Attribution, factors, statistical factors

The author discusses the search for return factors:  r = X * f + e

r is a vector of asset returns, f is a matrix of factor returns, X are exposures, and e is idiosyncratic returns. He discusses the difference between fundamental and statistical factors, where statiscal are derived from things like principle components, and derives the conditions under which statistical factors can be extracted with reasonably low error (e.g. a large time series of returns is available, a reasonably large percentage of assets are exposed to the factor, and the factor has reasonably high variance).  He notes that well constructed fundamental factors (e.g. industry, where each industry includes only a few stocks) can elude statistical techniques.

 Next he describes hybrid models, where returns are first expressed as a sum of exposures to fundamental factors and residuals, and statistical techniques are used to extract factors from the residuals. He shows that this hybrid techniques performed better in terms of bias and tracking error in matching a TOPIX index.





**Bayes Rule and FX lottery arbitrage (Wilmott)

Bayes Rule and FX lottery arbitrage (Wilmott).doc

Folder: Stats/Probability/Finance

Keywords: Probability Paradox

Online: (may need login)

Unpublished (except online at Wilmott)

Great example of probability paradox with money in an envelope, and of risk-free arbitrage opportunity if you can trade currency options and someone offers currency bets with flat payoffs. Clever.

Score: 8/10


**Iceberg Risk

Iceberg Risk.pdf

Keywords: Probability

Folder: Stats/Probability/Finance

Online: (PDF) (I apologize if this link goes down but, since this is an excerpt from a book, the author may have to remove it)

The presentation is fantastic (as a conversation). Presents probability concepts (given N uncorrelated bernoilli variables...) and results at various correlation levels. Though quite interesting, it offers little practical advice for investment managers.

Score: 5/10


**Online Data Mining for Co-evolving Time Sequences

File: MUSCLES - Online Data Mining for Co-evolving time sequences (incremental regression).pdf

Title: Online:

Keywords: Regression, Forecasting

Folder: Stats/Probability/Finance

Linear regression for data that streams in, with or without forgetting. Though the original version I read had a mistake in the equations in the appendices, after correcting that this is a clever way to efficiently do incremental regression.

Score: 7/10


----------CORPORATE FINANCE----------


**An Analytical Process for Generating the WACC Curve and Locating the Optimal Capital Structure

File: An Analytical Process for Generating the WACC curve.pdf


Published: Wilmott July 2002

Author: Ruben D Cohen (Corp. Finance, Structured Products, CitiGroup)

Keywords: WACC Curve, Corporate Finance

Folder: General Finance

Shows a simple monte-carlo method to find a firm's optimal capital struture under some ideal and slightly more realistic assumptions. Math is simple but the result is interesting.

Score: 6/10 


----------MARKET SIMULATION----------


**Penn-Lehman automated trading project

File: Kearns - Penn-Lehman automated trading project.pdf


Trading, Trading simulator

Folder: Market Simulation

Summarizes a project and recent results of the Penn-Lehman automated trading project. Short description.

Score: 6/10





**A Simple Model of Capital Market Equilibrium with Incomplete Information

Author: Robert C Merton

Published: The Journal of Finance 1987

Keywords: Return Models, incomplete information, equilibrium, return models, capm

Summary: Merton creates a single-period model where individual investors are mean-variance optimizing and all returns are driven by a single market factor (and idiosyncratic variance), but every investor is only willing to invest in a subset of stocks due to limited information.  Here there is a distribution over company cashflows at the end of the period (measured in dollars), and he derives the aggregate demand for each company (in dollars), leading to come equilibrium return.  He frames this problem from the individual investor perspective as a constrained maximization and solves it, and then aggregates results across investors to derive the market equilibrium, where all investors have the same wealth and preferences. He derives that, in this model, some stocks will have CAPM-alpha, especially ones with high idiosyncratic variance, and all stocks will have lower market values. Also, the estimated CAPM slope will be too low...that is, high beta stocks will appear to have negative alpha. The later part of the paper considers how companies should act, under the assumption that raising awareness of the firm can increase its base of investors willing to buy the firm (i.e. investors who will add the firm to their information set), and this can increase the firms share value (decreasing its return). The optimization and derivation of the market equilibrium will be interesting to some readers, but I will not repeat it all here.


**Equilibrium in a Dynamic Limit Order Market

File: Equilibrium in a Dynamic Limit Order Market.pdf

Won NYSE award (

Key Words: Market Microstructure,Price Formation, limit order book

Folder: Market Microstructure, Price Formation

Models individually-rational traders deciding if a limit or market order is best for them. Shows how mid-point of spread may not be best estimate of value and tries, using monte carlo simulation, to see how the order book changes over time. Though perhaps not particularly useful to a practitioner, the model in this paper has some interesting conclusions


**A Unified Theory of Underreaction, Momentum Trading, and overreaction in Asset Markets

Authors: Harrison Hong and Jeremy C. Stein

Published: The Journal of Finance 1999

Keywords: Momentum, Reversal, Utility modelling, equilibrium, underreaction, over-reaction

The authors model a market with two kinds of traders.  The first are news traders, who learn private news "shocks" about company valuation that disperse linearly among news traders over time. The second are momentum traders who trade conditional ONLY on past returns.  Both types of investors are risk-averse in general.

The authors find that, in any covariance-stationary equilibrium, the momentum traders will act as trend chasers. And, given any news shock at time t, there is always underreaction at first as information spreads among news traders, followed by over-reaction as momentum traders pile into the stock, which eventually returns to its correct value. Thus stocks will always exhibit short-term positive auto-correlation and long-term negative auto-correlation. Another interesting point is that early momentum traders (who buy before news has spread) give a negative externiality to later momentum traders, as early price changes from momentum traders are mis-interpreted as news by later momentum traders.

They find that adding "arbitrageurs" who trade conditional on past price changes over a longer period (i.e. who pursue reversal strategies to arbitrage away the momentum-induced over-reaction) cannot exist in equilibrium over a large range of parameter values, and if they do exist they attentuate, but do not eliminate, the momentum-induced overreaction.  Also, in their simple model slower diffusion of information among news traders leads to more momentum effects, and the authors argue that this explains an empirical result that, even holding size fixed, stocks with less analyst coverage have stronger momentum profits.

Though I admit I skipped over most of the equations where they derived their equilibria, the results are somewhat neat.