Estimating Hedge Fund Alpha with Factor Models and Fund Panels

# (Yearly) Hedge Fund alpha


# (Yearly) Hedge Fund alpha












I am currently trying to make the best of quarantine by getting familiar with Stata by working on a little project. As the title states, I want to compare yearly alphas of Hedge Fund strategies.

I would like to do two things:

- Finding the alpha of all (i.e. not individual) Hedge Funds over the last 25 years

- Finding the yearly alpha of different Hedge Fund strategies (e.g. Emerging Market or Fixed Income arb.)

To do so I am using the Fung-Hsieh 7-factor model. It is basically a linear factor model designed for the Hedge Fund universe, where the dependent var. is the after-fee excess return and the independent vars. are the seven factors and the intercept is alpha. My data is an unbalanced panel set where each fund is identified by its ID and observations are monthly. Leaving me with around 600,000 observations in total. Now on to my problem(s): To find the "overall" alpha I have tried various regressions.

`reg return EquityMarket Size BondMarket CreditSpread PTFSBD PTFSFX PTFSCOM, vce(hc3)`

`xtreg return EquityMarket Size BondMarket CreditSpread PTFSBD PTFSFX PTFSCOM, fe vce(robust)`

While the alpha and coeff. of the factors are roughly in line with the literature my adj. R2 is substantially lower. In the literature, depending on the timeframe, adj. R2 are generally above 0.5 mine is around 0.1. The only regression that has delivered a somewhat higher R2 (0.28) is:

`bysort IDD: asreg return EquityMarket Size BondMarket CreditSpread PTFSBD PTFSFX PTFSCOM`

asreg is a user-written command, the results are the same when I regress on the fund level using the normal reg command and then average the R2 of the individual funds.

For the yearly alpha exercise, I have also tried various approaches, e.g.

`by year, sort : regress return EquityMarket Size BondMarket CreditSpread PTFSBD PTFSFX PTFSCOM, vce(hc3)`

where R2 ranges from 0.04 to 0.2.

So in short, I am certain that there is something fundamentally wrong with my approach. I have already checked my monthly factors, so I am confident the issue is not caused by wrongly calculated monthly factors. I am aware that there are more complex models/approaches out there which I am planning to tackle once I have the basics sorted out.

As this is my first post I hope I did not make any mistakes while writing up this question and would also like to thank you all in advance.

References: Factor Model: https://faculty.fuqua.duke.edu/~dah7/HFRFData.htm

Tables 7&8 is roughly what I am trying to reproduce:

Roger G. Ibbotson, Peng Chen & Kevin X. Zhu (2011) The ABCs of Hedge Funds: Alphas, Betas, and Costs, Financial Analysts Journal, 67:1, 15-25, DOI: 10.2469/ faj.v67.n1.6

## Answer by RWP - Down by the Bay (score 1)

https://quant.stackexchange.com/a/53527

Non-Specific, Meta-answer: When I build something and it doesn't work I try to simplify the problem. In this case, first, work on just estimating the market exposure and alpha net of that market exposure. This should be simple enough and you should be able to find references either in your specific paper or other papers that will let you ballpark check your result. This will help you identify any problems with your code or data. You can also export the data to excel and check your results there as well.

I can tell you what you'll likely find: (1) Some small amount of hedge funds are really worth their fees. You'll only rarely get access to these funds. (2) Some larger but still small amount of hedge funds are worth their fees. You'll have better access to these funds, but still not great access. (3) Most funds are not worth their fees. You can invest all you want with these funds. The easier it is to tell which category a fund falls into the more pronounced an example they will be of that category w.r.t. access.