Quant Methods - Practice Exam 1 PM #30 hypothesis, question, returns, decide, period

stalkey

I e-mailed about this question as I think it is misguided or flat out wrong … agree, disagree?
A researcher has investigated the returns over the last five years to a long-short strategy based on mean reversion in equity returns volatility. His hypothesis test led to rejection of the hypothesis that abnormal (risk-adjusted) returns to the strategy over the period were less than or equal to zero at the 1% level of significance. He would most appropriate decide that:
A) his firm should employ the strategy for client accounts because the abnormal returns are positive and statistically significant.
B) while the abnormal returns are highly significant statistically, they may not be economically meaningful.
C) as long as the estimated statistical returns are greater than the transactions costs of the strategy, his firm should employ the strategy for client accounts.
I chose C, reasoning that if the risk-adjusted returns are greater than the transaction costs, the strategy would be worth implementing. Schweser’s answer is B. Here’s the explanation they give for answer B:
“There are many reasons that a statistically significant result may not be economically significant (meaningful). Besides transactions costs, we must consider the risk of the strategy as well. For example, although the mean abnormal return to the strategy over the 5-year sample period is greater than transactions costs, abnormal returns for various sub-periods may be highly variable. In this case the risk of the strategy return from month to month or quarter to quarter may be too great to make employing the strategy in client accounts economically attractive.”
So, I e-mailed arguing that since the returns were stated as being risk-adjusted in this question, the variability in returns would be accounted for. The answer I received back was unhelpful. The guy essentially said that the risk-adjustment may not have accounted for all risk. No kidding?
Am I right? If not, can you offer any clarity?

ramdabom

I agree that the explanation for the answer is off, but I still think B is the best answer.
When deciding on investments you don’t just look at returns being greater than transaction costs, you have to look at oppurtunity costs as well, i.e. you have to use NPV.

hassan

SimplyAdvised wrote:
Change the answer in “C” to this:
“as long as the estimated statistical returns are greater than the transactions costs of the strategy, his firm should CONSIDER EMPLOYING the strategy for client accounts.” and you’re correct.
Would a firm consider using a strategy like this, probably, but what if the mean return was 100% with a std of 200% (still better risk adjusted then the market, no?) - this would be a great investment if you had infinite funds but would probably not work for many people.
Anyway, there’s a sentence somewhere in the review material that says something along the lines of don’t try to outsmart the test.  You may think you could argue your answer in front of a panel of judges but you’ll never get that chance, so just pick the best answer according to what you’ve been taught.  In this case, the question is whether you recognize how stasticial methods may not have economic benefit.  I can tell B is probably the answer without even reading the question.
I did pick the answer that I thought best matched the text. The text says that statistically significant results may not be economically significant due to transaction costs or risk. How is that trying to outsmart the test? It seemed completely cut and dry.
Quote:
I agree that the explanation for the answer is off, but I still think B is the best answer.
When deciding on investments you don’t just look at returns being greater than transaction costs, you have to look at oppurtunity costs as well, i.e. you have to use NPV.
The NPV is positive and there’s nothing to suggest that the strategy is mutually exclusive.
Quote:Statisticallty significant should not mean economically meaningful.
The abnormal returns was highly significant over 5 years, but what if your client does not have a 5 year horizon, but instead a 6 month investment horizon, can you still prove that this strategy will definitely work for him?
Another way to look at this.  If it takes exactly 5 years for returns to revert back to mean values, then it will not be economically meaningful to employ such a strategy for a client that plans to only invest for the next 1.5 years.
No one would disagree with your comment about economic and statistical significance. The quesiton is where the line is. The books say that abnormal returns may not be economically significant because of transaction costs or risk. Both of these are accounted for.
There’s nothing to suggest that the investment horizon of this strategy is at all meaningful.
Thanks for the discusssion everyone.