When you use apprisal data, the correlation to other assets class are biased downwards compared to the true correlation, also the true variance of the asset is bisased downwards. can someone explain why the true variance is biased downwards? I thought the true variance should be biased upwards since apprisal data somewhat smoothes out the volalility.
Thats what I thought too, and I got the question wrong. Any explanations?
Appraisal data is “stale”, that is volatility in asset prices are not truly known due to the infrequent measurement of real estate values, therefore correlations are biased downwards.
BTON04 Wrote: ------------------------------------------------------- > Appraisal data is “stale”, that is volatility in > asset prices are not truly known due to the > infrequent measurement of real estate values, > therefore correlations are biased downwards. you stated the obvious, the question doesn’t pertain to correlation as I have noted above but rather to the VARIANCE of the asset.
ok you are right…but again the if you cannot truly observe the asset value because the same is not readily tradeable, the measured values as reported by appraisals will not exhibit the variance due to infrequent measurement of appraisals.
Since the appraisals are done infrequently, you will have infrequent price changes. This means you will not “see” any movements between appraisal dates. Since you only see “the smoothed” increase/decrease in price, your computed standard deviation will be lower compared to highly liquid assets
kurmanal Wrote: ------------------------------------------------------- > Since the appraisals are done infrequently, you > will have infrequent price changes. This means you > will not “see” any movements between appraisal > dates. Since you only see “the smoothed” > increase/decrease in price, your computed standard > deviation will be lower compared to highly liquid > assets exactly, but the CFA states that the TRUE variance (starndard deviation) will actually be the lower one having a downward bias. read the statement above.
think of the formula (the diversification effect) - Variance = (… + 2*w1*w2*stddev1*stddev2*correlation coefficient). Since correlation coefficient is lowered, so is the variance.
mp2438 Wrote: ------------------------------------------------------- > think of the formula (the diversification effect) > - Variance = (… + > 2*w1*w2*stddev1*stddev2*correlation coefficient). > Since correlation coefficient is lowered, so is > the variance. variance of the asset class to itself not with a another asset. that’s what it’s stating.
to me, variance is easy to get, but correlation is pretty hard to understand. I thought appraisal smooth >> higher Rsquared >> higher correlation. when data is smoothed, it is more likely develop dependent relationship with others; if volatile, highly possible the Rsquared will be low, hence lower correlation. did I miss anything?
I don’t know if this helps…but take a look at this… asset 1 (mvs’ are smoothed intra-quarter and refreshed at end of year) quarter 1 - 100 quarter 2 - 100 quarter 3 - 100 quarter 4 (end of year) - 90 asset 2 (mv’s are refreshed quarterly) quarter 1 - 100 quarter 2 - 85 quarter 3 - 105 quarter 4 (end of year) - 90 if you compare asset 1 and asset 2, their correlations will be biased downwards due to smoothing, even though they both end up at the same mv at end of the year. and asset 1 will have low variance and standard deviation than asset 2 due to smoothing… this is how I keep is straight in my mind.
bluecollar, when you say “compare asset 1 and asset 2 their correlation …” you mean the correlation between 1 and 2 or something else? biased downwards, but from what level?
I do mean the correlation between 1 and 2… biased downwards from their “real leve” or “actual level”
whystudy Wrote: ------------------------------------------------------- > When you use apprisal data, the correlation to > other assets class are biased downwards compared > to the true correlation, also the true variance of > the asset is bisased downwards. > > can someone explain why the true variance is > biased downwards? > > I thought the true variance should be biased > upwards since apprisal data somewhat smoothes out > the volalility. “true variance… is BIASED downwards” - this is how I read the statement. EDIT: By reading this statement, I interpret it as follows: Someone smooths out the returns… puts a bias on the actual picture… we only see how the price is nicely trending up… so the true variance is biased… how, you might ask? well, downward… it was supposed to be way higher than the picture shows… I think you are misinterpreting what you are reading…
haha, this is getting hillarious
I am pretty sure “true” as mentioned by the Institute, means “calculated”, because Schweser phrased it as “calculated standard deviation” is biased downwards. The actual value should not be the reference for anything biased anyway, if anything is biased, it should be the calculated value because of the smoothened data.