This is a copy, saved in Word 1997-2003 format, of my research paper that was submitted to and accepted by the graduate school for the Master of Science in Economics. It pertains to the hedge fund market using proprietary data from hedgefund.net.
1. AN ABSTRACT OF THE RESEARCH PAPER OF
Mark Hoaglund, for the Master of Science degree in economics, presented on August 14,
2008, at Southern Illinois University at Carbondale.
TITLE: Measuring the Performance of the Hedge Fund Market
MAJOR PROFESSOR: Scott Gilbert
The objective of this study was to determine some of the characteristics of the hedge
fund market and to compare the returns of the hedge fund market to the S&P 500 by
computing various statistical measures of performance for a representative sample of hedge
funds and imparting meaning to the results. In order to achieve the objective, Capital Asset
Pricing Models, polynomial regressions, variances, correlations, and mean averages were
computed and the results were analyzed. Finally, graphs were generated as tests for
heteroscedasticity and normality in the CAPM regressions, and plausible interpretive
meaning was suggested. The collective statistical analysis concluded that the performance
of hedge funds exceeds the market impressively. Specifically, the hedge fund market was
found to be far less volatile and more profitable than the S&P 500. Moreover, those
particular funds - as distinguished from the overall hedge fund market - with higher Sharpe
Ratios were found to be both less volatile and more profitable than the S&P 500. Thus,
within the hedge fund market, investment alternatives exist which are characterized by an
overall improvement to the index fund.
i
2. ACKNOWLEDGEMENTS
I’d like to thank Professor Scott Gilbert for helping me throughout the process of
developing this study.
ii
3. TABLE OF CONTENTS
ABSTRACT ……………………………………………………………..i
ACKNOWLEDGEMENTS ….......................................................................................ii
LIST OF TABLES ……………………………………………………………iv
LIST OF FIGURES …......................................................................................vi
TEXT ……………………………………………………………..1
REFERENCES …………………………………………………………..123
VITA …………………………………………………………..124
iii
13. Introduction
The purpose of the following study was to examine various measures of performance of
the hedge fund market, to compare the hedge fund market to the broader stock market by
way of the S&P 500 index, and to determine the implications of the hedge fund market
performance from the perspective of considering all the investigative results collectively.
The source of data used in the analysis of hedge funds was www.hedgefund.net which is a
service owned by Channel Capital Group Incorporated that provides hedge fund news and
proprietary performance data on approximately 8000 hedge funds.1 The hedge fund data
were drawn by conducting a search for funds according to the Sharpe Ratio in descending
order and then selecting the performance data from 30 hedge funds using an algorithm. By
arranging the funds in terms of the Sharpe Ratio, a sample of data more representative of
the overall hedge fund market was obtained because the data accounted better for the full
spectrum of both risk and return of the funds. Many interesting statistics began to emerge
once the data was arranged in Excel and analyzed.
The literature contained information that shared a complementary relationship with the
findings of this study, but also, that information yielded some cautionary reservations that
must be noted with respect to this study’s performance findings of hedge funds. In an
article by John Morgan on July 7, 20082, a warning was issued that the Securities and
Exchange Commission (SEC) is poised to initiate tighter regulation of the hedge fund
market depending on the political persuasions of those elected in the impending
presidential election. If these regulatory prospects materialize, then access might be further
restricted to investors, and fewer funds may form as a result of an inability of
xiii
14. smaller firms to raise capital. Perhaps the lack of smaller, unstable firms might actually
improve the statistical performance results, such as those that are found in this study,
because there would be fewer firms that collapse and pull the performance data of the
hedge fund market down. However, an October 2004 publication by Burton G. Malkiel3
extensively studied many ways that hedge fund performance data artificially inflate the true
returns of the hedge fund market. For example, hedge funds that are about to close stop
reporting their performance data during the last months of their existence, and because
hedge funds, unlike mutual funds, do not have to report their performance data to the SEC,
a hedge fund only begins offering its data to a database when the fund has established some
sustainable measure of success so that the initial performance remains unreported.
Nevertheless, if a hedge fund were to be chosen judiciously, such as the selection of one
with low volatility and a proven track record, then surely the integrity of the results will be
intact since the investor would not have to be as concerned about the hedge fund folding.
An additional concern is also noted in a September 11, 20064 article by Pascal Botteron
regarding the inflated perception of hedge fund performance. Namely, the fact that hedge
funds tend to have low volatility is only true insofar as the fund itself is solvent and viable.
For example, the volatility of stocks in a company reflects broadly disseminated reports
about the welfare of the company itself, but a hedge fund is not required to produce such
information, so an imperative for wise investment is the process of thoroughly vetting a
fund. These reservations about the hedge fund market performance must be taken into
context and temper any understanding about the results.
xiv
15. Models and Variables
Employed in the study of the hedge fund market were a number of statistical variables
and models which will be defined and explained next.
E ( Ri ) − Rf
The Sharp Ratio, mentioned in the introduction, is defined as where E(Ri) is
σ
the expected return of fund i, Rf is the risk free rate of return as measured by treasury
bonds, and σ is the standard deviation of the excess return as given by the entire numerator.
The Sharpe Ratio is considered a measure of the tradeoff between excess return and risk
from volatility.
The variance is a measure of the spread of the values of a random variable around the
expected value. The variance can be defined, in its most abstract sense, as
var(X) = E(X – μ)2 where X is a random variable and μ = E(X), the expected value of X.
The coefficient of correlation is a measure of the degree of association between two
variables. The coefficient always lies in the interval [-1, 1] where a high positive value
means that the two variables move closely together whereas a low negative value means
that the two variables move in opposition to each other. The coefficient of correlation is
defined differently for a population of data and a sample of data.
xv
16. 2
The definition of the population coefficient of correlation in its most abstract is
ρ= where X and Y are random variables and the rho values in the denominator
are their respective population standard deviations.
The definition in its most abstract form of the sample coefficient of correlation is
r= where X and Y are random variables and the s values are their sample
standard deviations. However the definition in a form best suited for interpretation in
terms of simple regression is
r= where X and Y are random variables and n
is the number of pairs observed.
A point of clarification must be addressed in preparation for the body of the study. In
conducting the analysis of the hedge fund market, the tables for the coefficient of
correlation values were computed for a population of data in Excel because the only Excel
function available to compute correlations uses the formula for populations of data. The
only difference between the correlation formulas for populations and samples of data is that
the sample standard deviation is divided by n-1 whereas the population standard deviation
is divided by n. Consequently, the denominator of the population correlation is smaller
than the denominator of the sample correlation, so the population correlation is larger than
the sample correlation when both the sample correlation and population correlation are
applied to the same set of data. In truth, the populations of data were known in this study,
17. 3
but these populations were often treated as samples in order to project future trends, so
whether the population correlation or sample correlation is more desirable is a matter of
interpretation. Also, the reader must know that the regressions are based on the sample
correlation formula when the discussion about the r2 = R2 values is encountered later in the
study.
The coefficient of determination, r2, is a measure of how well a regression line fits the
data. In other words, the coefficient measures the percentage of the regression that can be
explained by the regression where the remaining percentage can only be accounted for by
random error. In regression involving more than one explanatory variable, that is, in
multiple regression, the term used by convention for the coefficient of determination is R2,
and in regression involving only one explanatory variable, that is, in simple regression, the
term used is r2. However, R2 is often used interchangeably for both simple regression and
multiple regression. Since Excel used the R2 term for the simple regressions discussed
later, the reader must be aware that r2 = R2.
Average Returns
For both the S&P 500 and the individual hedge funds, each month of percentage returns
was annualized by multiplying each monthly return by 12. For the period of January, 1995
– April, 2008, the mean average of the annualized monthly returns for the S&P 500 index
was 9.21515%. The mean average of the annualized monthly returns for each hedge fund
was obtained similarly, but care must be taken to note that many of the hedge funds did not
span the same number of months as the time period stated above that was chosen for the
S&P 500. Regardless, when these averages for the individual hedge funds were themselves
18. 4
averaged, the result was 13.11473%, which is substantially higher than the return for the
S&P 500. Moreover, the hedge fund performance data was also approached somewhat
differently by first averaging, for any given month, across all 30 hedge funds so that, for
example, in September of 1995, the average annualized monthly performance across all the
hedge funds was 37.2%. When these monthly annualized averages were themselves
averaged, the result was 16.9934%, which was even higher than the 13.114% figure.
Therefore, the average returns of the hedge fund market yielded much higher returns than
the general stock market.
Correlations
The correlations between the annualized S&P 500 monthly market returns and each of
the 30 hedge fund monthly performances were computed to determine how closely hedge
fund investments behave like the market. The correlations are shown below in descending
order of the Sharpe Ratio as explained in the introduction.
Table 1. Fund Correlations With the S&P 500__________________________________
Fund #1 -0.173464524 Fund #2 0.649193451
Fund #3 -0.044247993 Fund #4 -0.0714241
Fund #5 -0.207737278 Fund #6 0.497261555
Fund #7 0.414375015 Fund #8 0.470509021
Fund #9 0.253269275 Fund #10 0.486760127
Fund #11 0.601120061 Fund #12 -0.190548726
19. 5
Fund #13 0.407577199 Fund #14 0.44922268
Fund #15 0.570891982 Fund #16 0.620327411
Fund #17 0.298016148 Fund #18 0.45629676
Fund #19 0.262008673 Fund #20 0.409339549
Fund #21 0.781050409 Fund #22 0.514196748
Fund #23 0.744336757 Fund #24 0.202843746
Fund #25 0.421641744 Fund #26 -0.585642472
Fund #27 0.582170503 Fund#28 0.35064364
Fund #29 0.65484412 Fund #30 -0.083269169
________________________________________________________________________
Upon inspection, the only detectable pattern in the behavior of the correlations is that, as
the fund number increases, that is, as the Sharpe Ratio decreases, the correlation between
the given fund and the market tends to grow larger. The increase in the correlation could
indicate that many of the fund selections, especially those with decreased Sharpe Ratios
that more closely resemble the volatility of the stock market, might be characterized by
investments intentionally designed to mimic the behavior of the market. In fact, the time
plots comparing the excess market returns with the excess fund returns corroborate the
suspicion that many of the selected funds were designed thusly. Consider the following
21. 7
Figure 2. Fund 6 Performance Comparison_____________________________________
Figure 3. Fund 10 Performance Comparison____________________________________
22. 8
Figure 4. Fund 16 Performance Comparison____________________________________
Figure 5. Fund 23 Performance Comparison____________________________________
23. 9
Figure 6. Fund 29 Performance Comparison____________________________________
Similarly, many of the funds appear to move negatively with the market by construct, and
the remainder, of course, appear to move neither with the market nor against the market,
and there are a significant number of these graphs seemingly unconnected to the market
movement in the 30 funds selected. The reader can observe the graphs for himself on page
74.
Variances
Interestingly, of all the first 10 hedge funds, the average annualized monthly return
exceeded that of the market, yet, as the reader can quickly verify from the time plots, the
variances are extremely small for most of the first 10 hedge funds compared to the variance
of the market. Thus, the hedge fund investments with high Sharpe Ratios offered both
24. 10
exceptionally-lower risk and higher returns than the market. Consider the raw variance
data for the first 10 hedge funds and the average hedge fund:
Table 2. Fund Variance and Average Return____________________________________
Market Variance: 2415.999957 Average Market Return: 9.21525
Ave. Fund Variance: 753.2803166 Ave. Fund Return: 16.99337751
Fund #1 Variance: 11.70683544 Average Return: 8.890983447
Fund #2 Variance: 726.1629818 Average Return: 26.14545455
Fund #3 Variance: 51.82846841 Average Return: 11.27661972
Fund #4 Variance: 453.5904889 Average Return: 14.82
Fund #5 Variance: 938.9087074 Average Return: 17.65830508
Fund #6 Variance: 1524.90216 Average Return: 19.965
Fund #7 Variance: 363.6100871 Average Return: 11.8096
Fund #8 Variance: 653.8928485 Average Return: 13.65818182
Fund #9 Variance: 1908.131577 Average Return: 19.17795918
Fund #10 Variance: 1129.798794 Average Return: 15.16941176
________________________________________________________________________
Notice that for the average over the entire Sharpe Ratio spectrum of funds, the variance is
only 753 compared to 2415 for the S&P 500. Such a comparatively low variance
25. 11
reinforces the position that the entire hedge fund market, even when fledgling hedge funds
with low Sharpe Ratios are included in the analysis, remains far less volatile than the stock
market. Another significant characteristic of this data is that, despite the low variability
compared to the market, the average annualized monthly return for these funds with the
highest Sharpe Ratios are typically higher than those latter 20 with the lower Sharpe
Ratios. Although the fact that all but fund one of the top ten Sharpe Ratio funds exceeded
the average market return of 9.21525 could be the result of a coincidental selection of
funds, a trend seems more likely that most funds in the market with favorable Sharpe
Ratios do not merely compromise high returns with excessively low volatility. Hence, the
evidence supports the hypothesis that the hedge fund market in general forms a powerful
apparatus for generating inordinate returns.
26. 12
Regressions
Regressions of the excess annualized monthly fund returns on the excess annualized
monthly market returns were performed for all 30 hedge funds with the intention of
examining the results for the overall volatility of the hedge fund market as measured
against the stock market and the degree to which the overall hedge fund market moves with
the stock market when, in fact, the hedge fund market actually does move with the stock
market.
The regressions revealed that the volatility of the hedge fund market and the degree to
which the hedge fund market moves with the S&P 500 depend on the perspective from
which the regression results are considered. By averaging the excess returns across each
month and then regressing those average monthly returns on the excess S&P 500 returns,
results were determined for the general hedge fund market. Consider the graph of that
regression as shown below as an overview of the data.
27. 13
Figure 7. Regression of the Average Fund Return on the S&P 500__________________
The regression equation demonstrates that the degree of movement in the hedge fund
market is not very responsive to the S&P 500. Specifically, an increase or decrease in S&P
500 returns of 1% corresponds to an increase or decrease, respectively, of only .3576% in
the hedge fund market. The observation must be noted that the R2 value is .428, which
means that only 42.8% of the variation in the hedge fund market is being explained by the
regression. Furthermore, examining each of the hedge fund regressions individually yields
more perspective by revealing some potential hazards, but also some detectable trends.
Inspection of the regressions shows that some patterns emerge. The funds with the
highest Sharpe Ratios tend to have, in terms of absolute value, the smallest beta
28. 14
coefficients because low risk implies lower volatility. The following regression graph
illustrates the effect.
Figure 8. Regression of Fund 1 on the S&P 500_________________________________
As can be seen, the beta coefficient indicates that a change of 1% in the stock market
corresponds to a change of only 0.01%. Such a small coefficient might simply reflect a
hedge fund which is volatile but which has data points that are more randomly dispersed
thereby representing a fund which is neither highly positively nor highly negatively
correlated with the S&P 500. There exist a few regressions matching that description for
which polynomial regressions were fitted to the data for somewhat better results in the last
part of this section, but for many of the regressions with extremely low beta coefficients,
the time plots confirm that the low coefficients reflect low volatility. In the case of fund 1
shown above, the associated time plot is shown below.
29. 15
Figure 9. Time Plot Returns Comparison Between Fund 1 and the S&P 500___________
An additional example pair of graphs is shown below for fund 3.
Figure 10. Regression of Fund 3 on the S&P 500________________________________
30. 16
Figure 11. Time Plot Comparison Between Fund 3 and the S&P 500________________
Traversing the list of funds toward the funds with lower Sharpe Ratios leads to regressions
with beta coefficients increasing in absolute value. Ultimately, the purpose of illustrating
how the Sharpe Ratios affect the beta coefficients is to add interpretive meaning to the
average excess fund regression. For example, a citation of the .428 beta coefficient would
be remiss without attributing some of that coefficient’s meaning to the Sharp Ratio’s effect.
A few examples of the increased beta coefficients are shown below.
31. 17
Figure 12. Regression of Fund 16 on the S&P 500_______________________________
Figure 13. Regression of Fund 23 on the S&P 500_______________________________
32. 18
Figure 14. Regression of Fund 26 on the S&P 500_______________________________
Of greatest importance regarding the trend toward increasing beta coefficients is that,
with the exception of a single graph, the highest coefficient of any of the regressions is .
7875, and consequently, the hedge fund market is significantly less volatile than the stock
market. In order to facilitate the attainment of some sense of the extent to which the hedge
fund market trails the increases and decreases of the stock market, the top 15 correlations,
in terms of absolute value, from the section entitled “Correlations” above, have been
juxtaposed below with their corresponding fund numbers and associated regression beta
coefficients.
33. 19
Table 3. Fund Correlations and Beta Coefficients________________________________
Fund #21: Correlation: .781 Beta Coefficient: .7249
Fund #23: Correlation: .7443 Beta Coefficient: .7522
Fund #29: Correlation: .654844 Beta Coefficient: .4571
Fund #2: Correlation: .649193 Beta Coefficient: .5707
Fund #16: Correlation: .62032 Beta Coefficient: .6344
Fund #11: Correlation: .60112 Beta Coefficient: .6685
Fund #26: Correlation: -.58564 Beta Coefficient: -1.2199
Fund #27: Correlation: .58217 Beta Coefficient: .3053
Fund #15: Correlation: .57089 Beta Coefficient: .7875
Fund #22: Correlation: .514196 Beta Coefficient: .5018
Fund #6. Correlation: .49726 Beta Coefficient: .5249
Fund #10: Correlation: .48676 Beta Coefficient: .5338
Fund #8: Correlation: .4705 Beta Coefficient: .401
Fund #18: Correlation: .456296 Beta Coefficient: .4087
Fund #14: Correlation: .4492 Beta Coefficient: .7716
________________________________________________________________________
34. 20
The underlying assumption of analyzing these values is that the funds with the highest
correlations follow the market either naturally or by design such that the results of the
analysis can be used as predictors of the degree to which the broader market of hedge funds
that mimic the S&P 500 follows the stock market. A cursory overview of the data shows
that the beta coefficients are quite high in terms of absolute value, but none of them, except
fund #26, exceeds .8 indicating that hedge funds might actually be a safer investment than
the stock market.
The reader must be cognizant of some discrepancies in the regressions and data. First,
some of the regressions are based on a limited number of performance data. This
deficiency is attributable to the fact that many of the hedge funds that were selected have
not long been in existence, so the sum of 12 data points per year is not many points to plot
over the course of three or fewer years. Second, many of the regressions have very low R2
values. The fact that so many of the regressions have these low values is especially
disturbing because there is no immediate, non-statistical way to account for the proportion
of the regression attributable to error. There is one statistical remedy, however, that has
been employed in the regression graphs on page 57: since many of the graphs seemed to
exhibit non-linear trends, polynomial curves were fitted to the data and generated some
improvement in the R2 values. Some of the scatter plots, however, were so widely
dispersed that even polynomials of orders five and six, which most uniquely fitted the data,
did not yield much improvement in R2. Moreover, interpretation of the polynomial
regressions becomes unwieldy at the higher powers. All of the polynomial curves are only
of order two, and in most of the regressions, the coefficient of the variable raised to the first
power is greater than the beta coefficient in the linear fit, and the coefficients in the
35. 21
polynomial regressions are either both positive or both negative, so the reader can interpret
those results as meaning that a 1% increase or decrease in the S&P 500 corresponds to at
least an increase or at least a decrease of the coefficient of the variable raised to the first
power.
Heteroscedasticity
Scatter plots were derived by first obtaining the residuals from regressing each of the
funds on the S&P 500, squaring the residuals, and then plotting those squared residuals
against the S&P 500 for the purpose of determining the presence or absence of
heteroscedasticity. The reader can see the graphs and SAS regression tables on page 91.
An inspection of the graphs shows that the presence of heteroscedasticity is very weak.
Two primary factors to explain why the variability in hedge fund performance is so weakly
related to the performance of the stock market are immediately suspects. First, hedge fund
investors are typically required to relinquish control over the money invested for a period
of six months or a year unless the investors are willing to accept a penalty for withdrawing
in the middle of that time interval, so whereas stock market investors may invest and
withdraw continually, hedge fund managers can continue their investment strategy with
impunity. Second, because access to hedge fund investing is extremely limited, and
because hedge funds are not required to report their performance to the SEC and, by
extension, the general public, hedge funds are not subject to the same nature of stock
market speculation as issuers of stock are subject to. These two aforementioned possible
reasons, however, imply only that hedge funds are facilitated, not coerced, to lack a strong
presence of heteroscedasticity with the stock market. For example, as will be shown
further into this section, if the nature of a hedge fund, perhaps by construction, is to
36. 22
respond to the stock market, then some meaningful degree of predictive force of the
variability in a hedge fund might exist. Regardless, observation of the scatter plot for the
squared residuals associated with the average fund confirms that, although the residuals are
somewhat widely dispersed or that there are some outliers, depending on how the graph is
interpreted, there is simply no pronounced directional pattern of the residuals other than
strictly homoscedastic horizontal movement. Most of the individual plots are constituted
similarly to this average fund plot.
Before proceeding, the reader should consider the following table containing the betas
from regressing the squared residuals on the S&P 500 as discussed at the beginning of this
section. The funds, from which the residuals were originally obtained when the funds were
regressed on the S&P 500, are numbered in the left column, and the p-values for the t-tests
on the betas, obtained from regressing the squared residuals on the S&P 500, are in the
rightmost column.
Table 4. Betas and p-values of Regressing the Squared Residuals on the S&P 500______
Fund #1: Beta: .02147 p-value: .3414
Fund #2: Beta: -.35204 p-value: .9110
Fund #3: Beta: -.95092 p-value: .0482
Fund #4: Beta: .29539 p-value: .9208
Fund #5: Beta: -14.92568 p-value: .0559
Fund #6: Beta: -11.25687 p-value: .3120
37. 23
Fund #7: Beta: -.78789 p-value: .2328
Fund #8: Beta: 4.70992 p-value: .4046
Fund #9: Beta: 3.86371 p-value: .5341
Fund #10: Beta: -18.79702 p-value: .0031
Fund #11: Beta: -2.54423 p-value: .7317
Fund #12: Beta: -.31229 p-value: .7679
Fund #13: Beta: 1.80260 p-value: .9044
Fund #14: Beta: 1.99910 p-value: .9409
Fund #15: Beta: 12.07186 p-value: .5855
Fund #16: Beta: -13.25197 p-value: .1044
Fund #17: Beta: -7.80340 p-value: .6205
Fund #18: Beta: 8.00835 p-value: .2370
Fund #19: Beta: -2.39933 p-value: .1891
Fund #20: Beta: 4.56654 p-value: .5358
Fund #21: Beta: -6.43843 p-value: .0260
Fund #22: Beta: -22.24671 p-value: .0034
Fund #23: Beta: -2.16036 p-value: .3398
38. 24
Fund #24: Beta: 3.1539 p-value: .1856
Fund #25: Beta: -.01672 p-value: .9693
Fund #26: Beta: -46.81454 p-value: .0245
Fund #27: Beta: -.02682 p-value: .9847
Fund #28: Beta: -32.74194 p-value: .1814
Fund #29: Beta: -4.2439 p-value: .2619
Fund #30: Beta: .33543 p-value: .8406
Average Fund: Beta: -3.41761 p-value: .0215
If the betas were consistently positive or consistently negative, then inference could be
made about the volatility of the hedge fund market when the stock market performed well
or poorly, but 30% of the hedge funds had positive betas which is a percentage too high to
conclude that there is a definitive trend. However, the absence of a trend in the entirety of
the hedge fund market does not imply such an absence in any individual fund, and, in fact,
fund 26 provides an excellent illustration. The graph is shown on the next page. Fund 26
was chosen because it contains a sufficiently-large number of data points, a large beta
coefficient in terms of absolute value, and an R2 value relatively larger than the other funds
with similar numbers of data points: few data points can exaggerate the regression results, a
large beta indicates that the variability of the fund increases or decreases with a movement
in the stock market, and a higher R2 value suggests that more of the changing variability in
39. 25
the fund is attributable to the stock market rather than
Figure 15. The Original Fund 26 Trend Line___________________________________
40. 26
Figure 16. Performance Comparison For Fund 26_______________________________
random error. Now let the reader consider the time plot performance comparison between
fund 26 and the S&P 500. It is shown above. Whenever the S&P 500 is in a state of
decreasing, the performance of the fund tends to be extremely positive or extremely
negative. Thus, whereas the betas did not yield any trend, predicting the variability in
individual funds is possible.
Generally, the graphs tended to be homoscedastic. Some choice examples are given in
the graphs below.
41. 27
Figure 17. Fund 13 Example of Homoscedasticity______________________________
Figure 18. Fund 15 Example of Homoscedasticity______________________________
Some of the graphs were less obviously homoscedastic for reasons that were common to
other funds with similar characteristics. Fund 24 below is the first example. There
42. 28
Figure 19. Fund 24 Example of Aberrant Homoscedasticity_______________________
appear to be outliers present in this graph, but beware that this appearance is illusory,
because the scale on the vertical axis does not extend far compared to other graphs
suffering true outlier effects. Fund 6 below is the second example. Because there are so
43. 29
Figure 20. Fund 6 Example of Insufficient Sample Size__________________________
few data points available, a solid homoscedastic or heteroscedastic trend simply cannot be
known.
Although the scatter plots do not seem to assert the presence of heteroscedasticity, the
betas for funds 3, 10, 21, 22, and 26 and for the average fund were statistically significant
at the .05 level as computed in SAS. Every one of these funds, however, becomes
statistically insignificant when an outlier is removed. Fund 21 functions as an ideal
example. The fund 21 graph is below, and the outlier can clearly be seen in the upper left
corner. When the graph is recomputed without the outlier, not only does the beta become
statistically insignificant, but also the beta, R2, and intercept are changed dramatically.
Figure 21. Fund 21 Squared Residuals With the Outlier vs. S&P 500_______________
44. 30
Figure 22. Fund 21 Squared Residuals Without the Outlier vs. S&P 500_____________
The results can be seen in the graph above. The table below has been prepared to show
relevant information and the t-statistics for the data sets excluding the outliers.
Table 5. t-Statistics and Other Relevant Information for the Modified Graphs________
new intercept new R2
Fund critical t new t df(n-2) new beta
#3: 1.980 > .547058 68 -.77692 48.54227 .0327
#10: 2.021 > 1.65064 31 -9.034 730.2802 .0808
#21: 1.960 > .738383 144 1.509757 741.10505 .0038
#22: 1.960 > .613519 135 -3.56885 1614.52462 .0028
#26: 1.980 > 1.43195 65 -29.0355 4401.04655 .0306
45. 31
Ave. 1.960 > .123604 150 .1486 373.7932 .0001
Because the betas all become statistically insignificant when an outlier is removed from
each of the funds, the p-values generated from the data sets including the outliers are
spurious detectors of heteroscedasticity, and the trend in the data does in fact appear to be
horizontal and homoscedastic in each of these funds in the table.
Normality
In order to test whether the assumptions of the classical regression model were satisfied,
a test of normality for the residuals was conducted by regressing all 30 of the funds and the
average fund on the S&P 500 and plotting the residuals into histograms. The graphs can be
seen on page 40. With the exception of funds 1, 6, 11, 14, 18, 23, and 24, all of the
histograms conformed reasonably well to the normal curve, and in most of those funds
which did not readily conform, there were so few residual data that concluding that the
fund residuals were either normally distributed or not normally distributed in future trends
was premature. In particular, funds 1, 2, 14, and 18 are represented by too few residuals.
A normal fit for funds 6, 11, and 24 might actually be deemed acceptable, but the shape
isn’t as pronounced as it is for the other funds. Fund 23 is the only case for which there
were a sufficient number of residuals available to exclude an obviously normal fit. Most
importantly, the average fund residuals appeared to assume a very strong normal curve
shape, and because many of the arguments presented in this study have been embodied and
buttressed by the results of the average fund regression, the weights of those arguments are
more securely anchored. Some sample graphs depicting the strong tendency toward a
normal curve shape, especially for the average fund, are shown below.
46. 32
Figure 23. Fund 7 Normal Shaped Residuals___________________________________
Figure 24. Fund 21 Normal Shaped Residuals__________________________________
47. 33
Figure 25. Fund 27 Normal Shaped Residuals__________________________________
Figure 26. Average Fund Normal Shaped Residuals______________________________
48. 34
Conclusion
When all the aspects of hedge fund performance are assessed collectively, the hedge
fund market dramatically outperforms the stock market. The average returns discussed in
the introduction show that, purely in terms of generating profit, the hedge fund market
outstripped the S&P 500. In terms of volatility, the hedge fund market performance again
defeated the market. Specifically, the variance data showed that most of the funds,
especially for high Sharpe Ratios, were far less variable and yielded greater returns than the
S&P 500. Moreover, the regression analysis confirmed that the hedge fund market as a
whole remained less volatile than the S&P 500 and that, for those funds highly correlated
to the stock market, the performance fluctuations were more dampened than the stock
market. In addition to the performance and volatility attributes, hedge funds did not exhibit
any heteroscedasticity which effectively translates into more stable expectations on returns
because of the independent nature of hedge fund operations. All of these performance
qualities affirm the superiority of the hedge fund market over the stock market.
49. 35
Descriptive Statistics
Complete tables of the statistical measures used in the study are given here for the
reader who wishes to gain comprehensive insight into the arguments that were proposed.
50. 36
Table 6. Performance Averages______________________________________________
S&P 500: 9.21525 Fund #1: 11.70683544
Fund #2: 26.14545455 Fund #3: 11.27661972
Fund #4: 14.82 Fund #5: 17.65830508
Fund #6: 19.965 Fund #7: 11.8096
Fund #8: 13.65818182 Fund #9: 19.17795918
Fund #10: 15.16941176 Fund #11: 15.28421053
Fund #12: 8.902702703 Fund #13: 21.87630252
Fund #14: 18.70909091 Fund #15: 19.87175258
Fund #16: 15.20047059 Fund #17: 14.57454545
Fund #18: 11.286 Fund #19: 7.771111111
Fund #20: 12.79418182 Fund #21: 12.28
Fund #22: 12.40956522 Fund #23: 9.06375
Fund #24: 7.451320755 Fund #25: 6.177391304
Fund #26: 12.76058824 Fund #27: 6.142702703
Fund #28: 8.341132075 Fund #29: 5.942608696
Fund #30: 5.215 Average Fund: 16.99337751
51. 37
________________________________________________________________________
Table 7. Variance_________________________________________________________
S&P 500: 2415.999957 Fund #1: 8.890983447
Fund #2: 726.1629818 Fund #3: 51.82846841
Fund #4: 453.5904889 Fund #5: 938.9087074
Fund #6: 1524.90216 Fund #7: 363.6100871
Fund #8: 653.8928485 Fund #9: 1908.131577
Fund #10: 1129.798794 Fund #11: 1126.69836
Fund #12: 193.7927049 Fund #13: 3808.637586
Fund #14: 3189.621818 Fund #15: 4362.841398
Fund #16: 2253.336476 Fund #17: 2108.215882
Fund #18: 993.25512 Fund #19: 250.9952252
Fund #20: 2125.89921 Fund #21: 2205.50663
Fund #22: 2575.932906 Fund #23: 985.9317532
Fund #24: 420.0005155 Fund #25: 137.1764503
Fund #26: 7179.599546 Fund #27: 254.4790703
Fund #28: 3352.015176 Fund #29: 555.4719747
52. 38
Fund #30: 253.566817 Average Fund: 753.2803166
________________________________________________________________________
Table 8. Fund Correlation with the S&P 500___________________________________
Fund #1: -0.173464524 Fund #2: 0.649193451
Fund #3: -0.044247993 Fund #4: -0.0714241
Fund #5: -0.207737278 Fund #6: 0.497261555
Fund #7: 0.414375015 Fund #8: 0.470509021
Fund #9: 0.253269275 Fund #10: 0.486760127
Fund #11: 0.601120061 Fund #12: -0.190548726
Fund #13: 0.407577199 Fund #14: 0.44922268
Fund #15: 0.570891982 Fund #16: 0.620327411
Fund #17: 0.298016148 Fund #18: 0.45629676
Fund #19: 0.262008673 Fund #20: 0.409339549
Fund #21: 0.781050409 Fund #22: 0.514196748
Fund #23: 0.744336757 Fund #24: 0.202843746
Fund #25: 0.421641744 Fund #26: -0.585642472
Fund #27: 0.582170503 Fund #28: 0.35064364
Fund #29: 0.65484412 Fund #30: -0.083269169
53. 39
Average Fund: 0.654979938 __________________________________________
Residual Histograms
The residuals from regressing each of the funds on the S&P 500 were plotted and placed
into histograms as given here.
54. 40
Figure 27. Fund 1 Residuals________________________________________________
Figure 28. Fund 2 Residuals________________________________________________
55. 41
Figure 29. Fund 3 Residuals________________________________________________
Figure 30. Fund 4 Residuals________________________________________________
56. 42
Figure 31. Fund 5 Residuals________________________________________________
Figure 32. Fund 6 Residuals________________________________________________
57. 43
Figure 33. Fund 7 Residuals________________________________________________
Figure 34. Fund 8 Residuals________________________________________________
58. 44
Figure 35. Fund 9 Residuals________________________________________________
Figure 36. Fund 10 Residuals_______________________________________________
59. 45
Figure 37. Fund 11 Residuals_______________________________________________
Figure 38. Fund 12 Residuals_______________________________________________
60. 46
Figure 39. Fund 13 Residuals_______________________________________________
Figure 40. Fund 14 Residuals_______________________________________________
61. 47
Figure 41. Fund 15 Residuals_______________________________________________
Figure 42. Fund 16 Residuals_______________________________________________
62. 48
Figure 43. Fund 17 Residuals_______________________________________________
Figure 44. Fund 18 Residuals_______________________________________________
63. 49
Figure 45. Fund 19 Residuals_______________________________________________
Figure 46. Fund 20 Residuals_______________________________________________
64. 50
Figure 47. Fund 21 Residuals_______________________________________________
Figure 48. Fund 22 Residuals_______________________________________________
65. 51
Figure 49. Fund 23 Residuals_______________________________________________
Figure 50. Fund 24 Residuals_______________________________________________
66. 52
Figure 51. Fund 25 Residuals_______________________________________________
Figure 52. Fund 26 Residuals_______________________________________________
67. 53
Figure 53. Fund 27 Residuals_______________________________________________
Figure 54. Fund 28 Residuals_______________________________________________
68. 54
Figure 55. Fund 29 Residuals_______________________________________________
Figure 56. Fund 30 Residuals_______________________________________________
70. 56
Regressions
These graphs are the result of regressing each of the funds on the S&P 500 and then
determining a regression line. In many of the graphs, polynomial regressions were also
determined and plotted as curves.
71. 57
Figure 58. Fund 1 Regression_______________________________________________
72. 58
Figure 59. Fund 2 Regression_______________________________________________
73. 59
Figure 60. Fund 3 Regression_______________________________________________
Figure 61. Fund 4 Regression_______________________________________________
102. 88
Figure 89. Fund 1 Time Plot Comparison______________________________________
Figure 90. Fund 2 Time Plot Comparison______________________________________
104. 90
Figure 91. Fund 3 Time Plot Comparison______________________________________
Figure 92. Fund 4 Time Plot Comparison______________________________________
105. 91
Figure 93. Fund 5 Time Plot Comparison______________________________________
Figure 94. Fund 6 Time Plot Comparison______________________________________
106. 92
Figure 95. Fund 7 Time Plot Comparison______________________________________
Figure 96. Fund 8 Time Plot Comparison______________________________________
107. 93
Figure 97. Fund 9 Time Plot Comparison______________________________________
Figure 98. Fund 10 Time Plot Comparison_____________________________________
108. 94
Figure 99. Fund 11 Time Plot Comparison_____________________________________
Figure 100. Fund 12 Time Plot Comparison____________________________________
109. 95
Figure 101. Fund 13 Time Plot Comparison____________________________________
Figure 102. Fund 14 Time Plot Comparison____________________________________
110. 96
Figure 103. Fund 15 Time Plot Comparison____________________________________
Figure 104. Fund 16 Time Plot Comparison____________________________________
111. 97
Figure 105. Fund 17 Time Plot Comparison____________________________________
Figure 106. Fund 18 Time Plot Comparison____________________________________
112. 98
Figure 107. Fund 19 Time Plot Comparison____________________________________
Figure 108. Fund 20 Time Plot Comparison____________________________________
114. 100
Figure 109. Fund 21 Time Plot Comparison____________________________________
Figure 110. Fund 21 Time Plot Comparison____________________________________
116. 102
Figure 111. Fund 21 Time Plot Comparison____________________________________
Figure 112. Fund 21 Time Plot Comparison____________________________________
117. 103
Figure 113. Fund 21 Time Plot Comparison____________________________________
Figure 114. Fund 21 Time Plot Comparison____________________________________
118. 104
Figure 115. Fund 27 Time Plot Comparison____________________________________
Figure 116. Fund 28 Time Plot Comparison____________________________________
119. 105
Figure 117. Fund 29 Time Plot Comparison____________________________________
Figure 118. Fund 30 Time Plot Comparison____________________________________
121. 107
Heteroscedasticity graphs and SAS output tables
In order to determine the regression graphs discussed and shown in the body of this
paper, each of the funds was regressed on the S&P 500. The SAS output tables in this
section are the result of squaring the residuals from those regressions, and then regressing
the squared residuals on the S&P 500 for the purpose of analyzing the p-values of the test
statistic for the betas. The graphs are the result of plotting the squared residuals vs. the
S&P 500.
122. 108
Table 9. Squared Residuals on Fund 1_________________________________________
The REG Procedure
Model: MODEL1
Dependent Variable: squaredResidual
Number of Observations Read 79
Number of Observations Used 79
Analysis of Variance
Sum of Mean
Source DF Squares Square F Value Pr > F
Model 1 70.77308 70.77308 0.92 0.3417
Error 77 5954.70142 77.33378
Corrected Total 78 6025.47451
Root MSE 8.79396 R-Square 0.0117
Dependent Mean 8.48215 Adj R-Sq -0.0011
Coeff Var 103.67613
Parameter Estimates
Parameter Standard
Variable Label DF Estimate Error t Value Pr > |t|
Intercept Intercept 1 8.52905 0.99061 8.61 <.0001
excess_market_return excess_market_return 1 0.02147 0.02244 0.96 0.3417
Figure 120. Squared Residuals against Fund 1__________________________________
123. 109
Table 10. Squared Residuals on Fund 2________________________________________
Figure 121. Squared Residuals Against Fund 2__________________________________
124. 110
Table 11. Squared Residuals on Fund 3________________________________________
The REG Procedure
Model: MODEL1
Dependent Variable: squaredResidual
Number of Observations Read 71
Number of Observations Used 71
Analysis of Variance
Sum of Mean
Source DF Squares Square F Value Pr > F
Model 1 113937 113937 4.05 0.0482
Error 69 1942836 28157
Corrected Total 70 2056773
Root MSE 167.80060 R-Square 0.0554
Dependent Mean 51.70812 Adj R-Sq 0.0417
Coeff Var 324.51500
Parameter Estimates
Parameter Standard
Variable Label DF Estimate Error t Value Pr > |t|
Intercept Intercept 1 50.63853 19.92136 2.54 0.0133
excess_market_return excess_market_return 1 -0.95092 0.47272 -2.01 0.0482
Figure 122. Squared Residuals Against Fund 3__________________________________
125. 111
Table 12. Squared Residuals on Fund 4________________________________________
The REG Procedure
Model: MODEL1
Dependent Variable: squaredResidual
Number of Observations Read 82
Number of Observations Used 82
Analysis of Variance
Sum of Mean
Source DF Squares Square F Value Pr > F
Model 1 13521 13521 0.01 0.9208
Error 80 108791378 1359892
Corrected Total 81 108804898
Root MSE 1166.14417 R-Square 0.0001
Dependent Mean 445.13084 Adj R-Sq -0.0124
Coeff Var 261.97784
Parameter Estimates
Parameter Standard
Variable Label DF Estimate Error t Value Pr > |t|
Intercept Intercept 1 445.98346 129.06266 3.46 0.0009
excess_market_return excess_market_return 1 0.29539 2.96242 0.10 0.9208
Figure 123. Squared Residuals Against Fund 4__________________________________
Table 13. Squared Residuals on Fund 5________________________________________
126. 112
The REG Procedure
Model: MODEL1
Dependent Variable: squaredResidual
Number of Observations Read 59
Number of Observations Used 59
Analysis of Variance
Sum of Mean
Source DF Squares Square F Value Pr > F
Model 1 13202690 13202690 3.81 0.0559
Error 57 197554173 3465863
Corrected Total 58 210756862
Root MSE 1861.68275 R-Square 0.0626
Dependent Mean 883.62026 Adj R-Sq 0.0462
Coeff Var 210.68810
Parameter Estimates
Parameter Standard
Variable Label DF Estimate Error t Value Pr > |t|
Intercept Intercept 1 961.50702 245.63372 3.91 0.0002
excess_market_return excess_market_return 1 -14.92568 7.64731 -1.95 0.0559
Figure 124. Squared Residuals Against Fund 5__________________________________
Table 14. Squared Residuals on Fund 6________________________________________
The REG Procedure
Model: MODEL1
127. 113
Dependent Variable: squaredResidual
Number of Observations Read 16
Number of Observations Used 16
Analysis of Variance
Sum of Mean
Source DF Squares Square F Value Pr > F
Model 1 2539867 2539867 1.10 0.3120
Error 14 32314491 2308178
Corrected Total 15 34854358
Root MSE 1519.26887 R-Square 0.0729
Dependent Mean 1074.69491 Adj R-Sq 0.0066
Coeff Var 141.36746
Parameter Estimates
Parameter Standard
Variable Label DF Estimate Error t Value Pr > |t|
Intercept Intercept 1 1000.51918 386.34344 2.59 0.0214
excess_market_return excess_market_return 1 -11.25687 10.73116 -1.05 0.3120
Figure 125. Squared Residuals Against Fund 6__________________________________
Table 15. Squared Residuals on Fund 7________________________________________
The REG Procedure
Model: MODEL1
Dependent Variable: squaredResidual
128. 114
Number of Observations Read 150
Number of Observations Used 150
Analysis of Variance
Sum of Mean
Source DF Squares Square F Value Pr > F
Model 1 228957 228957 1.44 0.2328
Error 148 23604159 159488
Corrected Total 149 23833116
Root MSE 399.35893 R-Square 0.0096
Dependent Mean 291.99981 Adj R-Sq 0.0029
Coeff Var 136.76685
Parameter Estimates
Parameter Standard
Variable Label DF Estimate Error t Value Pr > |t|
Intercept Intercept 1 295.03138 32.70554 9.02 <.0001
excess_market_return excess_market_return 1 -0.78789 0.65758 -1.20 0.2328
Figure 126. Squared Residuals Against Fund 7__________________________________
Table 16. Squared Residuals on Fund 8________________________________________
The REG Procedure
Model: MODEL1
Dependent Variable: squaredResidual
Number of Observations Read 55
Number of Observations Used 55
129. 115
Analysis of Variance
Sum of Mean
Source DF Squares Square F Value Pr > F
Model 1 1069506 1069506 0.71 0.4046
Error 53 80292858 1514960
Corrected Total 54 81362364
Root MSE 1230.83694 R-Square 0.0131
Dependent Mean 499.47060 Adj R-Sq -0.0055
Coeff Var 246.42831
Parameter Estimates
Parameter Standard
Variable Label DF Estimate Error t Value Pr > |t|
Intercept Intercept 1 488.21475 166.50580 2.93 0.0050
excess_market_return excess_market_return 1 4.70992 5.60560 0.84 0.4046
Figure 127. Squared Residuals Against Fund 8__________________________________
Table 17. Fund 9 Squared Residuals on the S&P 500_____________________________
The REG Procedure
Model: MODEL1
Dependent Variable: squaredResidual
Number of Observations Read 147
Number of Observations Used 147
Analysis of Variance
130. 116
Sum of Mean
Source DF Squares Square F Value Pr > F
Model 1 5573957 5573957 0.39 0.5341
Error 145 2080216229 14346319
Corrected Total 146 2085790186
Root MSE 3787.65347 R-Square 0.0027
Dependent Mean 1764.95251 Adj R-Sq -0.0042
Coeff Var 214.60371
Parameter Estimates
Parameter Standard
Variable Label DF Estimate Error t Value Pr > |t|
Intercept Intercept 1 1756.26783 312.71094 5.62 <.0001
excess_market_return excess_market_return 1 3.86371 6.19859 0.62 0.5341
Figure 128. Fund 9 Squared Residuals Against the S&P 500_______________________
Table 18. Fund 10 Squared Residuals on the S&P 500____________________________
The REG Procedure
Model: MODEL1
Dependent Variable: squaredResidual
Number of Observations Read 34
Number of Observations Used 34
Analysis of Variance
Sum of Mean
Source DF Squares Square F Value Pr > F
131. 117
Model 1 10785697 10785697 10.21 0.0031
Error 32 33799777 1056243
Corrected Total 33 44585475
Root MSE 1027.73685 R-Square 0.2419
Dependent Mean 835.55238 Adj R-Sq 0.2182
Coeff Var 123.00089
Parameter Estimates
Parameter Standard
Variable Label DF Estimate Error t Value Pr > |t|
Intercept Intercept 1 850.42414 176.31685 4.82 <.0001
excess_market_return excess_market_return 1 -18.79702 5.88230 -3.20 0.0031
Figure 129. Fund 10 Squared Residuals Against the S&P 500______________________
Table 19. Fund 11 Squared Residuals on the S&P 500____________________________
The REG Procedure
Model: MODEL1
Dependent Variable: squaredResidual
Number of Observations Read 38
Number of Observations Used 38
Analysis of Variance
Sum of Mean
Source DF Squares Square F Value Pr > F
Model 1 217459 217459 0.12 0.7317
132. 118
Error 36 65562859 1821191
Corrected Total 37 65780317
Root MSE 1349.51492 R-Square 0.0033
Dependent Mean 701.61839 Adj R-Sq -0.0244
Coeff Var 192.34315
Parameter Estimates
Parameter Standard
Variable Label DF Estimate Error t Value Pr > |t|
Intercept Intercept 1 698.60214 219.09418 3.19 0.0030
excess_market_return excess_market_return 1 -2.54423 7.36285 -0.35 0.7317
Figure 130. Fund 11 Squared Residuals Against the S&P 500______________________
Table 20. Fund 12 Squared Residuals on the S&P 500____________________________
The REG Procedure
Model: MODEL1
Dependent Variable: squaredResidual
Number of Observations Read 74
Number of Observations Used 74
Analysis of Variance
Sum of Mean
Source DF Squares Square F Value Pr > F
Model 1 12574 12574 0.09 0.7679
Error 72 10313729 143246
Corrected Total 73 10326302
133. 119
Root MSE 378.47884 R-Square 0.0012
Dependent Mean 181.82405 Adj R-Sq -0.0127
Coeff Var 208.15664
Parameter Estimates
Parameter Standard
Variable Label DF Estimate Error t Value Pr > |t|
Intercept Intercept 1 181.42427 44.01796 4.12 <.0001
excess_market_return excess_market_return 1 -0.31229 1.05406 -0.30 0.7679
Figure 131. Fund 12 Squared Residuals Against the S&P 500______________________
Table 21. Fund 13 Squared Residuals on the S&P 500____________________________
The REG Procedure
Model: MODEL1
Dependent Variable: squaredResidual
Number of Observations Read 119
Number of Observations Used 119
Analysis of Variance
Sum of Mean
Source DF Squares Square F Value Pr > F
Model 1 999622 999622 0.01 0.9044
Error 117 8072698768 68997425
Corrected Total 118 8073698390
134. 120
Root MSE 8306.46889 R-Square 0.0001
Dependent Mean 3136.72300 Adj R-Sq -0.0084
Coeff Var 264.81359
Parameter Estimates
Parameter Standard
Variable Label DF Estimate Error t Value Pr > |t|
Intercept Intercept 1 3140.00131 761.93971 4.12 <.0001
excess_market_return excess_market_return 1 1.80260 14.97608 0.12 0.9044
Figure 132. Fund 13 Squared Residuals Against the S&P 500______________________
Table 22. Fund 14 Squared Residuals on the S&P 500____________________________
The REG Procedure
Model: MODEL1
Dependent Variable: squaredResidual
Number of Observations Read 22
Number of Observations Used 22
Analysis of Variance
Sum of Mean
Source DF Squares Square F Value Pr > F
Model 1 90674 90674 0.01 0.9409
Error 20 321555936 16077797
Corrected Total 21 321646610
Root MSE 4009.71281 R-Square 0.0003
Dependent Mean 2431.52719 Adj R-Sq -0.0497
135. 121
Coeff Var 164.90512
Parameter Estimates
Parameter Standard
Variable Label DF Estimate Error t Value Pr > |t|
Intercept Intercept 1 2433.15191 855.14736 2.85 0.0100
excess_market_return excess_market_return 1 1.99910 26.61988 0.08 0.9409
Figure 133. Fund 14 Squared Residuals Against the S&P 500______________________
Table 23. Fund 15 Squared Residuals on the S&P 500____________________________
The REG Procedure
Model: MODEL1
Dependent Variable: squaredResidual
Number of Observations Read 97
Number of Observations Used 97
Analysis of Variance
Sum of Mean
Source DF Squares Square F Value Pr > F
Model 1 32361346 32361346 0.30 0.5855
Error 95 10263473634 108036565
Corrected Total 96 10295834980
Root MSE 10394 R-Square 0.0031
Dependent Mean 2911.99771 Adj R-Sq -0.0074
Coeff Var 356.93929
136. 122
Parameter Estimates
Parameter Standard
Variable Label DF Estimate Error t Value Pr > |t|
Intercept Intercept 1 2966.53641 1060.05147 2.80 0.0062
excess_market_return excess_market_return 1 12.07186 22.05700 0.55 0.5855
Figure 134. Fund 15 Squared Residuals Against the S&P 500______________________
Table 24. Fund 16 Squared Residuals on the S&P 500____________________________
The REG Procedure
Model: MODEL1
Dependent Variable: squaredResidual
Number of Observations Read 85
Number of Observations Used 85
Analysis of Variance
Sum of Mean
Source DF Squares Square F Value Pr > F
Model 1 31913896 31913896 2.70 0.1044
Error 83 982792968 11840879
Corrected Total 84 1014706864
Root MSE 3441.05785 R-Square 0.0315
Dependent Mean 1369.74415 Adj R-Sq 0.0198
Coeff Var 251.21902
Parameter Estimates
137. 123
Parameter Standard
Variable Label DF Estimate Error t Value Pr > |t|
Intercept Intercept 1 1330.11919 374.01473 3.56 0.0006
excess_market_return excess_market_return 1 -13.25197 8.07203 -1.64 0.1044
Figure 135. Fund 16 Squared Residuals Against the S&P 500______________________
Table 25. Fund 17 Squared Residuals on the S&P 500____________________________
The REG Procedure
Model: MODEL1
Dependent Variable: squaredResidual
Number of Observations Read 33
Number of Observations Used 33
Analysis of Variance
Sum of Mean
Source DF Squares Square F Value Pr > F
Model 1 1798265 1798265 0.25 0.6205
Error 31 222881127 7189714
Corrected Total 32 224679392
Root MSE 2681.36417 R-Square 0.0080
Dependent Mean 1858.27275 Adj R-Sq -0.0240
Coeff Var 144.29336
Parameter Estimates
Parameter Standard
138. 124
Variable Label DF Estimate Error t Value Pr > |t|
Intercept Intercept 1 1857.09987 466.77148 3.98 0.0004
excess_market_return excess_market_return 1 -7.80340 15.60318 -0.50 0.6205
Figure 136. Fund 17 Squared Residuals Against the S&P 500______________________
Table 26. Fund 18 Squared Residuals on the S&P 500____________________________
The REG Procedure
Model: MODEL1
Dependent Variable: squaredResidual
Number of Observations Read 20
Number of Observations Used 20
Analysis of Variance
Sum of Mean
Source DF Squares Square F Value Pr > F
Model 1 1464962 1464962 1.50 0.2370
Error 18 17623855 979103
Corrected Total 19 19088816
Root MSE 989.49635 R-Square 0.0767
Dependent Mean 746.42674 Adj R-Sq 0.0255
Coeff Var 132.56443
Parameter Estimates
Parameter Standard
Variable Label DF Estimate Error t Value Pr > |t|