CHAPTER 13 The Effect of Management and Incentive Fees on the Performance of CTAs: A Note Fernando Diz This chapter examines the effect of management and incentive fees on the performanc
Trang 1CHAPTER 13 The Effect of Management and Incentive Fees on the Performance of CTAs: A Note
Fernando Diz
This chapter examines the effect of management and incentive fees on the performance and volatility of CTA track records Evidence of a struc-tural change in incentive compensation is presented that points to a larger reliance on incentive fees as opposed to management fees Management fees have no relationship to performance No systematic performance or volatil-ity penalty is suffered by investors by this type of compensation Incentive fees are found to be positively related to both net of fees returns and volatil-ity An increase in the incentive fee parameter from 10 percent to 20 per-cent will increase performance by an average of 6.58 perper-cent per year The performance increase is net of the effects of leverage and other variables affecting performance There is also a small tendency for CTAs with larger amounts of assets under management to have slightly better performance INTRODUCTION
This chapter empirically examines the effect of incentive compensation con-tracts of commodity trading advisors (CTAs) on their performance The analysis is an extension of Golec (1993) and examines the effects of incen-tive compensation contracts on the risk and return of CTAs The contribu-tion of this chapter is twofold In Golec, the sample used was too small to
248
Partial support for the completion of this study was provided by a grant from the Foundation for Managed Derivatives Research The author wishes to thank the Foundation for its support, and Sol Waksman for his invaluable comments.
Trang 2draw reliable inferences about the effect of incentive compensation on the risk and return of CTAs Current events in the money management world associated with manager compensation abuses have heightened the impor-tance of measuring the effects of compensation more accurately A much larger database than the one used in Golec allows us to measure these effects with less error The advantages of using a larger database are even more important in view of the structural changes in the composition of total compensation in managed futures, as documented by Diz and Shukla (2003) In addition, this study measures the effects of management and incentive fees on the risk and returns of CTAs more accurately by control-ling for known effects that other very important variables have on these measures of performance (see Diz 2003)
CTA COMPENSATION STRUCTURE
CTA compensation contracts generally contain two types of fees: a
man-agement fee, k m, which represents a fixed percentage of end-of-period assets
under management, and an incentive fee, k i, which represents a fixed per-centage of investment gains over a year period
The CTA total fee income for period t can be written as:
Φt = kmA t + k i max[0,A t − A t− 1]
where A t− 1and A t= dollar value of assets under management at the end
of periods t − 1 and t respectively.
Defining R pt as the CTA’s portfolio rate of return for one period (t− 1
to t), we can redefine A t as A t− 1(1 + R pt) We can then rewrite the total compensation equation as:
Φt = kmA t− 1(1+ R pt)+ k i max[0,A t − 1 R pt] (13.2) Equation 13.2 shows the dependence of CTA total compensation on
the level of assets under management (A t− 1), the one period performance
(R pt ), the management fee (k m ), and the incentive fee (k i) Base compensa-tion is a linear funccompensa-tion of the level of assets under management Incentive compensation is a nonlinear function of performance Table 13.1 contains summary statistics for the variables included in equations 13.1 and 13.2 The median management and incentive fees for a sample of 974 CTAs over
ki At At if Rpt
k A
− − >
,
Trang 3this study sample period (1974 to 1998) were 2 percent and 20 percent respectively
Although the management and incentive fees presented in Table 13.1 appear high when compared to mutual funds (e.g., Golec 1993), average monthly returns appear higher than what one finds for mutual funds for the same time period This is especially telling if one considers that CTA reported performance figures are net of all fees To date, no study has accu-rately accounted for all fee-adjusted performance of mutual funds when comparing them to fee-adjusted performance in the managed futures indus-try Further, it is a known fact that mutual fund fees have continued to increase and that this increase has not translated into higher returns for individual investors It is generally acknowledged that higher fees in the mutual fund industry have reduced returns to investors (Trzcinka 1998) Management fees in the managed futures industry have followed a down-ward trend from an average high of 2.81 percent in 1982 to an average low
of 1.85 percent in 2002 More of CTA compensation in 2002 came in the form of incentive fees (Diz and Shukla 2003) The results in Table 13.2 highlight the change in the total compensation structure for CTAs Almost
50 percent of total CTA compensation came from management fees in 1982 while in 2002 only 35 percent of total CTA compensation came from these asset-based fees Two-thirds of CTA compensation came from performance based fees in 2002 Golec’s study used CTA data from 1982 to 1987 The structural change that is evident in the 1990s is a third reason for review-ing Golec’s (1993) findreview-ings
Because the purpose of this study is not to explore the theory of com-pensation contracting, we refer the reader to Golec (1993) for such a review
TABLE 13.1 Summary Statistics for CTA Management and Incentive Fees,
Assets Under ManagementaVariables, and Returns
Variable Mean Std Dev Median Q(1) Q(3) Min Max
Management
Incentive fee (%) 20.27 0.044 20.00 20.00 20.00 0.00 50.00 Assets
(Millions$) 34.68 186.950 1.80 0.50 10.52 0.10 2,954.00 Monthly
returnb(%) 0.99 0.013 0.94 0.38 1.50 −5.23 10.00
aAssets under management can and often do include notional assets
b
Reported returns are net of management and incentive fees.
Trang 4What is clear from equation 13.2 is that total CTA compensation is a
func-tion of performance (R pt ), the level of assets under management (A t), and
the management and incentive fee rates (k m and k i)
DATA
The data used in this chapter consist of individual CTA monthly returns provided by the Barclay Trading Group The database contains records for 1,253 CTAs and includes both programs that were still listed as of February 1998 as well programs that were delisted anytime from 1975 to January 1998 Of the total 1,253 programs, 798 had been delisted by Feb-ruary 1998 Only 455 programs were listed as of FebFeb-ruary 1998 Of the 1,253 programs, only 989 (80 percent) reported margin to equity ratios
TABLE 13.2 Evolution of Management, Incentive Fees, and Total Compensation
in the Managed Futures Industry, 1982–2002
Average Average MF as % of IF as % of
N Year Management Feea Incentive Feea Fee Revenue Fee Revenue
aManagement and incentive fees are reported fees Actual average fees are likely to
be lower since these are subject to negotiation
Source: Diz and Shukla (2003).
Trang 5Fifteen programs were eliminated from the sample for various reasons rang-ing from missrang-ing observations to duplication This left us with a sample of
974 programs Golec’s sample includes only 80 CTAs The time spanned by the two samples is also worth noting Our sample spans a period of 24 years starting in 1975 and ending in 1998 Golec’s sample spans only a five-year period from May 1982 to December 1986 Summary statistics were calculated for each CTA in the sample Table 13.3 provides a summary of the averages for these statistics
The average length of a CTA track record for the sample is about 5.5 years The longest track record is 23 years and the shortest only 5 months The average monthly rate of return for the combined CTAs was 1.31 per-cent and the annual standard deviation of returns for the cross section of CTAs was 26.24 percent These results are consistent with Brorsen (1998) for his combined CTA sample Golec’s study reports a monthly average rate
of return of 1.35 percent with an annual standard deviation of 11.56 per-cent The sample used in this chapter is more similar in size, composition, and performance to Brorsen’s
The average management fee for the sample is 2.46 percent while the same average is 3.96 percent in Golec’s sample More strikingly, the median management fee for this study’s is 2.00 percent while it is 4.00 percent in Golec’s sample The average incentive fee for the sample in this study is 20.27 percent while the same average is 16.33 percent in Golec’s sample The median incentive fee for this study’s sample is 20.00 percent and only 15.00 percent in Golec’s sample Finally, the average assets under man-agement in this study were $34.68 million compared to $5.01 million in
TABLE 13.3 Summary of CTA Average Attributes, February 1974–February
1998, 974 CTA Programs
Margin to equity ratio (%) 19.40 10.58 1.03 100.00 Annual compounded rate
Annual standard deviation (%) 26.24 18.41 0.79 142.89
Trang 6Golec’s sample The median amount of assets under management for this sample was $1.8 million versus $1.5 million for Golec’s sample
It is clear from the data that the sample is our study is broader in coverage, size, composition, performance variability, and time span than Golec’s As such, it is perhaps more suitable to accurately measure the effects of compensation structure on CTA performance
CTA COMPENSATION PARAMETERS
AND PERFORMANCE
In this section we empirically explore the relationship between CTA returns and the standard deviation of returns to their compensation parameters by replicating Golec’s (1993) analysis We examined the issue
by fitting two ordinary least squares (OLS) cross-sectional regressions on the means and standard deviations of returns of the CTAs on their fee parameters as follows:
AROR j = b0+ b1k m + b2k i + b3ln(A t− 1)+ e j (13.3)
s j = a0+ a1k m + a2k i + a3ln(A t− 1)+ u j (13.4)
where AROR j = annual compounded rate of return for CTA j
s j = annual standard deviation of CTA jreturns
e j , u j = error terms
Because the distribution of assets under management is clearly skewed, we use the natural logarithm of assets under management as the “size” vari-able Significance tests use White’s (see Greene 2000) heteroskedasticity consistent standard errors Table 13.4 presents OLS estimates of regression
TABLE 13.4 Estimation of the Relationship between Compensation Parameters
and CTA Mean Annual Compounded Returns and Standard Deviation of Returns
Independent Variables Dependent Variables Intercept k m k i ln(A t − 1)
Mean Annual Returns −0.255* 0.580 0.693* 0.016*
(0.075) (0.583) (0.259) (0.003) Standard Deviation 0.229* 1.424* 0.654* −0.009*
(0.057) (0.482) (0.156) (0.003)
*Significant at the 1 percent level under H = 0
Trang 7coefficients from equations 13.3 and 13.4, along with white standard errors
in parentheses
The results in Table 13.4 show that cross-sectional variation in mean returns is not related to management fees This result is in agreement with Golec (1993), and it is good news for investors as it suggests that there are
no systematic abuses in management fees that penalize performance The cross-sectional variation in mean returns also is shown to be positively asso-ciated with the incentive fee parameter This is also in agreement with Golec’s results, and it is also good news for investors because greater incen-tive fee parameters lead to greater CTA effort or ability that in turn leads to
higher performance If the incentive fee parameter k iwere to increase from
10 percent to 20 percent, performance should be expected to increase by 5.8 percent The magnitude of the increase is roughly half of what was found in Golec and seems like a much more reasonable number A 10 percent increase
in Golec’s study would have accounted for a 1 percent per month increase in performance or more than 12 percent per year, a very large number It is important to highlight that the performance increase is net of all fees Other things being equal, a CTA with higher incentive fees is likely to deliver larger performance after fees
We find the amount of assets under management to have a positive effect on performance while Golec (1993) finds the opposite result Our finding is likely to reflect a known fact in the industry that successful CTAs tend to capture the bulk of assets under management The amount of assets under management tends to reflect performance The newly created Barclay BTOP50 Index for managed futures is only a reflection of this known fact The increase in performance associated with assets under management is not spectacular An increase in assets under management from $100,000 to $3 billion is associated with a 16-basis-point increase in performance Although the effect appears to be statistically different from zero, its economic importance is very small A similar increase in assets under management is associated with a decrease in performance of 71 basis points in Golec Figure 13.1 illustrates the annual increases/decreases
in performance as a function of assets under management found in this study and in Golec (1993).1
The volatility of CTAs’ track records appears to be positively associated with the incentive fee parameter (Table 13.4) The relationship supports the idea that CTAs who charge larger incentive fees take on larger risks It also appears that risk taking pays off as viewed from the relationship between mean returns and the incentive fee parameter The amount of assets under
1 Golec’s results were annualized to make them comparable to the results of this study.
Trang 8management appears to be negatively associated with the volatility of CTAs’ track records Although the effect is rather small, this result is con-sistent with Golec’s findings Golec’s explanation of this empirical observa-tion is appealing Risk aversion is likely to rise with wealth, and this in turn may induce CTAs to reduce risk levels Some indirect support for this expla-nation is found in Diz (2003), where the level of leverage of “surviving’’ CTAs (the larger ones) appears to be smaller
One surprising finding is that management fees appear to be positively associated with the volatility of CTAs’ track records There is no clear explanation for this finding other than measurement error
Because a substantial amount of relative total compensation is contin-gent on positive performance (incentive fee), common sense and theory suggest that all factors associated with performance have a potential impact
on total compensation For example, Diz (2003) shows that CTAs’ level of leverage is related to performance CTAs with larger margin to equity ratios tend to have larger returns and volatility As other variables such as lever-age are strongly associated with the performance of a cross section of CTAs, the exclusion of such variables in regression equations 13.2 and 13.3 may substantially alter the size, sign, and level of statistical significance of their
40 20 0
−20
−40
−60 –80 –100 –120 –140 –160
Assets under Management
0 5e+08 1e+09 1.5e+09 2e+09
This Study Golec (1993)
FIGURE 13.1 Effects of Assets under Management on Average Annual Returns
Trang 9coefficients In an effort to reduce the omitted variable problem, we fit this augmented model to the data:
AROR j = b0+ b1k m + b2k i + b3ln(A t− 1)+ b4mdd + b5me+
+ b6vr + b7surv + b8Diver + b9Syst + b10Disc + e j
(13.5)
s j = a0+ a1k m + a2k i + a3ln(A t− 1)+ a4mdd + a5me+
+ a6vr + a7surv + a8Diver + a9Syst + a10Disc + u j
(13.6)
where:
k m = management fee parameter in %
k i = incentive fee variable in %
ln(A t− 1) = natural log of the amount of assets under management
in the previous month
mdd = maximum drawdown variable (drawdown is defined as
the percentage size of an equity retracement)
me = margin to equity ratio
vr = ratio of “positive” to “negative” volatility
surv = dummy variable that takes a value of 1 when the CTA
is still in business and 0 when the CTA or program is
no longer available
Diver = dummy variable that takes a value of 1 when the CTA
is diversified and 0 otherwise
Syst = dummy variable that takes a value of 1 when the CTA
is systematic in trading approach and 0 otherwise
Discr = dummy variable that takes a value of 1 when the CTA
is discretionary in trading and 0 otherwise
e j , u j = error terms The results in Table 13.5 suggest that management fees are unrelated to both the level and volatility of CTA returns The effect of the incentive fee parameter remains positive and statistically significantly different from zero under the augmented model specification Moreover, the magnitude of the effect of the incentive fee parameter on the level of returns appears to be the same as in the previous model specification The robustness of the incentive fee parameter to different model specifications lends credence to the conclusion that CTAs’ incentive fee structure is strongly associated with their level of net returns Under the augmented model, an increase in the incentive fee parameter from 10 percent to 20 percent will increase per-formance by an average of 6.58 percent annually
Incentive fees continue to be associated with the overall volatility of CTA track records Larger incentive fee parameters are associated with
Trang 10larger levels of volatility, although this effect is reduced considerably in the augmented model The amount of assets under management continues to be associated with the mean level of returns The effect appears to be of the same order of magnitude in the augmented model The level of assets under management is unrelated to the volatility of the CTA track record This is
in contrast with Golec (1993), who finds a negative and significant rela-tionship between assets under management and volatility and casts doubts about the existence of any relationship between size and volatility once one accounts for other volatility variables
CONCLUSION
This study examines the effect of incentive contracting on CTA perform-ance and volatility Evidence of structural changes in incentive compensa-tion is presented that points to a larger reliance on incentive fees as opposed
to management fees Management fees are shown to have no relationship with performance This is good news for investors, as the evidence seems to suggest that this type of compensation results in no systematic performance
TABLE 13.5 Estimation of the Relationship between Compensation Parameters and CTA Mean Annual Compounded Returns and Standard Deviation of Returns,
Augmented Specification
Performance
Variable Coefficient S.E Coefficient S.E.
**Significant at the 5 percent level for H0= 0.
**Significant at the 1 percent level for H0= 0.