ICI Submits Supplementary Comments on Regulatory Impact Analysis Supporting Proposed DOL Fiduciary Rule (pdf)

December 1, 2015 Email: piacentini.joseph@dol.gov; e-ORI@dol.gov; e-OED@dol.gov Mr. Joseph Piacentini Director, Office of Policy and Research & Chief Economist Employee Benefits Security Administration U.S. Department of Labor 200 Constitution Avenue NW, Suite N-5718 Washington, DC 20210 Re: RIN 1210-AB32: Definition of the Term “Fiduciary”; Conflict of Interest Rule – Retirement Investment Advice/ZRIN 1210-ZA25: Proposed Best Interest Contract Exemption Dear Mr. Piacentini: This letter supplements previous letters from the Investment Company Institute (ICI) commenting on the Department of Labor’s (Department) Regulatory Impact Analysis (RIA)1 supporting the proposed Fiduciary Rule (proposed rule).2 Since the Institute filed its comment letters, two authors of a study cited heavily by the RIA filed a comment letter with the Department responding to our comments. Those authors argue that ICI misinterpreted their study. While we appreciate that the comment period has closed, we trust that the Department’s stated purpose of ensuring a full and fair public comment process will allow consideration of what we intend as helpful clarification of our analysis. 1 U.S. Department of Labor, Employee Benefits Security Administration, Fiduciary Investment Advice Regulatory Impact Analysis, (April 14, 2015), (“RIA”). 2 Notice of Proposed Rulemaking – Definition of the Term “Fiduciary”; Conflict of Interest Rule – Retirement Investment Advice (“Fiduciary Rule Notice”), 80 Fed. Reg. 21928 (April 20, 2015). Mr. Joseph Piacentini December 1, 2015 Page 2 of 13 Our original comment letter to the Department3 identified several significant flaws in the RIA. As we explained, rather than conducting its own investigation of current, publicly available data to assess how its rule might affect fund investors, the Department turned to academic studies in an effort to find evidence supporting its proposed rule. In our comment letter, we listed specific problems with the Department’s application of the findings in such studies. In particular, the RIA relied heavily on a study by Christoffersen, Evans, and Musto (CEM) in its analysis.4 As we argued in our comment letter, the Department misapplied the results in the CEM study, leading the RIA to vastly overstate the potential benefits of the proposed rule. Two of those authors, Christoffersen and Evans, recently filed a comment letter5 with the Department responding to ICI’s original comment letter. Christoffersen and Evans note that their letter is intended to “make clear that our results [i.e., the results in the CEM study] are correctly calculated and interpreted.” Christoffersen and Evans address three concerns6 that ICI raised in its comment letter about the Department’s interpretation and usage of the results in the CEM paper: • ICI pointed out that the CEM data stops in 2009 and the market has since changed. • ICI questioned the Department’s interpretation and use of a coefficient estimate in the CEM study that is a linchpin of Department’s RIA. • ICI pointed out that performance results in most of the academic studies cited by the RIA, including those in the CEM study, were not asset-weighted, and thus did not reflect whether investors were holding higher- or lower-performing funds. Having read carefully the responses to these points from Christoffersen and Evans in their latest comment letter, we find nothing which changes our view that the Department’s RIA misapplied the results in the CEM paper and other academic research, causing the Department to incorrectly 3 See Brian Reid and David W. Blass, letter to Office of Regulations and Interpretations, Employee Benefit Security Administration, U.S. Department of Labor (July 21, 2015) (“ICI Comment Letter”), available at http://www.dol.gov/ebsa/pdf/1210-AB32-2-00749.pdf. ICI has also filed a supplemental comment letter responding to issues surrounding the RIA that were raised at the Department’s hearings; that letter is referenced in Section 4 below. 4 See Susan E.K. Christoffersen, Richard Evans, and David K. Musto, “What Do Consumers’ Fund Flows Maximize? Evidence from Their Brokers’ Incentives,” Journal of Finance, vol. LXVIII, no. 1, February 2013, 201–235. 5 See Susan Christoffersen and Richard Evans, letter to Office of Regulations and Interpretations, Employee Benefit Security Administration, U.S. Department of Labor (“Christoffersen and Evans”), available at http://www.dol.gov/ebsa/pdf/1210- AB32-2-02766.pdf. 6 The letter by Christoffersen and Evans highlights a fourth concern, relating to whether front-load fees and residual loads are included in one of their regressions. We do not address that concern in this comment letter because it does not bear directly on the Department’s analysis in the RIA of the relative performance of broker-sold funds. Mr. Joseph Piacentini December 1, 2015 Page 3 of 13 conclude that investors using brokers are underperforming “by an average of 100 basis points per year.” As we explain in the remainder of this letter: 1. The mutual fund market has indeed changed, and broker-sold and no-load funds now compete head-to-head. 2. The Department did in fact misapply a regression coefficient taken from the CEM study. To be clear, this has very little to do with the results in the CEM study, which stand on their own merits. The issue instead is the Department’s misuse of the coefficient in its calculations of the economic impact of its proposed regulation. The coefficient measures the response of the excess return on a broker-sold fund to the fund’s “excess load,” which they measure as the difference between the front-load fee that a fund pays to a broker and the front-load fee that one would expect the fund to pay the broker, as predicted by regression results in the CEM study (which for clarity we call “residual load”). Thus, the magnitude of any fund’s “residual load” is typically much smaller than the front-load that a fund actually pays to brokers. The Department erred by applying the CEM coefficient to the front-load paid to brokers rather than to the residual load paid to brokers, inflating the Department’s estimate of the benefit of its proposed regulation. 3. The Department needed to weight the research findings taken from the CEM study by assets or sales of fund shares. This problem remains even though the CEM study properly adjusted for the levels of funds’ assets in its regressions. The CEM study, like most of the other academic studies the RIA cites, conducts its analysis at the fund level. The RIA seeks to convert this fund- level analysis into aggregate dollar effects on the total IRA assets invested through broker-sold funds. To translate fund-level findings into market-level dollar effects, the Department would need to weight the fund-by-fund effects predicted by the CEM regression by the asset levels or sales of those funds. The RIA did not do that. Consequently, the RIA allowed funds with very small assets or sales to have an outsized influence on its benefits calculations7—an influence that also tended to inflate the Department’s estimate of the benefit of its proposed regulation. With respect to (1), we note that a recent academic study suggests that the RIA’s assumption that broker-sold funds underperform by 100 basis points may be nearly six times too high. Problems (2) and (3) together imply that the RIA’s benefits estimates are 15 to 50 times too high. We also would like to address an additional issue. During the August 11, 2015 hearing regarding the Department’s fiduciary rule proposal, Employee Benefits Security Administration 7 ICI Comment Letter at 5. Mr. Joseph Piacentini December 1, 2015 Page 4 of 13 (EBSA) staff asked about using risk-adjusted measures of returns, such as Sharpe ratios,8 in calculating the relative performance of funds sold with or without loads. For completeness, Section 4 in this comment letter provides Sharpe ratios for broker-sold funds using recent data. 1. Data in the CEM Study End in 2009, but the Market Has Changed We noted in our original comment letter that many of the studies cited in the RIA used data for broker-sold funds going back well into the mid-1990s.9 Some of the academic studies cited in the RIA explain their findings that load funds underperformed no-load funds by conjecturing that the load and no-load markets were segmented and that the two types of funds did not compete directly. For example, the RIA cites an article by Diane Del Guercio and Jonathan Reuter, which argues that “the market for retail funds is segmented, catering to two distinct clienteles … [W]e hypothesize that retail mutual funds sold through intermediaries, or brokers, face a weaker incentive to generate alpha than mutual funds sold directly to retail investors.”10 In fact, the mutual fund market has changed dramatically in the past 15 years—a fact the RIA ignores. For example, we presented evidence that most broker-sold funds now offer at least one no-load share class, thus competing head-on with “traditional” no-load funds. Although the CEM study did not suggest that broker-sold and no-load fund markets are segmented, in their recent letter Christoffersen and Evans challenge the notion that the market has changed. They present a table showing assets in front-load fund share classes compared to those in no- load share classes, which is derived from ICI’s 2015 Investment Company Fact Book.11 For clarity, we repeat their table below (Figure 1). If anything, the evidence in Figure 1 confirms that the market has changed significantly in recent years. Since 2009, front-end load share classes have experienced outflows and lost market share, while no-load share classes have experienced strong inflows and gained market share. For example, from 2009 to 2014, the modest growth in assets in front-end load share classes—a 21 percent increase—was entirely the result of asset appreciation, as front-end load share classes experienced net outflows from 8 A Sharpe ratio is the return on a security less the risk-free rate of return, all divided by the standard deviation of that quantity. It is a widely-used approach for risk-adjusting returns. 9 ICI Comment Letter at 6. 10 See Diane Del Guercio and Jonathan Reuter, “Mutual Fund Performance and the Incentive to Generate Alpha,” Journal of Finance, vol. LXIX, no. 4, August 2014, 1673-1704. 11 See Investment Company Institute, 2015 Investment Company Fact Book, 55th ed., p. 107. Mr. Joseph Piacentini December 1, 2015 Page 5 of 13 2009 to 2014. As a result, the share of long-term fund assets held in front-load share classes fell from 25 percent in 2005 to 22 percent in 2009 and fell further to 16 percent in 2014. In contrast, from 2005 to 2014, the assets in no-load share classes rose 143 percent, from $3.4 trillion to $8.3 trillion, due to both asset appreciation and strong inflows. As a result, the market share of no-load share classes, which stood at 50 percent in 2005, rose to 55 percent in 2009, and rose even further to 64 percent in 2014. As ICI’s Fact Book explains, there are several reasons for this change in the market.12 The key point is that the rising demand for no-load funds led broker-sold funds to offer no-load share classes so they could compete directly with no-load funds. In 2005, assets in no-load share classes of broker-sold funds totaled $708 billion; the average broker-sold fund had 30 percent of its assets in no-load share classes; and these assets accounted for 10 percent of the assets of all long-term funds. By 2014, assets in no-load share classes of broker-sold funds had grown to $2,677 billion; the average broker-sold fund had 55 percent of its assets in no-load share classes; and these assets accounted for 20 percent of the assets of long-term funds. In short, segmentation between broker-sold and no-load funds no longer exists. These funds compete head-to-head with each other. This change in the market makes it imperative to look at recent data when analyzing fund performance.13 12 See Investment Company Institute, 2015 Investment Company Fact Book, 55th ed., p. 108. 13 In a recent paper, Jonathan Reuter revisits the performance of broker-sold and direct-sold mutual funds using distribution channel data that cover 2003–2012. He reports results that he interprets as indicating that “the average broker-sold fund has become more competitive with the average direct-sold fund.” He reports that the broker-sold funds underperformed direct- sold funds (measured across all types of actively managed funds excluding municipal funds and adjusting for 12b-1 distribution fees) by only 18 basis points over the period 2003–2012 on an asset-weighted basis. This is less than one-fifth of the 100 basis point underperformance assumed in the RIA. See Jonathan Reuter, “Revisiting the Performance of Broker-sold Mutual Funds,” working paper, Boston College and NBER, November 2, 2015, (“Reuter”) available at https://www2.bc.edu/jonathan-reuter/research/brokers_revisited_201511.pdf. Mr. Joseph Piacentini December 1, 2015 Page 6 of 13 Figure 1: Total Net Assets of Long-Term Mutual Funds Are Concentrated in No-Load Share Classes Billions of dollars, 2005–2014 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 All long-term mutual funds $6,865 $8,060 $8,914 $5,771 $7,797 $9,029 $8,940 $10,359 $12,316 $13,127 Load 2,313 2,630 2,795 1,722 2,185 2,352 2,176 2,362 2,658 2,655 Front-end load1 1,728 2,027 2,190 1,374 1,750 1,882 1,751 1,893 2,148 2,117 Back-end load2 271 241 204 102 98 78 50 39 32 24 Level load3 287 340 379 237 328 381 367 417 459 502 Other load4 17 15 10 7 8 8 7 11 10 11 Unclassified 9 8 13 2 2 4 2 2 8 3 No-load5 3,427 4,073 4,587 3,055 4,255 5,091 5,224 6,262 7,594 8,344 Retail 2,404 2,799 3,091 1,940 2,666 3,069 2,991 3,469 4,144 4,605 Institutional 1,023 1,274 1,496 1,116 1,589 2,022 2,233 2,793 3,450 3,740 Variable annuities 1,039 1,225 1,346 854 1,130 1,291 1,250 1,396 1,614 1,649 "R" share classes⁶ 85 132 186 140 226 296 290 339 451 478 1 Front-end load > 1 percent. Primarily includes Class A shares; includes sales where front-end loads are waived. 2 Front-end load = 0 percent and contingent deferred sales load (CDSL) > 2 percent. Primarily includes Class B shares. 3 Front-end load ≤ 1 percent, CDSL ≤ 2 percent, and 12b-1 fee > 0.25 percent. Primarily includes Class C shares; excludes institutional share classes. 4 All other load share classes not classified as front-end load, back-end load, or level load. 5 Front-end load = 0 percent, CDSL = 0 percent, and 12b-1 fee ≤ 0.25 percent. ⁶ “R” shares include assets in any share class that ICI designates as a “retirement share class.” These share classes are sold predominantly to employer- sponsored retirement plans. However, other share classes—including retail and institutional share classes—also contain investments made through 401(k) plans or IRAs. Note: Components may not add to the total because of rounding. Data exclude mutual funds that invest primarily in other mutual funds. Sources: Investment Company Institute and Lipper Mr. Joseph Piacentini December 1, 2015 Page 7 of 13 2. The Department Misused Findings from the CEM Study The ICI explained in its original comment letter that the Department incorrectly applied a result from the CEM paper.14 The Department states that “Every 100 basis points in load sharing paid to unaffiliated adviser reduces future returns by 50 basis points.”15 The Department concludes that “Projecting these results onto the current IRA marketplace suggests that load fund holders could be down over $10 billion per year in loads and underperformance as a result.”16 This conclusion is incorrect and is a misuse of the findings in the CEM study. Let’s explain. CEM used a “two-stage” regression approach to identify whether fund performance is linked to broker incentives. In the first stage, CEM regressed front-load fees paid to brokers on front-load fees paid by investors and an array of other independent variables, including fund assets. This first-stage regression demonstrated how much of the front-load fees paid to brokers could be explained by these independent variables. For each fund, CEM calculated a variable called “excess load,” which is in fact the residual from the first-stage regression—the unexplained portion of the load paid to brokers. For clarity, we call this “residual load.” The residual load could be positive (i.e., the actual load paid to brokers is larger than that explained by the regression) or negative. By construction, the mean, or average value, of the residual loads for all the funds in the regression is zero. CEM then calculate the “excess return” of each front-load fund, defined as a fund’s return in the subsequent 12 months relative to the average return for all funds in the same Morningstar fund investment category. In the second-stage regression, they regress the excess return of each front-load fund on the fund’s residual load fee and an array of other independent variables. They find that for every 1 percent in residual loads paid to an unaffiliated broker, a fund would tend to underperform by 0.4972 percentage points. The error that the Department made was that it applied the coefficient from this second regression (i.e., –0.4972) to the average load paid to unaffiliated brokers to estimate the amount of the underperformance due to “payments from the mutual fund to the broker.”17 The correct calculation would have applied the coefficient to the average residual load. This is significant because, as noted earlier, residual load fees are generally quite small relative to the total load fees paid to brokers. The 14 ICI Comment Letter at 15. 15 RIA at 98. 16 RIA at 98. 17 RIA at 90. Mr. Joseph Piacentini December 1, 2015 Page 8 of 13 result of the Department’s flawed calculation was that the RIA overstates the estimated benefits of the proposed rule by 15 to 50 times. To demonstrate this point, we applied the CEM framework to mutual fund data reported to the Securities and Exchange Commission (SEC) on Form N-SAR for the most recent four years available (2010 to 2013).18 Figure 2 plots the data for 2013. In the figure, fund “residual loads” paid to brokers average to zero. The largest residual load fee in the figure, which is +1.4 percent, is not even as large as the +1.5 percent value that the Department applied to the entire IRA market for the year 2013.19 Thus, the Department multiplied the CEM coefficient of –0.4972 by an extremely large value, one that does not even appear in current data, resulting in a vast overstatement of the proposed rules’s benefits. Finally, the negative relationship between the residual loads paid to brokers and the subsequent fund performance is tiny. Examining fund returns from 2011 through 2014, based on sales of front- load funds from 2010 to 2013, we find that the model explains only 2.49 percent of the variance in fund performance. The residual load variable, which seeks to capture the effect of paying brokers higher-than-expected loads, by itself explains 0.03 percent (three one-hundredths of 1 percent) of the variation in fund performance. This point emphasizes that any increase in performance to investors from the Department’s rule likely will be extremely small, as we discussed in our original comment letter. When the costs of lost advice or the increased cost of fee-based accounts are added to the economic impact, shareholders are likely to experience $109 billion in lower returns if the Department adopts its rule as proposed. 18 Like CEM, our analysis uses a “two-stage” regression approach. In the first stage, following CEM, we regressed front-load fees paid to brokers on front-load fees paid by investors and an array of other independent variables including fund (log) assets. The coefficient estimates and fit of the first-stage model, though based on data for the years 2010 to 2013, are in all their important aspects similar to those reported in CEM. Like CEM, we use this first-stage regression to create an “residual load fee” variable, which is the residual from the first-stage regression. In the second stage, we regress the excess return of each front-load fund—measured as that fund’s return in the coming 12 months relative to the return on that fund’s Morningstar category in the coming 12 months—against the fund’s residual load fee and an array of other independent variable similar to those used in CEM. The results in the second-stage regression are also in all their important elements very similar to those reported in CEM. For example, we find a coefficient estimate on the residual load fee paid to brokers (both affiliated and unaffiliated) of –0.48 percent, a bit larger in absolute size than their reported estimate of –0.34 percent (see Table V on page 226 of the CEM study). We find a coefficient estimate on the residual load fee paid to unaffiliated brokers of –0.64 percent, which implies an even larger effect than the –0.4972 coefficient reported in CEM. 19 RIA at 113, column (C) for the year 2013. Mr. Joseph Piacentini December 1, 2015 Page 9 of 13 Figure 2: Relationship between Fund Excess Returns and “Residual loads” Paid to Brokers, 2013 Source: ICI analysis of Form N-SAR data. 3. The Department Erred in Failing to Weight CEM’s Results by Fund Assets or Sales A third issue that we raised in our comment letter is how to take the findings in CEM and other academic studies and apply them to the IRA market as a whole. These studies examine how the performance of individual funds responds to various individual fund characteristics. In an effort to assess the aggregate dollar benefit of its rule proposal, the Department took a key coefficient from the CEM study (i.e., –0.4972)—a coefficient that relates individual fund performance to residual loads paid to brokers—and multiplied that coefficient by the unweighted average load that investors paid. The Department then asserted that the reduced performance it calculated applied to all assets in front-end load funds held in IRAs. That was an error. The results in the CEM study simply cannot be used in that manner. -4 0 -2 0 0 20 40 Fu nd e xc es s re tu rn (a nn ua l p er ce nt ) -3 -2 -1 0 1 Fund 'residual load fee' paid to broker (percent) Mr. Joseph Piacentini December 1, 2015 Page 10 of 13 In their recent letter, Christoffersen and Evans challenge the statements in our comment letter that their results were not asset-weighted. They note that they included fund assets in their key regressions and “effectively the variances in the regression are asset-weighted.” Our point, however, was not about the results in the CEM paper—it was about how the Department misapplied those results.20 The CEM study did not attempt to estimate the aggregate dollar impact from investors using broker-sold funds. Had the CEM study sought to do so—even though its regressions included assets as a right-hand variable—it too would have needed to sales-weight (or asset-weight) their results, such as by weighting the predictions for individual funds or through weighted least squares.21 What the Department needed to do to get the correct answer is to multiply the –0.4972 percent coefficient taken from the CEM study times the sales-weighted average residual load fee in the data. In effect, the Department took +1.50 percent as an approximation. Based on our application of the CEM methodology to more recent data, the sales-weighted average residual load fee—the figure the Department should have used—was actually 0.03 percent, or one-fiftieth of the +1.50 percent value the Department used.22 That implies that the Department’s benefits calculation is inflated 50-fold. According to the CEM approach, investors in funds paying higher-than-expected load fees (positive “residual loads”) to brokers would be more likely to experience underperformance. Thus, the Department could alternatively have focused only on funds paying positive residual loads. In that case, to arrive at the correct answer, the Department would have multiplied the –0.4972 percent coefficient taken from the CEM study times the sales-weighted average residual load fees paid to brokers when the residual load is positive. That value, not surprisingly, is larger than 0.03 percent—but still quite small, just 0.099 percent for 2013 data. Thus, even if one applies the CEM analysis in a light as favorable as possible to the Department, the RIA’s benefits calculations are still overstated more than 15-fold (1.50 percent/0.099 percent). 20 ICI Comment Letter at 5. 21 Reuter compares the performance of broker-sold to direct-sold funds using both asset-weighted least squares and by asset- weighting the performance of individual funds to arrive at aggregate asset-weighted estimates of performance. 22 In the N-SAR data for 2013, the simple average front load fee paid to brokers is 1.31 percent, a bit less than the 1.50 percent figure the Department used. The simple average front load paid to unaffiliated brokers was slightly less, 1.29 percent. Mr. Joseph Piacentini December 1, 2015 Page 11 of 13 4. How Would Risk-Adjusted Excess Returns Based on Sharpe Ratios Affect ICI’s Findings? The Department’s RIA stated that broker-sold funds were “underperforming” by an average of 1 percent. In our comment letter regarding the RIA, we pointed out that such underperformance could not be seen in recent data. As we noted, in recent years, broker-sold funds had outperformed their Morningstar averages on either an asset- or sales-weighted basis. We did find that front-end load share classess had underperformed compared to no-load share classes in recent years, but only by 7 basis points. During the August 11, 2015 hearing regarding the Department’s fiduciary rule proposal, EBSA staff asked whether ICI had adjusted for risk in our analysis of the relative returns of broker-sold funds. In fact, we did risk-adjust returns by comparing returns of broker-sold fund with the returns of funds in the same Morningstar category. As we pointed out in our supplemental comment letter to the Department,23 Bergstresser, Chalmers, and Tufano24 had used Sharpe ratios—a common measure of risk-adjusted returns—to assess fund performance, and found no difference between broker-sold funds and no-load funds. The Department may nevertheless be interested in knowing how ICI’s conclusions would be altered if we examine Sharpe ratios for a more recent period. For completeness, we present Sharpe ratios for fund excess returns calculated using recent data. Figure 3 presents Sharpe ratios for front-end load and no-load share classes for the years 2011 to 2014. In terms of simple averages, Sharpe ratios on front-end load share classes were lower than those on no-load share classes—implying that front-end load share classes underperform no-load share classes on a risk-adjusted basis when returns are not weighted by funds’ assets or sales. A closer look shows that front-end load shares classes of equity and bond funds underperform on a simple-average basis, while those of hybrid funds outperform. But when returns are weighted by fund sales—an approach that better reflects where investors actually are putting their dollars—Sharpe ratios were higher on front-end load than no-load share classes. Among equity funds, Sharpe ratios were identical between front-end load and no-load share classes when weighted by sales. Sharpe ratios on bond and hybrid front-end load funds were higher than 23 See Brian Reid and David Blass, letter to Office of Regulations and Interpretations, Employee Benefit Security Administration, U.S. Department of Labor (September 24, 2015), available at https://www.ici.org/pdf/15_ici_dol_ria_comment.pdf. 24 See Daniel Bergstresser, John Chalmers, and Peter Tufano, “Assessing the Costs and Benefits of Brokers in the Mutual Fund Industry,” Review of Financial Studies, 22, no. 10, 2009, 4129-4156. Mr. Joseph Piacentini December 1, 2015 Page 12 of 13 those of their no-load counterparts. Thus, front-end load bond and hybrid funds outperformed their no-load counterparts, while the performance of equity funds was identical for the two sales channels. Figure 3: Sharpe Ratios of Front-end Load and No-load Share Classes Monthly data, 2011–2014 Fund type Front-end load No-load Simple average 0.24 0.27 Equity 0.24 0.29 Bond 0.24 0.26 Hybrid 0.25 0.23 Sales-weighted average 0.28 0.27 Equity 0.24 0.24 Bond 0.32 0.30 Hybrid 0.28 0.25 Note: The Sharpe ratio is calculated by first taking the mean of the difference between monthly share class return and the monthly return on the four-week U.S. Treasury bill across the 2011–2014 period. This is then divided by the standard deviation of the monthly share class return across the same period. Lastly, this value is then averaged across all front-end load or no-load share classes to create a simple average. For front-end load share classes, the sales-weighted average weights the aforementioned Sharpe ratio for each front-end load share class by the fund's total prior fiscal year N-SAR sales subject to a load. For no-load share classes, the sales-weighted average weights the aforementioned Sharpe ratio for each no-load share class by the share classes previous year ICI gross sales. Sources: Investment Company Institute, Strategic Insight Simfund, Lipper, and Morningstar 5. Summary In sum, we continue to believe that the Department should redo its cost benefit analysis, relying on its own independent analysis of recent publicly available data. We continue to believe the Department misunderstood and misapplied the results in a key study it relied on, that of Christoffersen, Evans, and Musto (2013). We see nothing in the recent comment letter by Christoffersen and Evans to the Department regarding our analysis that convincingly suggests otherwise. We would appreciate the opportunity to meet with you in person to resolve any remaining questions you have with our analysis. Given the importance of the matter, we would like to meet in the near future. We will be contacting you to schedule a meeting. In the meantime, do not hesitate to contact either of us with any questions. Mr. Joseph Piacentini December 1, 2015 Page 13 of 13 Sincerely, /s/ Brian Reid /s/ Sean Collins Brian Reid Sean Collins Chief Economist Senior Director Investment Company Institute Investment Company Institute cc: The Honorable Phyllis Borzi Assistant Secretary Employee Benefits Security Administration U.S. Department of Labor The Honorable Timothy Hauser Deputy Assistant Secretary Employee Benefits Security Administration U.S. Department of Labor G. Christopher Cosby Office of Policy and Research, Division of Regulatory Policy Analysis U.S. Department of Labor, Employee Benefits Security Administration Office of Regulations and Interpretations Employee Benefits Security Administration Room N-5655 U.S. Department of Labor Office of Exemption Determinations Employee Benefits Security Administration Attention: D11712 U.S. Department of Labor