The World Bank Economic Review, 31(1), 2017, 241–270
doi: 10.1093/wber/lhv065
Article

On the Impact of Regulating Commissions:
Evidence from the Indian Mutual Funds Market
Santosh Anagol, Vijaya Marisetty, Renuka Sane, and
Buvaneshwaran Venugopal
Abstract
Commissions-motivated agents have historically helped the development of many markets, but research sug-
gests brokers motivated by commissions sometimes steer consumers towards inappropriate products. This is-
sue is particularly important in household ﬁnancial markets where consumers may be unable to evaluate prod-
ucts on their own. While reforms attempting to limit commission payments have been undertaken worldwide,
little research has evaluated the impact of these reforms. We study a major Indian investor protection reform
that attempted to reduce commissions tied to mutual fund sales by banning the distribution fees that mutual
funds had previously earmarked for commissions. We analyze the policy impact by comparing funds charging
high versus low distribution fees pre-reform and ﬁnd no evidence that the reform itself reduced fund ﬂows. We
argue that the most plausible explanation is that the Indian asset management industry maintained substantial
commissions to brokers through other revenue sources apart from the banned distribution fees.
JEL classiﬁcation: O16, G18
Key words: O16: Economic Development: Financial Markets, Saving and Capital Investment, G28: Financial
Institutions and Services: Government Policy and Regulation

Santosh Anagol (corresponding author) is an assistant professor at the Wharton School of the University of Pennsylvania;
his email address is: anagol@wharton.upenn.edu. Vijaya Marisetty is an associate professor at the Royal Melbourne
Institute of Technology; his email address is: vijayabhaskar.marisetty@rmit.edu.au. Renuka Sane is a visiting scientist at the
Indian Statistical Institute; her email address is: renukas@gmail.com. Buvaneshwaran Venugopal is a PhD student at the
C.T. Bauer College of Business at the University of Houston; his email address is: bgvenugopal@uh.edu. We thank Shawn
Cole, James Choi, Ruchi Chojer, Mark Duggan, Fernando Ferreira, Keith Gamble, Ginger Jin, G. Sethu, Saikat Deb, Sayee
Srinivasan, K. N. Vaidyanathan, Shing-Yi Wang, and participants at Wharton, the IGIDR Emerging Markets Finance
Conference, the 2013 American Economic Association, and the NBER Law and Economics Working Group for valuable
feedback. We thank ACE Fintech Pvt Ltd for providing access to the Indian mutual funds data and Karvy Computershare
Limited for access to disaggregated data on mutual fund flows. We thank the Center for the Advanced Study of India at the
University of Pennsylvania and Wharton for funding. Minkwang Jang, Maria Gao, Mengshu Shen, and Jason Tian provided
excellent research assistance. This paper was earlier circulated as: “Are Mutual Funds Sold or Bought? Evidence from the
Indian Mutual Funds Market.” A supplemental appendix to this article is available at http://wber.oxfordjournals.org.

C The Author 2016. Published by Oxford University Press on behalf of the International Bank for Reconstruction and Development / THE WORLD BANK.
V
All rights reserved. For permissions, please email: journals.permissions@oup.com
242 Anagol et al.

The financial crisis of 2008 spurred an active worldwide policy debate on the optimal way to pursue con-
sumer financial protection. Two main policy options have received attention. One approach is to empower
consumers to make better financial decisions through financial literacy training and disclosure regulation.
A growing literature has evaluated these financial literacy training programs, with mixed results.1
A second policy approach is direct intervention into how financial products are sold, such as capping
or banning commissions to financial product brokers. A small but growing literature suggests that
brokers may not provide advice in the best interest of their consumers and that therefore regulating com-
missions might improve consumer welfare (Anagol, Cole, and Sarkar 2012; Gine ´ , Martinez Cuellar, and
Mazer 2014; Mullainathan, Noeth, and Schoar 2012). On the other hand, regulations that limit the
incentives of brokers to sell products may slow the growth of financial markets, reducing the amount of
productive capital in the economy and causing households to invest in suboptimal instruments.
Historically, brokers have played an important role in the development of many major household finan-
cial markets (Zelizer 1983).2 In describing the development of the life insurance industry in the United
States, Burton Hendrick noted in a 1906 article in McClure’s magazine: “Men do not insure of their
own free will. They must be clubbed into it. The company that employs no agents does no business.”
A number of significant reforms have already been made. The Financial Services Authority in the
United Kingdom implemented a ban on commissions paid to independent financial advisors by financial
product providers as of January 1, 2013 (Collinson 2012). Australia implemented a similar ban on com-
missions in July 2013 (Bowen 2011). The United States is yet to implement commissions reform, although
there is an ongoing policy debate on whether mutual funds should be allowed to charge a separate class of
operating expenses, known as 12b-1 fees, specifically for the purpose of paying distribution expenses such
as broker commissions (Ferris and Chance 1987; Walsh 2004).3 However, there currently exists little
empirical evidence on how regulations that affect brokers affect the growth of financial product markets.
These issues are of particular importance in developing countries, where the transition of households
from informal financial products, such as physical savings in the form of land, jewelry, and livestock,
towards formal financial products, such as bank accounts and equity products, is an important part of
the process of economic development. In India the amount of savings in formal financial assets is
approximately 8 percent of GDP compared to financial asset savings of 384 percent of GDP in the
United States (Reserve Bank of India 2012; Federal Reserve Board 2015). There is substantial opportu-
nity for savings in formal financial products to grow in the Indian economy.
There is also good reason to believe that commissions-motivated brokers will play an important role in
the growth of savings in formal financial products in India. For example, in the mutual fund industry (the
formal savings product studied here), individual investors own 84 percent of assets in equity products
(Association of Mutual Funds in India 2015b). This fraction is quite similar to the fact that 89 percent of
mutual funds in the United States are owned by individuals (Investment Company Institute 2015).4 An

1 Carlin and Robinson (2012), Cole, Sampson, and Zia (2011), Hastings and Mitchell (2011), and Cole and Shastry
(2010) present evaluations of financial literacy programs in a variety of contexts. For a recent review of financial literacy
programs, see Hastings, Madrian, and Skimmyhorn (2013).
2 It is perhaps not surprising that brokers play an important role in the development of formal financial product markets.
Experimenting with new financial products is likely to be costly, so an investor may prefer to get the advice of a trusted
broker before making an investment. For evidence on trust playing a role in the demand for rainfall insurance in India
see Gaurav, Cole, and Tobacman (2011). Brokers may also lower the investors’ transactions costs, provide solace in
case the investment does poorly, and provide guidance on when to exit the investment.
3 Walsh (2004) references a 1995 Investment Company Institute report that shows 63 percent of 12b-1 fees are used for
compensation of broker-dealers and related expenses.
4 We were unable to find statistics on what fraction of pure equity mutual funds are owned by individuals in the United
States, but this 89 percent number is based on equity, hybrid, and bond funds that are specifically aimed at individuals
(i.e., it excludes the money-market funds that corporations use for cash management.
The World Bank Economic Review 243

important difference, however, is that a much larger fraction of Indian mutual fund investments come via
commissions-motivated brokers; in 2015, for example, 91 percent of inflows into equity mutual funds in
India came through brokers (Association of Mutual Funds in India 2015a), while only approximately 40
to 50 percent of US inflows comes through brokers (Investment Company Institute 2015; Christoffersen,
Evans, and Musto 2005; Bergstresser, Chalmers, and Tufano 2009). The main reason for this difference is
that in the United States much of the equity mutual fund investments made by individuals come through
employer sponsored retirement plans, where there is no broker intermediating the transaction. In India,
however, pension funds such as the Employees Provident Fund are permitted to invest very limited
amounts in equity markets. The National Pension System also allows limited equity investments, but is cur-
rently very small, with only 10 million subscribers and USD 15 billion in assets as of July 2015 (National
Pension System Trust 2015) (relative to the approximately 500 million total workers in India, or even the
40 million workers in the formal sector (Ministry of Labour and Employment 2015)). Given that brokers
are particularly important for financial intermediation in the Indian context, policies that affect broker
incentives could have large implications for the growth of the productive capital stock as a whole.5
This paper provides the first estimates of the impact of a major policy reform aimed at reducing com-
missions paid to brokers on the mutual fund market in a developing country. We evaluate a policy change
that occurred on August 1, 2009, in which the Securities Exchange Board of India (SEBI) banned entry
loads charged by all mutual fund firms operating in India.6 Prior to this reform, mutual fund firms used
entry loads primarily to pay commissions to brokers who sold their products. The stated goal of the policy
was to “empower the investor in deciding the commission paid to the distributors and also to ensure trans-
parency in commissions paid for mutual fund products.”7 In practical terms, the regulator envisioned a
financial intermediation process where the investor paid the broker directly for advice, instead of mutual
fund companies paying brokers commissions to sell their products. The intent of this reform was to reduce
the ability of mutual funds to pay broker commissions; funds could no longer use the immediate income
earned from an investor’s entry load to pay out a commission to the broker who made the sale.8
Our primary empirical strategy is to compare how the entry load ban affected funds that previously
charged high entry loads to how the entry load ban affected funds that previously charged low entry
loads. We present a simple conceptual framework that makes predictions on how this kind of reform
might affect the relative attractiveness of the high versus low entry load funds. The main idea in the
framework is that the effect of the reform is likely to differ for investors based on how much the investor
is influenced by brokers (following the literature, we term investors that are heavily influenced by com-
missions as less sophisticated). The framework describes how less sophisticated investors would be
expected to reduce investments in high entry load funds after the reform because brokers no longer have
the incentive to encourage these investments. Consistent with this idea, the Indian mutual fund industry
has argued strongly that the entry load ban has reduced flows by limiting the ability of fund companies

5 See Levine and Zervos (1998) and Bekaert, Harvey, and Lundblad (2001) for cross-country studies suggesting that stock
market development is associated with overall economic growth.
6 An entry load, typically called a “front-end load” in the United States, is a fee that is calculated as a percentage of the to-
tal investment made in the mutual fund. Entry loads are immediately deducted from the customer’s investment at the
time of investment.
7 For newspaper accounts of the importance of entry loads as the primary source of commissions, see (1) “MFs Look For
Life Beyond Entry Load Ban,” Times of India, July 19, 2010 (2) “Mutual Fund Industry Struggling to Woo Retail
Investors,” Business Today, February 2011 Edition.
8 For example, suppose an investor invested one hundred rupees in an Indian mutual fund prior to the law change.
Typically a mutual fund would take 2.25 rupees out of this as an investment as an entry load and pay it to the broker
who sold the product. The entry load ban prevents mutual fund companies from taking any of the investor’s initial in-
vestment as revenue; the full amount of the investment must be invested in the fund on the investor’s behalf. Thus, if the
mutual fund company wants to pay the broker a commission, they must use other sources of funds to do so.
244 Anagol et al.

to pay commissions.9 The framework also notes, however, that for highly sophisticated investors, the
entry load ban may actually induce shifting of investments away from low entry load funds towards
high entry load funds because the upfront cost of these funds are now lower.
We analyze a newly assembled monthly panel dataset on Indian mutual funds from April 2006
through June 2012. On average, funds in our high entry load group had their entry loads decreased from
2.22 percent to zero, whereas funds in our low entry load group only had their loads reduced from .42
percent to zero. This policy change, therefore, constitutes a large change in the relative attractiveness of
these two groups of funds for both brokers and consumers and could potentially have impacts on both
groups of funds. In our main sample of funds we find that funds that charged higher entry loads prior to
the policy do not experience relatively lower flows after the entry load ban.10 Our results persist when
we use a smaller group of index funds which are less susceptible to the criticism of the funds being differ-
ent prior to the policy change.
The main empirical results suggest that the entry load ban did not lead to a major net drop in flows to
high entry load funds relative to low entry load funds. We argue there are two, nonexclusive, mecha-
nisms for this result. One possibility, which we find some auxiliary evidence for, is that fund companies
increased other types of commissions to at least partially offset the decline in commissions induced by
the entry load ban. This result is important because it suggests that indirect efforts at reducing commis-
sions (such as the entry load ban) may be of limited use in practice as financial product companies inno-
vate to find other ways of paying commissions. A second explanation is that the Indian mutual funds
market, at least during this time period, did not have a large number of low sophistication investors who
were highly influenced by brokers (despite the claims of fund companies). This result is interesting in
that it suggests that commissions are not always important drivers of fund flows; understanding under
what market conditions commissions are particularly important for driving investment behavior appears
to be an interesting area for future research.
We lastly discuss why overall net flows into Indian mutual funds declined in the post-reform period.
Using newly available, nationally representative data on household financial decisions in India, we show
that there has been a substantial increase in investments in gold and real estate in the period after the
entry load ban. We argue that Indian investors, similar to investors around the world, have generally
shifted their money away from financial assets, such as mutual funds, towards real asset classes in
response to the financial crisis of 2008 and would have likely done so even in the absence of the entry
load ban.
Our paper contributes to the literature on understanding the demand for formal financial products in
developing countries. Prior work has focused on financial literacy (Cole, Sampson, and Zia 2011;
Gaurav, Cole, and Tobacman 2011; Cai and Song 2012; Song 2015), product price (Cole, Sampson, and
Zia 2011; Gaurav, Cole, and Tobacman 2011), or other demand factors such as the availability of infor-
mal substitutes, wealth, or risk aversion (Mobarak and Rosenzweig 2012; Gine ´ , Townsend, and Vickery
2008) as determinants of formal financial product demand. Our paper is the first in this literature to
focus on the role of brokers in the distribution process. This prior work has also focused on a limited set
of formal financial products (mainly bank accounts, rainfall insurance, and pensions); our paper expands
this set to mutual funds.
This paper is also, to our knowledge, the first to analyze a major policy intervention that attempted
to limit the ability of financial product providers to pay commissions to brokers. We believe our results

9 Prior work on the impact of sales loads on fund flows also has generally found that higher sales loads correlate with
greater fund flows, although none of this work has looked at natural experiments where loads were exogenously low-
ered by a government policy. See Christoffersen, Evans, and Musto (2013) for a recent example of this work as well as a
review of previous research.
10 Net flows are the percentage growth in the fund after taking out growth due to the appreciation of assets.
The World Bank Economic Review 245

will be useful for other policymakers considering regulating mutual fund distribution fees. In particular,
a policymaker that only observed the aggregate data on fund flows in India might conclude that the entry
load ban did have a large negative impact on flows; we conducted a Lexis-Nexis newspaper search of
Indian newspapers mentioning “entry-load ban” and found that sixty out of ninety-seven articles sug-
gested that the entry load ban played an important causal role in the decline in flows in the post entry-
load ban period.11 Our results suggest that the true effects of the policy were more subtle: mutual fund
firms responded to the entry load ban by using revenues from other sources to maintain substantial com-
missions, and these post–entry load ban commissions may have been high enough to undo any major
effects from the entry load ban itself. Our paper highlights the importance of carefully thinking through
supply-side responses when attempting to regulate commissions, as there appear to be strong incentives
for asset management companies to pay commissions to brokers.

I. The Indian Mutual Funds Industry and the Entry Load Ban
Indian mutual fund assets in 2009 amounted to approximately US$90 billion.12 Assets under manage-
ment in India have seen a real growth rate more than double that of the growth rate of assets under man-
agement in the United States (12 percent average annual real growth in assets under management in the
Indian mutual fund industry since 1997, versus 5.3 percent real average annual growth in the United
States).13 There are approximately 10 million mutual fund investors in India (Halan 2010) and about
forty asset management companies. Assets in Indian equity-oriented mutual funds constitute approxi-
mately seven percent of the market capitalization of the Bombay Stock Exchange. A large fraction of
mutual fund sales comes through a network of thousands of mutual fund brokers (Kamiyama 2007).14
There are approximately 92,000 such brokers in India, and in 2015, they mobilized 91 percent of new
assets under management to equity mutual funds (Association of Mutual Funds in India 2015a). In addi-
tion to mutual fund brokers, there are approximately 2.5 million insurance agents in India, some of
whom also sell mutual fund products.
Mutual funds in India have historically charged three types of fees to investors. The first are entry
loads, which is a percentage of the initial investment the investor makes in the fund (prior to the reform
we study these entry loads were often set at 2.25 percent of the initial investment). The second are exit
loads, which are a percentage fee deducted at the time an investor exits the fund. The third type of fee is
called the management expense ratio. The management expense ratio is a percentage removed from the
value of the fund on an annual basis, to be used by the fund company for the following purposes: (a)
investment fees, (b) advisory fees, and (c) recurring expenses including custodian fees, audit fees, market-
ing expenses, brokerage (including commissions), and other miscellaneous expenses.
There are two types of commissions that asset management companies in India pay to brokers.15 The
first are “up-front” commissions, which are commissions that are paid by the mutual fund company to
the broker at the time of investment. In the period before the policy reform we analyze, the sales process
typically worked as follows. An individual investor would pay the amount they wanted to invest, say
one hundred rupees, to the broker. The broker would transfer this whole amount to the mutual fund
company, which would then deduct the entry load fee (typically 2.25 percent) and invest the remaining

11 Twenty-four articles did not make a statement about whether the entry load ban affected flows, and seven articles only
discuss positive aspects of the entry load ban like lower fees for investors.
12 The India Rupee / United States dollar exchange rate taken from finance.yahoo.com on Monday, October 26, 2009.
13 Growth rates of assets under management calculated from monthly reports of the Association Mutual Funds in India
monthly reports.
14 In the Indian context these brokers are referred to as “Individual Financial Advisors” (IFAs).
15 Throughout the paper we use the term “fees” to refer to payments made by investors to the mutual fund company and
“commissions” to indicate payments made from the mutual fund company to brokers who sell funds.
246 Anagol et al.

97.75 rupees in the mutual fund. The 2.25 rupee entry load fee would then typically be sent back to the
broker as a commission. While most funds paid the broker an amount equal to the entry load, funds
were allowed to save revenue from entry loads over time and use those monies to pay commissions in
the future.
The second type of commission paid by asset management companies is a “trail” commission, which
is paid annually to the broker as compensation for the investor remaining invested in the mutual fund.
As an example, consider an investor who invests one hundred rupees in a fund that paid a 1 percent trail
commission on January 1, 2008, and suppose the value of this investment had appreciated to two hun-
dred rupees on January 1, 2009. In this case the broker who sold the investor this fund would receive a
trail commission of two rupees on January 1, 2009 (the trail commission is based on the value of the
investment at year end). Note, however, that if this investor were to exit on December 31, 2008, the
broker would not receive any trail commission. Trail commission rates are typically set separately for
the first year and then all other years after the first year. Trail commissions are paid out of management
expense ratio fees, which are deducted from the net asset value of the fund annually.16
On June 30, 2009, the Securities and Exchange Board of India (SEBI) announced that starting on
August 1, 2009, mutual funds would no longer be allowed to charge entry load fees. Specifically, the law
had three components:
1. Funds could not deduct any amount of an investor’s initial investment (entry load ban)
2. Funds could charge up to 1 percent of redemptions (“exit loads”) and use these to pay up-front
commissions
3. Brokers had to disclose any commissions paid to them
The most important impact of this policy is that it directly reduced the source of funds that mutual
fund companies could use to pay up-front commissions (item (1)). Funds were no longer allowed to
charge the entry load fee from investors directly, at the time of investment, and use those fees to pay up-
front commissions. Instead, funds had to rely on three other funding sources to pay commissions. First,
as noted in item (2), funds could use fees collected by investors at the time of exit to meet commission
costs. Second, funds could use any monies saved from entry and exit loads charged prior to the ban to
meet current commission costs.17 Third, funds could use revenues from the management expense ratio
an investor would pay over the course of their investment to pay an up-front commission to the broker.
Note that in the case of using this third source of funding, the asset management company would have
to finance the up-front commission out of their own balance sheet and be compensated for this cost later
as they collected the annual management expense ratio fees from the investor over time.
SEBI had two primary motivations for enacting this policy. First, SEBI argued that investors were gen-
erally not aware of entry loads and thus made suboptimal decisions when choosing mutual funds.18
SEBI envisioned a market where the customer would directly pay the broker for advice, instead of the

16 Because trail commissions are taken out of the net asset value of the fund, all investors, even those who invested di-
rectly in the fund without a broker, are charged these commissions. The trail commissions deducted from a direct inves-
tor’s investment are kept by the mutual fund company. In the case where an investor switches brokers before the full
year is completed, neither the old or new broker keeps the trail commission; instead, the trail commission is kept by the
mutual fund company.
17 These funds are called the fund’s “unutilized load balance.” Fund companies were allowed to use as much of this unu-
tilized load balance (including unutilized entry and exit loads) to pay commissions as they wished after the policy
change. On March 9, 2011, SEBI instituted a rule that funds were not allowed to use more than one third of the unutil-
ized load balance as of July 31, 2009, for commissions in any given year (Securities and Exchange Board of India,
2012b).
18 In essence, SEBI argue that entry loads are a shrouded fee in the sense of Gabaix and Laibson (2006), Heidhues,
Ko}szegi, and Murooka (2012), and Anagol and Kim (2012).
The World Bank Economic Review 247

broker being compensated by the mutual fund company via a commission financed by the entry load.19
Second, SEBI believed that entry load based commissions gave brokers an incentive to encourage invest-
ors to move in and out of different mutual fund investments too frequently; removing entry loads and
the associated upfront commissions would remove the incentive for brokers to “churn” investors in this
manner.20
It is important to note that, although the policy banned mutual funds from charging entry load fees,
which were commonly used to finance upfront commissions, the policy change did not explicitly ban
upfront commissions. To our knowledge, there is no publicly available information on the upfront and
trail commissions agents would earn for selling specific mutual funds after the entry load ban, and there-
fore it is difficult to know how exactly the reform changed the commission structure on average. Based
on private conversations with one major fund house regarding twenty-five of their popular funds,
appendix figure S1.1 presents what a typical agent might earn in upfront, trail, and total commissions
based on how long a hypothetical one hundred rupee investor chooses to hold high and low entry load
mutual funds before and after the entry load ban.21 High entry load funds are defined as those charging
2.25 percent or more in entry loads, and low entry load fees are defined as those charging less than 2.25
percent (typically these charged less than 1 percent). We report these estimates for the period before the
reform (June 2009), immediately after the reform (June 2010), and five years after the reform (June
2014).
For high entry load funds, upfront and trail commissions were typically reduced in the year after the
reform, with upfront commissions declining by .4 percent and trail commissions declining by approxi-
mately .5 percent. Over the longer run, upfront commissions continued to decline to 1 percent, whereas
trail commissions were actually increased likely to offset the lower payment of upfront commissions.
Although we do not have access to investor level information that would allow us to directly estimate
the holding duration of investors, Shah et. al. (2010) find that 44 percent of retail assets under manage-
ment had been held for longer than two years, and 64 percent longer than one year. This suggests that
offsetting increases in trail commissions could have meaningfully dampened the effect of the entry load
ban on upfront commissions. We return to this as a potential explanation of our findings after presenting
our main results. For low entry load funds (defined as those charging less than 2.25 percent), a key dif-
ference is that upfront commissions did not change due to the reform as they were typically zero to begin
with. Trail commissions for low entry load funds have also been increased over the last five year period.
While our analysis of the reform primarily focuses on the entry load ban and its associated impacts
on commissions, we caveat that theoretically the disclosure regime also changed due to this reform, and
our data do not allow us to separately identify entry load ban versus disclosure effects. We choose to
focus on the entry load ban portion of this reform for the following reasons. First, anecdotal evidence
from our discussions with industry participants and newspaper accounts at the time of the reform almost
never mention the disclosure requirement, while there was intense discussion of the entry load ban after
the reform was passed. Second, although the reform required distributors to disclose commissions, there
was no enforcement mechanism proposed, so it is not clear whether the reform actually changed distrib-
utors incentives regarding disclosure.
Nonetheless, it is also worth noting that our empirical strategy of comparing high versus low entry
load funds would also likely pick up disclosure effects, as the requirement to disclose commissions at the

19 Although no systematic data exist on the prevalence of investors making direct payments to brokers for advice, anecdo-
tal evidence suggests that this has been very uncommon.
20 Studying the impact of the entry load ban on “churning” is an interesting avenue for future research. We do not address
this question in this paper as we do not have access to investor-level data to measure the amount of churning individ-
uals did before and after the policy change.
21 Appendix figure S1.1 is in the supplemental appendix available at http://wber.oxfordjournals.org/.
248 Anagol et al.

Figure 1. Net Flows into Equity Open-Ended Mutual Funds.

This ﬁgure presents the net ﬂows into equity open-ended mutual funds on the left hand side y-axis in crores of rupees (1 Crore
Rupees equals approximately 200,000 USD). The level of the Bombay Sensex Index (Indian stock market index) is shown as a solid
line and is measured on the right hand side y-axis. The dashed vertical line indicates the policy change that eliminated the ability of
mutual funds to charge entry loads. Authors’ analysis based on data described in the text.

time of sale would naturally affect high load funds more than low load funds. The fact that we find little
relative change in the flows to these types of funds after the reform suggests that the disclosure portion
of this reform had little impact.22
There has been an active policy debate surrounding the entry load ban policy since its inception in
August 2009. The policy was further tweaked in September 2011, allowing fund houses to charge the
customer one hundred rupees at the time of investment and pass this on to the broker as a commission.23
In August 2012, SEBI further changed the mutual fund fee policy with the goal of forcing funds to use
income from management expense ratios instead of exit loads to pay commissions. SEBI argued that
fund companies had an incentive to encourage investor exit under the current regime.24

II. Describing Indian Mutual Fund Flows
We begin our empirical analysis of the entry load ban by presenting data on the aggregate evolution of flows
into Indian mutual funds. Figure 1 plots net-flows to existing equity open-ended mutual funds around the pol-
icy change studied here. This data was obtained from monthly reports posted at the Association of Mutual
Funds of India (AMFI) website.25 The vertical dashed line indicates August 2009, the date of the ban on entry
loads. Overall, the pattern of net flows in the pre-period follows the level of the Sensex stock index (an index
of the thirty largest stocks on the Bombay Stock Exchange) more closely than in the post-reform period, and
flows into mutual funds appear to be lower in the post reform period versus the pre-reform period.26

22 We naturally do not interpret this result as suggesting that disclosure does or does not work, however, as it is unclear
whether the disclosure portion of the reform was actually enforced.
23 For new customers, fund companies could charge up to 150 rupees to pass on to the broker as commission.
24 For details on the August, 2012, reform see Securities and Exchange Board of India (2012a).
25 AMFI is the main trade organization of the Indian mutual funds industry.
26 Net flow, inflow, and outflow estimates are adjusted for inflation using the India Wholesale Price Index and also for
general economic growth using the Reserve Bank of India’s Quarterly estimates of GDP at market prices (constant
price series) to allow for more meaningful comparisons over time.
The World Bank Economic Review 249

This pattern of net flows is consistent with a variety of underlying patterns in inflows and outflows,
so appendix figure S1.2 separately plots inflows and outflows. The plot of inflows shows that there are
still substantial inflows into mutual funds in the post–policy reform period. However, the inflows do not
seem to follow the market relative to how inflows followed the market in the prereform period. Unlike
inflows, outflows continue to follow the market closely in the post-reform period. These figures suggest
that the fall in net flows into Indian mutual funds in the post–entry load ban period has primarily been
due to a change in the amount of inflows, as opposed to an increase in outflows (relative to the stock
market). Another potential measure of investor participation in Indian mutual funds is the number of
retail mutual fund accounts. Appendix figure S1.3 presents the number of retail investor accounts. After
the 2009 ban, there was an approximately 9.4 percent decline in retail accounts.
Overall, the aggregate data suggest a decline in mutual fund growth after the entry load ban. It is
important to note, however, that the results in these plots conflate the impact of the entry load ban with
all other time-varying market conditions.

III. Conceptual Framework
In this section we provide a simple framework for thinking about how the entry load ban might affect
flows into high and low entry load funds after the reform. The theoretical literature on commissions has
typically focused on how brokers and commissions interact with consumers with different levels of
sophistication, so we consider the behavior of three stylized types of investors;27 those with low,
medium, and high levels of sophistication. Let i 2 ½L; M; H index these three types of investors.
Low sophistication investors are the most strongly influenced by brokers, in the sense that brokers
directly determine both how much these investors put into mutual funds overall, as well as which specific
funds they invest in. We assume brokers are interested in maximizing their commissions, and therefore
they encourage these investors to invest completely in high entry load funds. Let c be the amount of com-
missions paid when a broker sells one unit of the high entry load funds, and xL ðcÞ be a function that
gives the amount of investment in high entry load funds from low sophistication investors. The greater
the commissions available, the larger these flows will be, that is, @ xL ðcÞ=@ c > 0.
Highly sophisticated investors are not influenced by brokers at all. They choose the amount to invest
in mutual funds, as well as the allocation across high and low load funds. Assume these highly sophisti-
cated investors have heterogenous beliefs about the expected returns on high and low entry load funds,
and any given investor completely invests in either high or low entry load funds. Let rh À rl be the true
future return difference between high and low entry load funds. Sophisticated investors are distributed
uniformly on [–.5, .5] based on their expectations of rh À rl , that is, half the investors think that high
entry load funds will perform worse than low entry load funds, and half think they will perform better.
After fees, a sophisticated investor will invest in the high entry load fund if E½rh À rl > f , and the low
entry load fund otherwise (where f 2 ½0; :5 is the level of fees). Total investment in high and low entry
load funds is given by. 5–f and fþ.5, respectively.
Medium sophisticated investors are not influenced by brokers on the extensive margin, that is, they
choose how much to invest in mutual funds overall by themselves. However, brokers influence their per-
ceptions of returns on high versus low entry load funds; in particular, the higher commissions are, the
higher these investors believe rh À rl to be. We implement this idea by assuming their return expectations
are distributed uniformly on [–.5þc, .5þc]. When commissions c are zero, half these investors invest in
high and low entry load funds, respectively. When c>0 we have .5þc of these investors investing in high
entry load funds and .5–c investing in low entry load funds.

27 See Inderst and Ottaviani (2009) for discussions of consumer sophistication and the impact of commissions-motivated
sales agents.
250 Anagol et al.

0
Let indicate post-reform values of variables and vh and vl indicate flows into high and low entry
funds, respectively. We have then that the change in flows to high versus low entry load funds in the
post- minus pre-period is:28

ðv0h À vh Þ À ðv0l À vl Þ ¼ xL ðcÞ À xL ðc0 Þ þ 2½c0 À c þ f À f 0 (1)
0
L L
x ðcÞ À x ðc Þ < 0 is the decrease in flows to the high entry load funds because low sophistication
investors are less likely to be brought to the mutual funds market because commissions are lower.
0
2½c À c < 0 is the decrease in flows to high entry load funds from medium sophisticated investors; this
is multiplied by two because the decrease in flows to high entry loads leads to an equal sized increase in
flows into low entry load funds (i.e., the entry load ban has a spillover effect on the low entry load
0
funds). 2½f À f > 0 is the increase in flows to high entry load funds above the decrease in flows to the
low entry load group from highly sophisticated investors who believe the postfee returns on high entry
load funds are better after the entry loads are reduced.
The model suggests that the effect of the entry load ban on the change in flows to high entry load
funds after the reform relative to low entry load funds will depend on the sophistication levels of invest-
ors. Suppose the market is completely made up of low sophistication investors. In this case equation 1
0
simplifies to xL ðcÞ À xL ðc Þ < 0, and the entry load ban will have the effect of reducing flows to high
versus low entry load funds. Adding medium sophistication investors, equation 1 simplifies to
0 0
xL ðcÞ À xL ðc Þ þ 2½c À c < 0. In addition to the decline induced by the low sophistication investors,
marginal medium sophistication investors divert their investments from high to low entry load funds
as brokers have less strong incentives to encourage them into the high entry load funds. In this case
the entry load ban is changing the market both by reducing inflows overall, as well as inducing a spill-
over effect from the high entry load group to the low entry load group. Most of the popular press cov-
erage of this reform has focused on the case in our model with just low and medium sophistication
investors.
Adding highly sophisticated investors, however, can somewhat offset the decline in flows to high
0
entry load funds (2½f À f > 0). On the margin, some of these investors are now attracted to high entry
load funds after the reform because the lower fees induce them to believe that the postfee returns are
higher on high versus low entry load funds.
0 0
We now turn to our empirical approach which provides estimates of ðvh À vh Þ À ðvl À vl Þ and then
later discuss how our estimates can be interpreted in light of our conceptual framework.

IV. Empirical Analysis of Fund Level Data
We manually construct a new monthly data set of fund level net flows, assets under management, fees,
and other fund characteristics for the Indian mutual funds sector. For the period April 2006 through
September 2010, the AMFI website lists the average assets under management for each Indian mutual
fund in that month. From October 2010 through June 2012, average assets under management are
reported on a quarterly basis.29 We downloaded each of these listings and merged them over time to cre-
ate a panel data set of average assets under management for each fund in each month. This constitutes
our base sample of fund*month observations. In appendix section A.1 in the supplemental appendix
(available at http://wber.oxfordjournals.org), we detail the data sources used to construct all of the varia-
bles in our analysis.

28 Appendix table S1.1 illustrates this calculation.
29 For the period when only quarterly data on assets under management are available, we linearly impute values for the
months where data was not reported.
The World Bank Economic Review 251

We study two primary outcome variables to measure the impact of the entry load ban on fund
growth. We first present results on the standard measure of fund growth used in the literature,
NetFlowi;t , as defined in Sirri and Tufano (1998):
AUMi;tþ1 À ð1 þ Ri;t ÞAUMi;t
NetFlowi;t ¼
AUMi;t

Ri;t is the return earned on the securities held by the fund. AUMi;t is fund i’s assets under management
at time t.
This net flow measure displays a significant amount of noise over time, even when we average across
a large number of funds. The main issue appears to be that our assets-under-management measure is an
average of the assets under management in the fund within a month. However, our returns measure is
based on the return in the fund from the first day of the month to the first day of the next month. This
mismatch can lead to systematically over-estimated net flows in one month and underestimated net
flows in the next. Unfortunately, without daily data on assets under management, we are unable to
determine to what extent the noise in this measure is due to this measurement issue.
Another important issue with this definition of net flows is that, in cases where a fund has very small
total net assets in the prior period and large growth, it can produce very large net flow measures that are
not necessarily indicative of fund growth. We therefore choose a trimmed sample as our baseline sample,
where we remove the top and bottom one percent of observations in terms of net flows. In practice, this
means that funds with net flow growth rates of less than À88 percent or greater than 350 percent are
excluded from the sample. Appendix figure S1.4 shows the histogram of net flows after the top and bot-
tom one percent of observations have been trimmed.
To explore the robustness of our results, we also look at how assets under management have evolved
after the entry load ban. Assets under management within a fund change for two reasons. First, the value
of the existing assets in the fund changes based on the return earned on the securities within the fund.
Second, investors purchase and sell units of the fund. We show that the trends in returns earned on high
and low entry load funds were very similar throughout our whole period, and thus any comparison of
assets under management across these two groups essentially already controls for changes in returns
over time. The main advantage of the assets under management variable is that it is consistently and
cleanly measured for all of the funds in our sample and does not appear as noisy as the net flow measure.

Methodology
Our primary empirical methodology is to compare the impact of the entry load ban on funds that
charged high entry loads prior to the ban versus funds that charged low entry loads prior to the ban.
Figure 2 presents the distribution of month*fund observations across the levels of entry loads
observed in the data in the pre-reform period. The figure shows two important mass points, one at the
zero entry load point, and one at the 2.25 entry load point. For simplicity, we thus define two types of
funds prior to the reform. High entry load funds are defined as those funds that charged an average entry
load of 2.25 percent or higher prior to the reform.30 Low entry load funds are those funds that charged
an average entry load of less than 2.25 percent prior to the reform. We test whether high entry load
funds have attracted differentially more or less net flows after the imposition of the entry load ban.
Our results include all funds in existence prior to the reform. Funds that appear in our data but then
exit prior to the reform are categorized as high or low entry load based on the level of entry load they

30 We also categorize funds that charged an entry load of 2.25 percent for the majority of the periods prior to the reform
as being in the high entry load group. Thirty-two out of the 650 funds in the high entry load group fall into this cate-
gory; these funds charged an average of 2.1 percent entry loads so they are more similar to the high entry load group
than the low entry load group.
252 Anagol et al.

Figure 2. Distribution of Fund Observations Across Entry Load Levels.

This ﬁgure presents a histogram of monthly fund observations across the levels of entry loads charged in the pre-reform sample (all
entry loads were mandated to be equal to zero in the post-reform sample). Authors’ analysis based on data described in the text.

charged prior to the reform. We believe these funds are useful observations on how entry loads impacted
flows prior to the law change. However, it is not possible to categorize funds that were started after the
entry load ban as high or low entry load funds, as they were mandated by law to have zero entry loads.
5,372 fund*month observations are dropped for funds that were started after the entry load ban was
introduced in August of 2009.

The Impact of the Entry Load Ban on All Fund Flows
Table 1 presents summary statistics on high and low entry load funds from the beginning of our data
(April 2006) through the implementation of the reform in August 2009. The average entry load charged
by high entry load funds is 2.23 percent, whereas low entry load funds charged .48 percent. The differ-
ence in entry loads charged by high versus low entry load funds is statistically significant at the one per-
cent level; this is consistent with the idea that the entry load ban should have a stronger effect on high
versus low entry load funds. Appendix figure S1.5 plots the average entry load charged by funds in our
high entry load group and our low entry load group. The figure shows that average entry loads in the
high entry load group were essentially flat at approximately two percent prior to the reform and then
experienced a discrete and large drop to zero after the reform. In the low entry load group, the average
entry load was slightly declining over time prior to the reform.
The mean size of funds in the high entry load group was 46.5 million dollars, where as the mean size
of funds in the low entry load group was 25.2 million dollars (significantly different at the one percent
level). Net flows into low entry load funds were approximately twenty-one basis points higher than high
entry load funds, but this difference is not statistically significant.
The mean return in the high entry load group is ten basis points higher per month than in the low
entry load group although this difference is not statistically significant.31 We also test for whether the
high entry load funds had higher risk-adjusted returns by regressing the average difference in returns

31 The returns we report in this paper are net of annual operating expenses but do not include entry or exit loads. It is dif-
ficult to incorporate the affect of entry and exit loads on returns, as these loads will vary based on the investor’s dura-
tion of investment.
The World Bank Economic Review 253

Table 1. Summary Statistics: All Funds in the Pre-reform Period

Low entry load fund High entry load fund Difference

Entry load (%) 0.48 2.23 À1.75***
(0.68) (0.37)
Net ﬂow (1% Trim) 0.55 0.34 0.21
(9.67) (7.69)
Assets under management (rupees millions) 1258.65 2323.24 À1064.59***
(5230.14) (4138.55)
Assets under management (US dollars millions) 25.17 46.46 À21.29***
(104.60) (82.77)
Log(AUM(t)) 4.71 6.21 À1.50***
(2.41) (2.28)
Return(t) (%) 0.70 0.80 À0.09
(5.29) (8.67)
Sector adjusted return À0.02 À0.00 À0.02
(3.78) (3.32)
Management expense ratio 1.10 2.14 À1.04***
(0.61) (0.46)
Exit load ﬁrst 30 days 0.48 0.69 À0.21***
(0.45) (0.53)
Exit load 31 to 180 days 0.37 0.67 À0.30***
(0.44) (0.53)
Exit load 181 to 365 days 0.33 0.66 À0.32***
(0.45) (0.55)
Minimum investment (rupees) 7854.58 4486.80 3367.78***
(6698.30) (2361.40)
Minimum investment (US dollars) 157.09 89.74 67.36***
(133.97) (47.23)
#Funds 237 588
#Funds*month observations 5998 16723

Authors’ analysis based on data described in the text. This table presents summary statistics on low entry load funds and high entry load funds prior to the reform.
The high entry load group is deﬁned as any fund that charged an entry load of 2.25 percent or higher prior to the reform; the low entry load group is the set of funds
that charged entry loads less than 2.25 percent prior to the reform. The Difference column presents the difference in the mean of the variable across the two groups.
Conversions from Indian Rupees to US Dollars are done at ﬁfty rupees per dollar. Standard deviations of the variable are presented in parentheses below the mean.
The signiﬁcance stars in the Difference column indicate whether the mean value of the variable is signiﬁcantly different across the high and low entry load groups.
***, **, * indicate signiﬁcance at 1%, 5%, 10%, respectively.

across the high and low entry load group on a four factor model for India described here (Agarwalla,
Jacob, and Varma 2013). We find the a term is small and insignificant in these tests, suggesting that the
high entry load funds did not have higher risk-adjusted returns. Figure 3 plots Carhart four factor risk-
adjusted returns of these two groups of funds separately.32 The figure shows that, at least in terms of
risk-adjusted returns (arguably the most important product characteristic of a fund), there does not

32 We estimate risk-adjusted returns in the high and low entry load groups as follows. We first take the average return
within each group for all months we have data and subtract the monthly risk-free rate provided for the Indian market
from Agarwalla, Jacob, and Varma (2013). Within each group, we then regress this excess return on the market return,
a small market capitalization versus big market capitalization factor, a high book-to-market minus low book-to-
market factor, and a momentum factor. For details on the construction of these factors see Agarwalla, Jacob, and
Varma (2013). These factors can be downloaded at http://www.iimahd.ernet.in/ jrvarma/Indian-Fama-French-
Momentum/. Our risk-adjusted return is the a term from this regression within each group plus the estimated residuals
(i.e., the excess return minus the expected excess return). In both the high and low entry load groups, the a terms are
small and insignificant.
254 Anagol et al.

Figure 3. Risk-Adjusted Returns on High Vs. Low Fee Funds.

This ﬁgure shows the average monthly returns earned in the high and low entry load groups over time. The high entry load group
is deﬁned as any fund that charged an entry load of 2.25 percent or higher prior to the reform; the low entry load group is the set of
funds that charged entry loads less than 2.25 percent prior to the reform. The dashed vertical line indicates the date the policy went
into force (August 2009). We estimate risk-adjusted returns in the high and low entry load groups as follows. We ﬁrst take the aver-
age return within each group for all months we have data and subtract the monthly risk-free rate provided for the Indian market
from Agarwalla, Jacob, and Varma (2013). Within each group, we then regress this excess return on the market return, a small mar-
ket capitalization versus big market capitalization factor, a high book-to-market minus low book-to-market factor, and a momentum
factor. For details on the construction of these factors see Agarwalla, Jacob, and Varma (2013). Our risk-adjusted return is the a
term from this regression within each group plus the estimated residuals (i.e., the excess return minus the expected excess return).
Authors’ analysis based on data described in text.

seem to be important trend differences across high and low entry load funds prior to the entry load
ban.33 Thus, any difference we might see in fund growth across these two types of funds after the policy
reform are not driven by a major change in return performance after the policy change.
Table 1 also shows that high entry load funds are funds that generally charged higher fees overall;
average annual management fees were approximately 1 percent higher in the high entry load group. Exit
loads were also higher in the high entry load group, although the size of the difference in exit loads is
small. High entry load funds also had lower minimum investment requirements on average.
Table 2 shows the proportion of pre-reform observations that are in ten major categories of Indian
funds. In both groups, the most common type of fund are general equity funds that invest in a variety of
equity instruments. Sector funds are those that focus on specific sectors such as infrastructure, banking,
agriculture, etc. Balance funds are funds that invest a substantial portion of assets in debt and equities. It
is important to note that the allocation of the low entry load funds across these fund categories is sub-
stantially different from the fund categories in the high entry load group. There are two major differen-
ces between the distribution of low entry load and high entry load funds across these categories. First,
approximately 16 percent of the low entry load observations are index funds, while only 1.5 percent of
the high entry load funds are index funds. Second, 31 percent of the low entry load group are “Income”
funds. These are funds that primarily invest in debt securities but allocate a small (unobserved) propor-
tion to equities. We suspect that these funds have lower entry loads because they catered to more

33 Appendix figure S1.6 plots the mean monthly raw return for the high entry load and low entry load groups separately.
The results are similar.
The World Bank Economic Review 255

Table 2. Fund Categories by Entry Load Levels Prior to the Reform

Low entry load fund High entry load fund Difference

Index fund 0.16 0.01 0.15***
Tax fund 0.02 0.09 À0.07***
General equity fund 0.31 0.50 À0.19***
Large cap fund 0.01 0.03 À0.03***
Sector fund 0.02 0.18 À0.16***
Income fund 0.31 0.00 0.31***
International fund 0.00 0.04 À0.04***
Small and mid cap fund 0.00 0.08 À0.08***
Balance fund 0.15 0.06 0.09***
Gold fund 0.02 0.00 0.01***
#Funds 237 588
#Funds*month observations 5998 16723

Authors’ analysis based on data described in the text. This table presents the proportion of funds in broadly deﬁned categories in the low and high entry load
groups. See table 1 for further description.

sophisticated investors who were interested in avoiding fees, although we do not have data on investor
characteristics to test this.
Before presenting regression-based results on the impact of the entry load ban on fund growth in our
two groups of funds, we first present simple graphical evidence on how fund growth has evolved in these
two types of funds over time. The left panel of figure 4 plots the mean logarithm of assets under manage-
ment for our high entry load and low entry load groups over the period of our sample. As these two types
of funds have essentially the same trends in returns (figure 3), we can use changes in assets under man-
agement as a signal of how fund growth has varied for the two types of funds. The trends in log assets
under management in both groups prior to the reform are very similar. Both series are highest in early
2008, hit a bottom in mid 2009, and show a large increase in the few months before the policy ban.
Despite the fact that funds in the low entry load group were statistically and economically different from
the high entry load group along a number of observable characteristics in the pre-reform period, these
differences did not cause these two types of funds to have substantially different patterns of asset growth
prior to the reform. Given the similarity in trends prior the reform, we argue that it is unlikely that any
patterns we observe after the reform would be due to differential trends in the attractiveness of these
funds before the reform.
Figure 4 also shows the main result of our paper. After the policy reform in August 2009, there does
not seem to be a major decline in assets under management in the high entry load group versus the low
entry load group. Both groups appear to experience a small decline in assets under management in the
post-reform period. Note that, as shown in figure 3, monthly returns in the post reform period were gener-
ally positive, so the fall in assets under management for both types of funds implies substantial negative
flows out of mutual funds during the post-reform period. What is interesting, however, is the fact that
both high and low entry load funds experienced drops in asset growth. This result is inconsistent with the
hypothesis that the entry load ban had an important impact on fund growth (we formally test whether the
trends in flows in these two groups differ post-reform in section IV). If anything, the figure shows that high
entry load funds have grown more in the post-reform period versus low entry load funds; this result is con-
sistent with the behavior of the highly sophisticated investors described in section 4.
The right panel of figure 4 shows the average monthly net flows as calculated in Sirri and Tufano
(1998). The series is much noisier than the assets under management series, perhaps because of the mis-
match between our assets under management data and the returns data described earlier. As was shown
in the summary statistics, the mean net flow for both groups is close to zero, although there is substantial
256 Anagol et al.

Figure 4. Asset Growth in High vs. Low Entry Load Funds.

This ﬁgure presents the average log asset under management (left panel) and net ﬂows (right panel) in the high entry load fund
group (solid line) and the low entry load fund group (dashed line). The high entry load group is deﬁned as any fund that charged
an entry load of 2.25 percent or higher prior to the reform; the low entry load group is the set of funds that charged entry loads less
than 2.25 percent prior to the reform. The dashed vertical line indicates the date the policy went into force (August 2009). Authors’
analysis based on data described in text.

variation in net flow rates over time. Given the noisiness of this measure over time, it is difficult to visu-
ally compare the pre-trends using this outcome measure. One discernable pattern is that starting in early
2008, both groups see a decline in net flows (on average), and then both groups display an increase in
net flows starting in early 2009. After the reform, the figure suggests that high entry load funds have
received higher net flows than low entry load funds, although it is difficult to conclude anything based
on visual inspection of these averages.
The World Bank Economic Review 257

All Funds: Empirical Results
While the figures suggest that the entry load ban was not a major cause of the Indian mutual fund indus-
try’s negative net flows in the post reform period, tables 1 and 2 did show a number of differences across
the high and low entry load groups that would be useful to control for when comparing post-reform
asset growth. We now turn to a regression approach where we explicitly control for all time invariant
fund characteristic differences across these two groups (using fund fixed effects), as well as time varying
characteristics such as return performance, which should be important in explaining asset growth at the
fund level. These tests allow us to determine whether the negative impact of the entry load ban on
mutual funds might be obfuscated by other important changes occurring across these two groups during
our study period.
Our primary statistical results are produced using the following estimating equation where we sepa-
rately estimate the impact of the entry load ban on high versus low entry load funds:
Yi;t ¼ b0 þ b1 Post Reform Ã High Entry Load Fundt þ b2 Post Reformt
(2)
þb3 High Entry Load Fundi þ bPi;t þ ci þ ei;t

Yi;t is our outcome variable (either net flows or log assets under management) for fund i in month t. The
variable Post Reformt is an indicator for observations in months after the reform was implemented
(August 2009 and afterwards). The variable High Entry Load Fundi is an indicator for those funds that
charged an entry load of 2.25 percent or greater before the policy’s implementation. We are interested in
estimating b1 , which is the difference in our outcome variable across high versus low entry load funds in
the period after the policy change.
Pi;t is a vector of covariates that allow us to control for the effect of prior performance (potentially
convex) on fund growth (Sirri and Tufano 1998). ci is a fund level fixed effect which controls for the
fund type, fund’s asset management company, and any other time invariant fund features. Note that, if
we had a balanced panel, then introducing fund fixed effects would not change our estimated program
effect, as the program effect is only estimated off differences across groups and over time, as opposed to
differences within each fund. Our panel is not balanced, however, because we have funds entering and
exiting the data at different times (limiting our sample to a balanced panel would introduce survivorship
bias), so it is possible that introducing fund fixed effects could change our estimates. Introducing fund
fixed effects could also potentially reduce the standard errors in our estimates, as the fund fixed effects
absorb variation within funds over time. Introducing fund fixed effects does not change our estimated
program coefficient meaningfully, nor does it cause our estimated program impact to be statistically sig-
nificant. Standard errors are heteroskedasticity robust and are clustered at the fund level.
Table 3 presents our estimation results where we use net flows into the fund as a measure of fund growth.
Column (1) includes only the variables necessary to assess the differential impact of the policy change on
the high and low entry load groups. Column (2) introduces fund fixed effects to the specification.34 Column
(3) adds the logarithm of the fund’s assets under management in the prior period as a control variable.
In general, we find that larger funds tended to have lower net flows over the period of our sample.35

34 In a balanced panel, the introduction of fixed effects would not change our main coefficient of interest (b2). Our data,
as mentioned above, is an unbalanced panel in that we include all funds that had at least one observation before the im-
position of the entry load ban. If there is some correlation between the types of funds that exited the data and the tim-
ing of the policy change, then our estimate of b2 can differ when we include fund fixed effects. We find, however, that
the differences in our estimates with and without fund fixed effects to generally be economically insignificant, suggest-
ing that there is not an important correlation between the types of funds that entered and exited and the policy change.
35 Using US data on mutual fund flows, Christoffersen, Evans, and Musto (2013) also finds a negative relationship be-
tween the logarithm of fund size and inflows.
258 Anagol et al.

Table 3. Net Flows to High vs. Low Entry Load Funds

(1) (2) (3) (4) (5) (6)

Post*high entry load fund 0.466 0.653* 0.534 0.499 0.0976 0.658
(0.346) (0.370) (0.391) (0.392) (0.387) (0.409)
Post reform À1.022*** À1.476*** À1.033***
(0.313) (0.339) (0.360)
High entry load fund À0.167
(0.255)
Log(AUM(t-1)) À1.346*** À1.497*** À1.199*** À1.520***
(0.153) (0.169) (0.168) (0.181)
Lag ranked returns low À0.00200 À0.00397
(0.00380) (0.00391)
Lag ranked returns high 0.0462*** 0.0441***
(0.00465) (0.00490)
Observations 46498 46498 46498 46498 41511 41511
Mean net ﬂow 0.0360 0.0360 0.0360 0.0360 À0.0220 À0.0220
St. dev. net ﬂow 7.471 7.471 7.471 7.471 7.160 7.160
Fund FE No Yes Yes Yes Yes Yes
Month*year FE No No No Yes Yes Yes
Month*year*family FE No No No No No Yes

Authors’ analysis based on data described in the text. This table presents regression results on the impact of the entry load ban on net ﬂows into high entry load
funds versus low entry load funds. The dependent variable is the monthly net ﬂow into the fund. The High Entry Load Fund variable takes a value of one for funds
that charged an entry load of 2.25 percent or higher in the pre-reform period. Post*High Entry Load Fund is an interaction of the Post Reform variable and the High
Entry Load Fund variable. The Log(AUM(t-1)) variable is the one-month lagged value of the logarithm of assets under management. The variable Lag Ranked
Returns Low is deﬁned as min (.5, Rank), where Rank is deﬁned as the percentile ranking (0–100) of the fund’s past six month returns within its category. The varia-
ble Lag Ranked Returns High is deﬁned as Rank – Lag Ranked Returns Low. Month*Year*Family FE are a set of ﬁxed effects for each fund family separately for
each month in the data.

Column (4) introduces Month*Year fixed effects (i.e., a fixed effect for January 2007, February 2007, etc.)
to control for aggregate time effects.
Column (5) introduces a measure of fund performance as an explanatory variable defined similarly to
the past performance controls used in Christoffersen, Evans, and Musto (2013). This measure is defined
as follows. For each month t, we calculate the fund’s total return over the period t–7 through t–1 (i.e.,
the total return in the six months prior to the current month). We then define the variable Rank as the
fund’s percentile rank within its fund category for that month. For example, a fund that was in the 10th
percentile in its fund category based on its past six month returns would have a value of 10 for this varia-
ble.36 A fund in the 90th percentile would have a value of 90. To allow for a potentially nonlinear rela-
tionship between past fund performance, as shown in Sirri and Tufano (1998), we include two variables
to measure this relationship as is done in Christoffersen, Evans, and Musto (2013). The variable Lag
Ranked Returns Low is defined as min(.5, Rank). The variable Lag Ranked Returns High is defined as
Rank – Lag Ranked Returns Low (i.e., it takes a value of zero for ranks below .5 and the difference
between the rank and .5 for ranks above .5). The inclusion of both these variables allows us to estimate
a different slope on the performance variable below and above median performance.
In column (5) we control for the past performance of the fund using the Lag Ranked Returns Low
and Lag Ranked Returns High variables. The sample size is lower in this column because we require six
months of lagged return to form the fund ranking variables. Similar to previous studies, we find evidence

36 The fund categories in our data are Index, Tax Savings, General Equity, Large Cap, Sector Fund, Bond Equity Mix,
International, Small and Mid Cap, Balance, and Gold. We use broadly defined categories, similar to the six broad cate-
gories used in Christoffersen, Evans, and Musto (2013).
The World Bank Economic Review 259

for a convex performance-flow relationship. The slope in the low performance range (i.e., the coefficient
on the Lag Ranked Returns Low variable) is estimated to be negative and statistically insignificant. The
coefficient on the Lag Ranked Returns High variable is estimated to be positive and significant. Given
the noisiness observed in our net flow measure, it is re-assuring to know that we observe the standard
convex relationship between past performance and fund flows found in other contexts (Sirri and Tufano
1998. In terms of economic magnitude, a 10 percentage point increase in a fund’s ranking above the
median ranking is associated with a 46 basis point increase in net flows in the current month.
Column (6) includes separate month*year fixed effects for each different asset management company
in the data. These company-by-time fixed effects allow us to control for differential trends across differ-
ent companies that might have been correlated with the timing of the entry load policy, in the way sug-
gested by Gormley and Matsa (2014).37
The difference between the high and low entry load funds after the entry load ban, as measured by
the coefficient on the Post*High Entry Load Fund variable, is estimated as positive across all of these
specifications (although not statistically significant except in column (2)). While the coefficient of .66
percentage points in the full specification (column (6)) is not statistically significant at the 5 percent level,
the lower bound on the 95 percent confidence interval around this estimate (À.19 percentage points)
effectively rules out the possibility of a large negative impact of the reform on net flows. For example,
this lower bound on the confidence interval is small in absolute value terms relative to the standard devi-
ation in monthly net flows of 7.19 percentage points in this sample.38,39,40,41

Index Fund Results
In this section we look at a specific set of funds, index funds, where it is unlikely that there were impor-
tant differences across high and low entry load versions prior to the reform. We compare the asset
growth of index funds that charged high entry loads versus those that charged low entry loads, then see
if high entry load funds have experienced lower asset growth in the period after the reform. We point
out two major differences in our analysis of index funds versus all funds. First, index funds were substan-
tially more likely to charge zero entry loads prior to the reform and less likely to charge an entry load of

37 We have also estimated this equation including fund category by month fixed effects. In this case the results are very
similar to those when we exclude the income type funds (presented later), as including these fixed effects causes us to
estimate the effect mainly within General Equity funds; General Equity funds are the only set of funds where there is
substantial variation within category in pre–entry load ban entry loads (1).
38 Given that one important difference between our high and low entry load groups is the preponderance of income type
funds in the low entry load group, we redo the analysis, dropping income funds from both groups. Appendix table S1.2
presents summary statistics on the high and low entry load funds dropping income funds and appendix table S1.3 pre-
sents the regression analysis. The removal of the income type funds leads to an even larger, positive, and significant es-
timated difference between the high and low entry load funds in the period after the policy.
39 Given the noisiness in the net flow measure, appendix table S1.4 presents the same results as table 3 but with the loga-
rithm of assets under management as the dependent variable (results discussed in appendix section S1). The results are
similar.
40 As an additional robustness check, we also estimated results where we define our low entry load group as funds that
charged zero entry loads prior to the reform. This comparison group of funds experienced no change in the entry loads
they could charge so should not be directly affected by the entry load ban (although spillover effects could naturally af-
fect this group as well). Under this redefined control group, we again find no evidence to suggest that the entry load
ban has caused a decline in flows into mutual funds.
41 Appendix table S1.5 conducts a robustness test where we do the analysis at the fund family level, where there is no en-
try and exit. We define fund families as high entry load if they had greater than the median proportion of their assets in
high entry load funds. We find little evidence that the policy reduced flows to these high entry load fund companies rel-
ative to low entry load companies.
260 Anagol et al.

Table 4. Summary Statistics on Index Funds Sample in the Pre-reform Period

Low entry load fund High entry load fund Difference

Entry load (%) 0.00 1.39 À1.39***
(0.00) (0.63)
Net ﬂow (1% trim) 2.90 2.46 0.43
(12.34) (12.77)
Assets under management (rupees millions) 2765.50 236.53 2528.96***
(10205.82) (811.63)
Assets under management (US dollars millions) 55.31 4.73 50.58***
(204.12) (16.23)
Log(AUM(t)) 5.77 3.84 1.93***
(1.69) (1.94)
Return(t) (%) 1.55 0.71 0.84*
(8.78) (8.25)
Management expense ratio 0.87 1.11 À0.24***
(0.18) (0.13)
Exit load ﬁrst 30 days 0.73 0.11 0.62***
(0.49) (0.26)
Exit load 31 to 180 days 0.10 0.04 0.06**
(0.37) (0.17)
Exit load 181 to 365 days 0.10 0.03 0.07***
(0.37) (0.15)
Minimum investment (rupees) 6645.70 3901.53 2744.18***
(2351.97) (1558.58)
Minimum investment (US dollars) 132.91 78.03 54.88***
(47.04) (31.17)
#Funds 19 26
#Funds*month observations 493 727

Authors’ analysis based on data described in the text. This table presents summary statistics on low entry load index funds and high entry load index funds. The
high entry load group is deﬁned as funds that charged an entry load greater than zero prior to the entry load ban. The Difference column presents the difference in the
mean value of the variable across the low and high entry load groups. Conversions from Indian Rupees to US Dollars are done at ﬁfty rupees per dollar. Standard devi-
ations of the variable are presented in parentheses below the mean. The signiﬁcance stars in the Difference column indicate whether the mean value of the variable is
signiﬁcantly different across the high and low entry load groups. ***, **, * indicate signiﬁcance at 1%, 5%, 10%, respectively.

2.25%. Thus, we define our low entry load group for index funds as the group of funds that charged
zero percent entry loads prior to the reform, and our high entry load group as the set of funds that charge
an entry load greater than zero percent prior to the entry load ban. Second, the sample of index funds in
both the high and low entry load groups is relatively small, and thus the statistical results are less precise
than the full sample.
Table 4 presents summary statistics on the high and low entry load index funds prior to the reform.
High entry load index funds charged 1.39 percent in entry loads on average, while (by definition) the
low entry load index funds charged zero percent. Appendix figure S1.7 shows that the entry load ban led
to a large and discrete drop in the level of entry loads charged by the high entry load index funds. Net
flows into the low and high load groups were 2.9 percent per month and 2.46 percent per month, respec-
tively. Net flows were not statistically different at the 10 percent level. However, the low entry load
funds are significantly larger, with the average assets under management in low entry load funds equal
to 55.3 million dollars versus only 4.73 million dollars in high entry load funds. On average, the high
entry load index funds earned approximately 80 basis points less per month (before fees).42 However,

42 We have checked whether this is due to a small number of funds in either group that earned different returns, but we
found no evidence that this difference is due to outliers.
The World Bank Economic Review 261

this difference does not appear to have changed much over time and the trends in returns in these two
groups of funds are very similar as shown in appendix figure S1.8. High entry load funds also charge
higher management expense ratios, although they charge lower exit loads within the first thirty days
after investment. Low entry load funds have a higher minimum investment level.
Figure 5 presents the average net flows into high and low fee index funds. From this figure, it does
appear that high entry load funds have experienced perhaps slightly lower net flows in the period after
the entry load ban relative to low entry load funds. In particular, net flows into the low entry load funds
were visibly higher than net flows into the high entry load funds in the six months after the reform.
Given the noisiness of these series it is difficult to determine from the figure alone whether these differen-
ces are statistically significant. Table 5 conducts an empirical test to determine whether the difference in
net flows into high versus low entry load funds in the post-reform period is significant. The empirical
results suggest that these two types of funds did not have significantly different net flows after the policy
change, either in a raw comparison (column (1)) or a specification that includes fund fixed effects, lagged
performance, month and year fixed effects, and family fixed effects (column (4)).
We have also tried this specification excluding the six months after the reform to determine whether
the negative point estimates are due mainly to this short period after the policy change. We do find that
the coefficients are less negative (but not positive) in this restricted sample. However, the standard errors
are too large to determine whether the estimates from this restricted sample are different from the full
sample. It is also worth noting that all of the negative estimates we find on the Post*High Entry Load
Fund interaction in table 5 are equal to less than .2 standard deviations in net flows. Thus, if there was a
negative policy impact, it has a very small effect relative to the regular monthly variation in net flows in
our sample period.
The left panel of figure 5 presents the average log assets under management for the high and low entry
load groups. The trends in assets under management across the two groups here are less similar than the
corresponding pictures in the sample of all funds; in particular the low entry load group experienced a
much larger decline in assets under management during 2008 and early 2009. After the policy change,
the high entry load group has experienced a small decline, while the low entry load group has experi-
enced a small increase in assets under management. At the time of the policy change there were eighteen
index funds in each of the high and low entry load index fund groups. This figure alone suggests that
high entry load funds might have fared slightly worse in the period after the entry load ban versus funds
in the low entry load group. However, given the differences in trends prior to the reform, it is perhaps
harder to argue that these groups of index funds would have been similar in the absence of the policy
change.43
Overall, when we focus on index funds, which are funds with perhaps the lowest level of heterogene-
ity prior the reform, we find little evidence to suggest that the entry load ban has caused a reduction in
the growth of Indian mutual funds. One important caveat to these results is that the standard errors on
the regression coefficients are large; we cannot decisively rule out large negative coefficients associated
with the reform. For example, the point estimates on the net flow regression are negative, so it is possible
that with a larger sample size, we would have been able to reject the null hypothesis of no difference
between the high and low entry load groups. Nonetheless, we argue that the combination of the graphi-
cal evidence, which shows little difference in the two types of funds, as well as the generally small size of
the point estimates from the regressions, suggests that the policy did not have an important differential
impact on asset growth in high versus low entry load index funds.

43 We also estimated our main equation using the index funds sample and log assets under management as the dependent
variable; we did not find evidence of a robust drop in high entry load assets under management relative to low entry
load funds.
262 Anagol et al.

Figure 5. Asset Growth in High Vs. Low Fee Index Funds.

This ﬁgure presents the average log assets under management (left panel) and net ﬂows (right panel) in the high entry load index
fund group (solid line) and the low entry load index fund group (dashed line). The high entry load group is deﬁned as any fund that
charged an entry load greater than zero prior to the reform; the low entry load group is the set of funds that charged zero entry
loads prior to the reform. The dashed vertical line indicates the date the policy went into force (August 2009). Authors’ analysis
based on data described in text.

V. Interpretation of Results
Overall, our results suggest that the period after the policy change was a time when previously high entry
load funds experienced similar growth rates as funds that charged low entry loads prior to the ban. If any-
thing, we find that as we tighten the comparison between the high and low entry load groups (i.e., add
more controls in tables 3), the coefficient on the Post*High Entry Load Group variable tends to increase.
The World Bank Economic Review 263

Table 5. Net Flows in High vs. Low Entry Load Index Funds

(1) (2) (3) (4)

Post*high entry load fund À0.805 À0.0492 À0.260 À1.275
(0.921) (1.387) (1.444) (1.319)
Post reform À2.054** À2.527**
(0.811) (1.187)
High entry load fund À0.620
(0.945)
Lag ranked returns low 0.0495***
(0.0179)
Lag ranked returns high À0.00259
(0.0311)
Observations 2240 2240 2240 2016
Mean net ﬂow 1.374 1.374 1.374 1.352
St. dev. net ﬂow 10.46 10.46 10.46 10.27
Fund FE No Yes Yes Yes
Month*year FE No No Yes Yes

Authors’ analysis based on data described in the text. This table presents regression results on the impact of the entry load ban on net ﬂows in high entry load index
funds versus low entry load index funds. See table 3 for further description.

While it is beyond the scope of our paper to definitively determine the causes of this result, we believe
our conceptual framework provides two useful possible nonexclusive explanations for why the entry
load ban may not have led to a relative decline in the high versus low entry load funds. First, it is possible
that the fund companies used other sources of fees besides entry loads to continue to pay substantial
0
commissions (i.e., in terms of equation 1, c À c was small). A second possible explanation is that the
number of less sophisticated investors, whose behavior is influenced by brokers, is not substantially
larger than the number of more sophisticated investors, who might have been attracted to high entry
load funds by the reform.

Evidence on Commissions
Ideally, we would have data on the commission rates offered by mutual fund companies to brokers for
all Indian mutual funds before and after the entry load ban. Unfortunately, there are no comprehensive
sources of data on commissions paid to brokers prior to the ban. Shah et al. (2010) estimate using data
on approximately 70 percent of fund flows in the postban period that the average upfront commissions
paid to brokers equaled 1.78 percent of the initial investment in 2008, 1.39 percent in 2009, and .94 per-
cent in the first quarter of 2010. Interestingly, this data suggests that even in the first quarter of 2010,
after the entry load ban had been passed, funds were still able to pay approximately 1 percent in upfront
commissions to brokers.
The data from Shah et al. (2010) are corroborated by anecdotal evidence. Shah and Kant (2011) note
that according to mutual fund industry executives, commissions have come down from approximately
1.2 to 1.5 percent to approximately .75 percent after the entry load ban. Price Waterhouse Coopers
India (2012) states: “Prior to the no-load regime, the distributor could earn commission between three
to four percent on NFOs [new fund offers] and two to 2.5 percent on existing schemes. Post the restric-
tion on entry loads, this has been reduced to a range of .75 percent to one percent.” This evidence sug-
gests that the regulator’s goal of eliminating commissions paid from asset management companies to
broker was not fully achieved by the entry load ban policy and that perhaps these commissions were
high enough to undo any major effects of the entry load ban on net flows.
We also obtained proprietary data on upfront and trail commissions paid to brokers in the south
Indian states of Tamil Nadu and Pondicherry that sell the funds of the Unit Trust of India (UTI) mutual
264 Anagol et al.

fund. A fund only appears in this data if there was at least one transaction where the commission was
earned or there was a change in the commission rates offered on this fund. Thus, we have a selected sam-
ple, and in particular, it is likely that the commission rates in this sample will be higher, as having a high
commission may make it more likely for the fund to experience a transaction. The UTI data covers on
average 6.6 funds per month (depending on the month) and includes the minimum and maximum
upfront, first year trail, and second year trail commissions.44 The data covers the months August 2008
through January 2009 and then March 2010 through March 2013; unfortunately there is a gap in our
data series between February 2009 and February 2010.
Appendix figure S1.9 plots the average maximum upfront commission offered on this particular set
of UTI funds; average upfront commissions were 2.06 percent prior to the ban. Interestingly, average
upfront commissions on this set of funds, even after the ban, have been equal to 1.82 percent. Appendix
figure S1.10 plots the average first and second year trail commissions offered for this set of funds. The
average first year trail commission for this set of funds has increased slightly from .75 percent per year
to .90 percent per year, and the average second year trail commissions has increased from .39 percent
per year to .55 percent per year. These results suggest, that, if anything, trail commissions have been
increased in the period after the entry load ban. A broker who sold the average fund in this sample of
UTI funds prior to the reform and expected the investor to stay invested for two years could expect
approximately 3.2 percent in commissions earnings, whereas after the reform this broker would expect
3.22 percent in commissions earnings. The results from this small sample of funds suggest that the entry
load ban might not have had as strong an effect on broker’s incentives as the regulator’s intended impact
of eliminating all commissions paid from mutual fund companies to brokers.45
One possibility is that the maintenance of relatively high commissions immediately after the entry
load ban was primarily to maintain short-term market share but that, in the longer run, fund companies
were forced to reduce commissions as revenues from entry loads were no longer available to pay them.
Appendix figure S1.1 showed estimated commissions from twenty-five UTI funds for investors with
holding durations of six, twelve, thirty-six, and seventy-two months. Total commissions in June 2014
are lower than those in June 2010 at the six month duration, which is consistent with the idea that over
time the entry load ban may have reduced commissions more substantially. However, at the longer dura-
tions, total commissions have not changed much between June 2010 and June 2014, suggesting that the
maintenance of relatively high commissions has persisted past just the period immediately after the
reform.

Adjustment of Fees After the Entry Load Ban to Maintain Commissions
Given the evidence that fund companies appear to have maintained substantial commissions despite the
entry load ban, it is interesting to test to what extent they raised other fees to offset the policy-induced
decline in entry loads. Indian mutual funds had three methods of raising money to pay commissions after
the entry load ban. First, they could charge up to 1 percent in exit loads and use those revenues to pay
commissions. Second, they could use any loads saved up prior to the entry load ban to pay commissions.
And third, they could use a portion of the management expense ratio charged to the fund as a whole to

44 Minimum and maximum commissions are included because different individual broker’s within this broker’s network
receive different commissions; the minimum level is almost always zero, so all the statistics we report refer to the maxi-
mum commission.
45 In appendix section S1 we discuss the related fact that Systematic Investment Plans (SIPs) became substantially more
popular after entry load ban (see appendix figure S1.11); this is consistent with the idea that brokers were now more
focused on trail commissions, as SIP investments are typically of longer duration.
The World Bank Economic Review 265

Table 6. Changes in Fees

Dependant variable: Entry load MER Exit load Total fees on 100 rs. investment

Holding period: <1 Month 1–6 Months > 6 Months 6 Months 1 Year 6 Years
(1) (2) (3) (4) (5) (6) (7) (8)

Post*high entry load fund À1.71*** À.00184 .256*** .282*** .303*** À1.432*** À1.41*** À1.421***
(.109) (.0269) (.0468) (.0571) (.0697) (.115) (.117) (.170)
Observations 5329 5329 5329 5329 5329 5329 5329 5329
Mean dep. var .755 2.064 .802 .751 .731 2.538 3.550 13.87
St. dev. dep. var.) 1.063 .555 .451 .478 .499 1.225 1.378 3.784
Fund FE Yes Yes Yes Yes Yes Yes Yes Yes
Month*year FE Yes Yes Yes Yes Yes Yes Yes Yes

Authors’ analysis based on data described in the text. This table presents regression results on the impact of the entry load ban on the level of entry loads, manage-
ment expense ratios (MER), and exit loads charged by funds, as well as the total costs for a six-month, one-year, and six-year horizon investor. Post*High Entry Load
Fund is an interaction of the Post Reform variable and the High Entry Load Fund variable.

pay for commissions. In this section we discuss to what extent funds adjusted along these various mar-
gins to maintain commissions after the entry load ban.
Table 6 presents an analysis of how funds changed their fees after the entry load ban and also how
these fee changes translated into different costs for investors of different holding horizons.46 One impor-
tant feature of this table is that the sample is limited to fund*month observations in the time period
August 2008 through August 2010, where we have data on entry loads, exit loads, and management
expense ratios.47 While funds almost always report their entry and exit loads in monthly fact sheets, it is
much less common for funds to report their management expense ratios. Of the 17,057 fund*month
observations in our main sample in this time frame, we were only able to locate the management expense
ratio charged in that month for 5,329 observations. Therefore, an important caveat to this analysis is
that it pertains only to a relatively small set of funds where all fee data was available.48
The main independent variable of interest is the Post*High Entry Load Group interaction between a
dummy for the post-reform period and whether the fund had a high entry load prior to the reform. All
regressions include fund fixed effects and month*year fixed effects. In column (1) the dependent variable
is the entry load charged by the fund in a given month. As expected, the entry load ban is associated
with a large and significant drop (1.7 percentage points) in the level of entry loads in this sample. In col-
umn (2) the fund’s management expense ratio is the dependent variable. We find little difference in the
level of management expense ratios charged by high entry load funds after the ban. In columns (3)–(5),
the dependent variable is the exit load (measured in percentage points). In this sample, we find that high
entry load funds charged approximately .26, .28, or .30 percentage points more for exiting the fund in
between 0 and 30 days, 30 and 180 days, or beyond 180 days, respectively. Interestingly, these results
suggest that high entry load funds did somewhat offset the mandatory reduction in their entry loads by
charging higher exit loads.
In columns (6)–(8), the dependent variables are the total rupees paid in fees on a one hundred rupee
investment that is sold after six months, one year, and six years, respectively. Overall, the entry load ban
causes an approximate 1.4 percentage point reduction in fees across these horizons; note that this effect

46 For a general discussion of the economic rationale for different types of mutual fund fees, see Chordia (1996).
47 We focused on this period to reduce the cost of data collection as management expense ratios were manually collected.
48 We found similar results to the main sample when we tested whether the entry load ban caused a reduction in fund
growth in this more limited sample.
266 Anagol et al.

does not vary over time because changes in entry and exit loads charged do not depend on the investor’s
holding period (and we estimate no change in management expense ratios).49 Overall, the data suggests
there may have been some adjustment in exit loads but that these higher exit loads were most likely
swamped by the major decrease in entry loads caused by the policy change.

Relative Importance of Different Types of Investors
Our conceptual framework also highlights the possibility that for some high-sophistication investors, the
entry load ban might increase the flows to high entry load funds relative to low entry load funds. If these
types of investors are quantitatively important, then it is possible that the positive effects on flows
induced in this group offset the negative flow effects in the less sophisticated groups, yielding our find-
ings of little overall difference across the two groups. Assessing the importance of this explanation would
require proxy data on investor sophistication as well as investor-level investment information, which is
currently not available. However, we believe our results are very useful in ruling out the idea that, at
least in this particular context, the mutual funds market was made up primarily of low sophistication
investors whose behavior is strongly determined by commissions-motivated brokers.50

Did the Entry Load Ban Cause an Aggregate Decline in Fund Flows?
The results so far are inconsistent with the entry load ban being the main reason for the massive outflows
out of mutual funds after the entry load ban. One possible impact of the ban that our conceptual frame-
work and empirical strategy cannot address, however, is that both high and low entry load funds experi-
enced lower net flows after the reform. Our empirical strategy exploits differences in flows to high and
low entry load groups, with the assumption that high entry load funds should have been more affected
by the entry load ban than low entry load funds. It is possible that an additional effect of the ban is that
both high and low entry load groups experienced lower flows due to the ban (perhaps because some
brokers chose to exit and stop selling all funds). Our strategy only picks up impacts of the ban that cause
differential flow changes across high and low funds.
While it is possible that the entry load ban caused reduced flows into both groups (i.e., the industry as
a whole) to fall after the reform, we believe it is unlikely that the entry load ban was the main cause of
the fall in net flows into mutual funds overall. The main reason is that low rates of investment during
this period were observed in other financial assets as well, including bank accounts, equities, and bonds.
We would expect that if the entry load ban, which only applied to mutual funds, was an important cause
of the generally low flows into mutual funds, that only mutual funds would show low flows in the post
entry load ban period. What we observe, however, is that essentially all financial assets, including equi-
ties, bank deposits, and bonds, had low investments in the post entry load ban period.
The period between 2009 and 2012 followed the 2008 financial crisis. Anecdotal evidence suggests
this made households wary of investing in financial assets, and led them towards real assets such as real
estate and gold. Aggregate data from the Reserve Bank of India suggests that the period after the entry
load ban was in general a time when households were moving out of financial assets and into real assets
such as real estate and gold. Net financial saving (i.e., savings in financial instruments such as stocks,
mutual funds, bonds, and bank accounts) fell from 12.2 percent of GDP in financial year 2009–2010 to
9.3 percent in 2010–2011, and to 7.8 percent in 2011–2012 (Reserve Bank of India 2012). The 2011–
2012 Annual Report of the Reserve Bank of India highlights this issue specifically:

49 We have conducted the analysis on the much larger set of funds that have entry and exit load data but are missing ex-
pense ratio data. In that set of funds we find no evidence that funds raised exit loads after the entry load ban (see ap-
pendix section S1 and appendix figures S1.12, S1.13, and S1.14.
50 We note that this may be a very context-specific result; in times when markets are rising, the number of unsophisticated
investors could increase substantially and brokers may be more influential.
The World Bank Economic Review 267

With real interest rates on bank deposits and instruments such as small savings remaining relatively low on
account of the persistent high inflation, and the stock market adversely impacted by global developments, house-
holds seemed to have favored investment in valuables, such as gold. In the post-global crisis period, valuables
have increased from 1.3 percent of GDP at current market prices in 2008–09 to 2.8 per cent in 2011–12; the share
of valuables in investment (gross capital formation) has also increased from 3.7 per cent to 7.9 per cent, over this
period. The apparent proclivity of households towards investment in valuables such as gold could have also
impacted the pace of their investment in physical assets such as housing in 2011–12.

To confirm that the above aggregate patterns are also likely to be true for household investments into
mutual funds, we use data from Consumer Pyramids, a newly available representative household survey
of all households in India conducted by the Centre for Monitoring Indian Economy (CMIE) on a quar-
terly basis. This is a panel data set where each survey household is contacted quarterly to answer ques-
tions on income, consumption, saving, and borrowing. This survey began in March 2009 with 120,000
households. In March 2011, new households were added to the panel, and the sample size was increased
to 150,000 households. In appendix figure S1.15, we present the proportion of households with out-
standing investments in various asset classes. This includes financial instruments, such as mutual funds,
life insurance, and fixed deposits, as well as real assets, such as gold and real estate. The figure shows a
substantial increase in the proportion of total households with outstanding investments in gold and real
estate.

VI. Conclusion
Expanding formal financial markets is a key policy goal in many developing countries. Policymakers
face an important trade-off in regulating the role that brokers play in the expansion of these markets.
On the one hand, brokers have historically played a very important role in the development of formal
financial markets throughout the world, and so regulating firms’ ability to pay commissions to brokers
may slow the growth of formal financial markets. On the other hand, recent anecdotal and empirical evi-
dence suggests that commissions-motivated brokers may provide unsuitable advice.
Our paper provides some first evidence on the impact of a regulation that attempted to reduce the
importance of brokers in the intermediation process. We study the impact of a ban on entry loads whose
purpose was to reduce the amount of commissions that fund companies would pay to brokers. A first
order concern with this type of policy is the possibility that reducing commissions would slow the
growth of the mutual funds market. Consistent with this, aggregate data shows that net flows into
mutual funds declined dramatically in the three years after the entry load ban was introduced. However,
our comparison of funds that were heavily affected by the policy (high entry load funds) versus funds
that were less affected (low and zero entry load funds) reveals that previously high entry load funds have
not attracted lower flows after the entry load ban. We conclude from this analysis that this particular
commissions reform did not have a major impact on the relative attractiveness of funds and that the
entry load ban is unlikely to have played an important causal role in the aggregate decline in flows into
mutual funds.
The entry load ban’s stated goal was to discourage mutual fund companies from paying commissions
directly to brokers; instead, the regulator envisioned a market where investors paid brokers directly for
advice. The available, albeit limited, data on commission levels paid to brokers in the period after the
reform suggest that the entry load ban was not successful in eliminating commissions paid from fund
companies to brokers. We suspect that the entry load ban ultimately did not have an important impact
on fund flows because fund companies were able to maintain substantial commissions even after the
ban. Although upfront commissions were approximately 1 percent lower in the period after the entry
load ban, our results suggest that the marginal flows lost from this decline in upfront commissions was
small. Another complementary explanation for our results is that the number of investors who are
268 Anagol et al.

strongly influenced by commissions is actually relatively small (at least during this time period); under-
standing what types of market conditions are associated with commissions influencing behavior most is
an interesting area for future research.
Our paper, while not a complete welfare evaluation of the Indian entry load ban, does suggest that
the entry load ban did not have major negative consequences for the growth of the Indian mutual fund
industry. Fees and commissions paid by investors appear to have declined modestly while our estimates
suggest that flows were not substantially affected. Given that higher fees typically did not earn investors
higher returns prior to the reform, investors are likely to experience higher net of fee returns in the future
due to the reform. The reform may also have had other benefits that are outside the scope of our analy-
sis. It is possible that brokers are less likely to encourage investors to quickly buy and sell investments to
maximize commissions as the reform reduced upfront commissions. The reform also appears to have
caused the industry to innovate towards using trail commissions (i.e., commissions paid based on how
long an investor holds a mutual fund, as opposed to at the time of entry), which at least in theory should
better align the incentives of brokers and investors (Barbora 2015). We leave the evaluation of these ben-
efits to future work, perhaps in other contexts, where better data and additional reforms allow for a
more holistic evaluation.

References
Agarwalla, S. K., J. Jacob, and J. R. Varma. 2013. “Four Factor Model in Indian Equities Market.” (2013–09–05).
Indian Institute of Management Working Paper, Ahmedabad, Gujarat.
Anagol, S., S. Cole, and S. Sarkar. 2012. “Understanding the Incentives of Commissions Motivated Agents: Theory
and Evidence from the Indian Life Insurance Market.” Harvard Business School Working Paper 12–055,
Cambridge, MA.
Anagol, S., and H. Kim. 2012. “The Impact of Shrouded Fees: Evidence from a Natural Experiment in the Indian
Mutual Funds Market.” American Economic Review 102 (1): 576–93.
Association of Mutual Funds in India. 2015a. “Individual Investors July 2015.” https://www.amﬁindia.com/Themes/
Theme1/downloads/home/individual-investors.pdf.
———. 2015b. “Industry Trends July 2015.” Tech. rep., Association of Mutual Funds in India. https://www.
amﬁindia.com/Themes/Theme1/downloads/home/industry-trends.pdf.
Barbora, L. 2015. “Does it Work to Cap Upfront Commissions?” http://www.livemint.com/Money/
xhNov6D7T32dradk2rt7JJ/Does-it-work-to-cap-upfront-commissions.html.
Bekaert, G., C. R. Harvey, and C. Lundblad. 2001. “Emerging Equity Markets and Economic Development.” Journal
of Development Economics 66 (2): 465–504.
Bergstresser, D., J. M. R. Chalmers, and P. Tufano. 2009. “Assessing the Costs and Beneﬁts of Brokers in the Mutual
Fund Industry.” Review of Financial Studies 22 (10): 4129–56.
Bowen, C. 2011. “Future of Financial Advice: Information Pack.” Tech. rep., The Treasury of Commonwealth of
Australia. Canberra, Australian Capital Territory.
Cai, J., and C. Song. 2012. “Insurance Take-up in Rural China: Learning from Hypothetical Experience.” Available
at SSRN 2161649.
Carlin, B., and D. Robinson. 2012. “What Does Financial Literacy Training Teach Us?” Journal of Economic
Education 43 (3): 235–47.
Chordia, T. 1996. “The Structure of Mutual Fund Charges.” Journal of Financial Economics 41 (1): 3–39.
Christoffersen, S. E. K., R. Evans, and D. K. Musto. 2013. “What Do Consumers’ Fund Flows Maximize? Evidence
from Their Brokers’ Incentives.” Journal of Finance 68 (1): 201–35.
Christoffersen, S. K., R. B. Evans, and D. K. Musto. 2005. “The Economics of Mutual-Fund Brokerage: Evidence
from the Cross Section of Investment Channels.” AFA 2006 Boston Meetings Paper, Philadelphia, PA.
Cole, S., T. Sampson, and B. Zia. 2011. “Prices or Knowledge? What Drives Demand for Financial Services in
Emerging Markets?” Journal of Finance 66 (6): 1933–67.
The World Bank Economic Review 269

Cole, S., and G. K. Shastry. 2010. “Smart Money: The Effect of Education, Cognitive Ability, and Financial Literacy
on Financial Market Participation.” Harvard Business School Working Paper 09–071, Cambridge, MA.
Collinson, P. 2012. “FSA Ban on Commission-Based Selling Sparks “Death of Salesman” Fears.” http://www.
theguardian.com/business/2012/dec/30/fsa-ban-commission-selling-death.
Federal Reserve Board. 2015. “Financial Accounts of the United States.” Tech. rep., Federal Reserve Board,
Washington, DC.
Ferris, S. P., and D. M. Chance. 1987. “The Effect of 12b-1 Plans on Mutual Fund Expense Ratios: A Note.” Journal
of Finance 42 (4): 1077–90.
Gabaix, X., and D. Laibson. 2006. “Shrouded Attributes, Consumer Myopia, and Information Suppression in
Competitive Markets.” Quarterly Journal of Economics 121 (2): 505–40.
Gaurav, S., S. Cole, and J. Tobacman. 2011. “Marketing Complex Financial Products in Emerging Markets: Evidence
from Rainfall Insurance in India.” Journal of Marketing Research 48 (SPL): S150–S162.
Gine´ , X., C. M. Cuellar, and R. K. Mazer. 2014. “Financial (Dis-) Information: Evidence from an Audit Study in
Mexico.” World Bank Policy Research Working Paper 6902, Washington, DC.
e, X., R. Townsend, and J. Vickery. 2008. “Patterns of Rainfall Insurance Participation in Rural India.” World
Gin
Bank Economic Review 22 (3): 539–66.
Gormley, T. A., and D. A. Matsa. 2014. “Common Errors: How to (and Not to) Control for Unobserved
Heterogeneity.” Review of Financial Studies 27 (2): 617–61.
Halan, M. 2010. “Mint 50: Mutual Funds to Invest In.” http://www.livemint.com/Home-Page/
41MXNN8Â5DHA6nmqCIfP3O/Mint-50-mutual-funds-to-invest-in.html.
Hastings, J., and O. Mitchell. 2011. “How Financial Literacy and Impatience Shape Retirement Wealth and
Investment Behaviors.” NBER Working Paper 16740, Cambridge, MA.
Hastings, J. S., B. C. Madrian, and W. L. Skimmyhorn. 2013. “Financial Literacy, Financial Education, and
Economic Outcomes.” Annual Review of Economics 5 (1): 347–73.
Heidhues, P., B. Ko }szegi, and T. Murooka. 2012. “The Market for Deceptive Products.” University of California
Working Paper, Berkeley, CA.
Inderst, R., and M. Ottaviani. 2009. “Misselling through Agents.” American Economic Review 99 (3):883–08.
Investment Company Institute. 2015. “2015 Investment Company Factbook.” Tech. rep., Investment Company
Institute. https://www.ici.org/pdf/2015_factbook.pdf.
Kamiyama, T. 2007. “India’s Mutual Fund Industry.” Nomura Capital Markets Review 10 (4).
Levine, R., and S. Zervos. 1998. “Stock Markets, Banks, and Economic Growth.” American Economic Review 88
(3): 537–58.
Ministry of Labour and Employment. 2015. “2015 Indian Labour Journal.” Tech. rep., Ministry of Labour and
Employment Bureau, Shimla, Shimla. http://labourbureau.nic.in/ILJ_Jan_2015.pdf.
Mobarak, A. M., and M. R. Rosenzweig. 2012. “Selling formal insurance to the informally insured.” Yale Economic
Growth Center Discussion Paper 1007, New Haven, CT.
Mullainathan, S., M. Noeth, and A. Schoar. 2012. “The Market for Financial Advice: An Audit Study.” NBER
Working Paper 17929, Cambridge, MA.
Pension National, and System Trust. 2015. “Assets Under Management and Number of Subscribers—NPS.” Tech.
rep., National Pension System Trust. New Delhi, New Delhi.
Waterhouse Price, and Coopers India. 2012. “Distribution Spectrum and the Changing Business Environment: Indian
Mutual Fund Industry.” Tech. rep. https://www.pwc.in/assets/pdfs/ﬁnancial-service/cii_mf_summit_fv.pdf.
Mumbai, Maharashtra.
Reserve Bank of India. 2012. “Reserve Bank of India Annual Report.” Tech. rep. Mumbai, Maharashtra.
Securities and Exchange Board of India. 2012a. “The Gazette of India: Extraordinary Part - III – Section 4.” Tech.
rep., Securities and Exchange Board of India, Mumbai, Maharashtra.
———. 2012b. “Master Circular for Mutual Funds.” Circular, Securities and Exchange Board of India, Mumbai,
Maharashtra.
Shah, A., A. Garg, K. Radhakrishnan, and N. K. Prasad. 2010. “Equity Mutual Funds: Charting Your Course with a
Compass.” Boston Consulting Group Mumbai, Maharashtra.
Shah, M., and C. Kant. 2011. “Top Distributors Make Millions as Fund Houses Bleed.” http://business-standard.
com/india/news/top-distributors-make-millions-as-fund-houses-bleed/470840/.
270 Anagol et al.

Sirri, E. R., and P. Tufano. 1998. “Costly Search and Mutual Fund Flows.” Journal of Finance 53 (5): 1589–622.
Song, C. 2015. “Financial Illiteracy and Pension Contributions: A Field Experiment on Compound Interest in China.”
Available at SSRN 2580856.
Walsh, L. 2004. “The Costs and Beneﬁts to Fund Shareholders of 12b-1 Plans: An Examination of Fund Flows,
Expenses and Returns.” Securities and Exchange Commission Working Paper, Washington, DC.
Zelizer, V. 1983. Morals and Markets. Transaction Publishers, Piscataway, NJ.