智能投顾与财富管理可及性（英）-22页

免费

3.0 babyfacer 2024-05-15 145 22 1.11MB 21 页免费

侵权投诉

Journal of Financial Economics 155 (2024) 103829

Available online 22 March 2024

Contents lists available at ScienceDirect

Journal of Financial Economics

journal homepage: www.elsevier.com/locate/jfec

Robo advisors and access to wealth management ✩

Michael Reher a,∗, Stanislav Sokolinski b

aUniversity of California San Diego, Rady School of Management, United States of America

bBroad College of Business, Michigan State University, United States of America

A R T I C L E I N F O A B S T R A C T

Dataset link: https://

data .mendeley .com /datasets /3f88g2cb7x /2

JEL classiﬁcation:

G11

G24

Keywords:

FinTech

Financial advice

Portfolio delegation

Inequality

We investigate how access to robo-advisors impacts the ﬁnancial investment and welfare of less-wealthy

investors. We leverage a quasi-experiment where a major U.S. robo-advisor signiﬁcantly expands access by

reducing its account minimum, increasing participation by middle-class investors but not the poor. A benchmark

model calibrated to portfolio-level data rationalizes this increase: middle-class investors want sophisticated

investing but cannot achieve it themselves. Their welfare rises moderately, driven by advanced features like

multi-dimensional glide-paths and additional priced risk factors. Middle-age investors gain three times more

than millennials. Our results reveal novel margins of demand for robo-advisors, helping explain their sustained

growth.

“The wealth-management industry stratiﬁes customers in a manner

rather similar to airlines. ‘High-net-worth’ clients ﬂy business class, pick-

ing stocks and chatting in person with named advisors. Cattle class gets no

service at all. Technology is conspiring to change that.”

The Economist Magazine

1. Introduction

Recent technological innovation has enabled ﬁnancial intermedi-

aries to scale the provision of many traditional services. In the wealth

✩Nikolai Roussanov was the editor for this article. We thank Christoph Aymanns, John Campbell, Michael Boutros, Francesco D’Acunto, Ben Friedman, Pedro Gete,

Steﬀen Hitzemann, Jakub Jurek, Iva Kalcheva, Yaron Levi, Jun Liu, Matteo Maggiori, Apurva Mehta, Will Mullins, Alessandro Previtero, Franklin Qian, Enrichetta

Ravina, Jonathan Reuter, Alberto Rossi, Nikolai Roussanov, Andrei Shleifer, Celine Sun, Allan Timmermann, Ross Valkanov, Boris Vallee, Ken Wilbur, anonymous

referees, and seminar participants at UC Irvine Finance Conference, Michigan State University, University of Houston, Indiana University, University of Texas at

Dallas, CEAR-RSI Household Finance Workshop, Northeastern University Finance Conference, University of Mannheim, USC, University of Wisconsin, Jackson Hole

Winter Finance Conference, FINRA NORC Diversity Conference, FMA Annual Conference, Paris FinTech Conference, Toronto FinTech Conference, the New York

Fed’s FinTech Conference, Georgetown’s Finance and Policy Center, UC San Diego, Rutgers Business School, California Corporate Finance Conference, CAFR FinTech

Workshop, and Harvard’s ﬁnance PhD lunch for comments. Ruikai Ji provided excellent research assistance. The views expressed in this paper are those of the

authors and do not necessarily reﬂect the position of Wealthfront Inc.

* Corresponding author.

E-mail addresses: mreher@ucsd.edu (M. Reher),

sokolins@msu.edu (S. Sokolinski).

1The top ﬁve robo advisors managed $283 billion in 2020 versus $30.4 billion in 2015 (Appendix Table A1).

management market, this scaling has taken the form of “robo advi-

sors”: intermediaries that use automation to provide services similar

to a traditional wealth manager (D’Acunto and Rossi 2020). Robo ad-

visors have grown rapidly in both popularity and size over the past

decade. For example, the largest ﬁve robo advisors have grown roughly

tenfold, and all of the Big-4 U.S. banks now oﬀer a robo advising

service.1

Robo advisors claim to have two key advantages over their tradi-

tional counterparts. First, by adhering to a transparent investing algo-

rithm, robo advisors can beneﬁt their clients by circumventing inef-

ﬁciencies documented among ﬁnancial advisors, many of which stem

https://doi.org/10.1016/j.jﬁneco.2024.103829

Received 12 July 2023; Received in revised form 1 March 2024; Accepted 9 March 2024

Journal of Financial Economics 155 (2024) 103829

M. Reher and S. Sokolinski

from portfolio-by-portfolio discretion.2Second, because automation

lowers per-portfolio management costs, robo advisors can extend the

beneﬁts of professional wealth management to less-wealthy investors,

who do not have enough assets to invest with traditional managers. Our

paper examines the latter claim, which is popular and intuitive but has

not been studied. We use a quasi-experiment and a quantitative model

to trace out the eﬀect of access to robo advisors down to the welfare of

new, less-wealthy robo investors.

In the ﬁrst part of the paper, we study how access to robo advisors

aﬀects participation by the less-wealthy. Not only is this exercise impor-

tant in its own right, but the results will also help discipline the model

that we develop in the second part of the paper. However, identifying

the eﬀect of interest is challenging for two reasons: lack of standardized

data on robo participants; and a need for a setting in which robo advi-

sors suddenly become more accessible to the less-wealthy. We overcome

these challenges by obtaining a novel dataset directly from a major U.S.

robo advisor and by studying a quasi-experiment. In our setting, de-

scribed in Sections 2and 3, the same robo advisor suddenly reduces its

account minimum from $5,000 to $500. This shock represents one of

the ﬁrst examples in which sophisticated wealth management becomes

available to a wide range of less-wealthy investors.

In Section 4, we ﬁnd that the reduction democratizes the market for

robo advisors by bringing in new, middle-class investors. The wealth

distribution of robo participants shifts sharply leftward after the re-

duction, while showing no pre-trend in the prior months. The share of

participants from the second and third U.S. wealth quintiles (the “mid-

dle class”) increases by 107% (16 pps). This increase reﬂects a sharp

break from trend in the number of middle-class participants, whereas

there is no such change in the number of participants from the upper

two quintiles (the “upper class”). However, the democratization is in-

complete, in that we ﬁnd no change in participation among the bottom

quintile (the “lower class”), as conjectured by Philippon (2019).

According to our theory, a relaxation of minimum-account con-

straints drives this shift in the robo wealth distribution. We sharpen this

interpretation through a diﬀerence-in-diﬀerence (DiD) analysis. The

middle class represents the treatment group, since it experiences a relax-

ation of minimum-account constraints due to the reduction. The upper

class is the control group. As a benchmark, we ﬁnd that middle-class

investors are 14 pps more likely to participate with the robo advisor af-

ter the reduction, relative to the upper class. We obtain similar results

from dynamic DiD designs at various time frequencies, and we perform

an event study that strongly supports the assumption of parallel trends.

Section 5further tests that the empirical eﬀect works by relaxing

minimum-account constraints. We ﬁrst show that the majority of new,

middle-class robo participants bunched their investment at the previ-

ous minimum of $5,000 prior to the reduction. Such bunching is a

hallmark of binding constraints. After the reduction, the bunching im-

mediately disappears, and most new middle-class participants make a

previously infeasible investment of under $5,000. Additionally, we ﬁnd

no evidence that the results are driven by other channels such as: het-

erogeneous response to targeted or non-targeted advertising; gambling

motives; business stealing from competitors; heterogeneous trends by

demographics or risk attitude; or measurement error in self-reported

wealth.

In the second part of the paper, we develop and calibrate a quan-

titative model of asset allocation with endogenous portfolio delegation

and an account minimum. This exercise addresses two major questions

that cannot be answered by the reduced-form DiD analysis from the

2A partial list of ineﬃciencies includes impulsive recommendations (e.g., Lin-

nainmaa et al. 2021), pandering to client biases (e.g., Mullainathan et al. 2012),

overly generic recommendations (e.g., Foerster et al. 2017), and recommenda-

tions of inferior products due to commissions (e.g., Egan 2019; Chalmers and

Reuter 2020). D’Acunto and Rossi (2020)discuss how robo advisors can correct

some of these ineﬃciencies.

ﬁrst part. First, we ask whether diﬀerences in asset allocation suﬃce to

explain the increase in middle-class participation with the advisor, or

whether we must appeal to non-ﬁnancial channels, like peace-of-mind

from portfolio delegation (e.g., Gennaioli et al. 2015). Second, we as-

sess the welfare gain from access to robo advisors, using the standard

lifetime consumption metric (e.g., Gomes 2020). This approach lets us

examine the potential beneﬁts of robo advisors more comprehensively

than, say, using realized returns, as it accounts for the investor’s hori-

zon and human capital.3We are particularly interested in how these

potential beneﬁts may vary across investors and which features of robo

portfolios drive the results.

Our comparison group consists of retail investors who would other-

wise manage their risky assets on their own without access to a robo

advisor (“self-managed portfolios”). Section 6uses portfolio-level mi-

crodata to document three advantages that robo portfolios have over

self-managed ones. First, robo portfolios contain greater exposure to

priced risk factors, such as bond and value premia. Consequently, they

provide a 2 pps higher expected return. Second, they are much better

diversiﬁed. Third, robo portfolios exhibit personalization by both age

and wealth, which we call a “double glide path”: exposure to stocks

falls as an investor grows older, holding wealth ﬁxed; but it rises as an

investor becomes wealthier, holding age ﬁxed. So, robo portfolios are

more personalized than target date funds (TDF), which oﬀer a single

glide path by age.

In Section 7, we embed these portfolio characteristics in the model

and reproduce the empirical results from the ﬁrst part of the paper.

The model closely matches the eﬀects of the reduction in minimum

on the wealth distribution of robo participants. Thus, simply account-

ing for how robo advisors invest diﬀerently than retail investors on

their own explains the reduced-form evidence quite well, without a

clear need for non-ﬁnancial channels. The economic mechanism that

generates these results within the model is precisely a relaxation of

minimum-account constraints, with the following key trade-oﬀ. Absent

the minimum, all investors would like to invest with the robo advi-

sor because of the aforementioned advantages. However, the account

minimum requires a larger-than-optimal risky share and savings rate

for less-wealthy investors. As a result, these investors prefer to manage

their own portfolios, even though they cannot do so as well as the robo

advisor. Reducing the minimum relaxes this constraint, prompting in-

vestors with low-to-intermediate wealth (middle class) to invest with

the advisor. However, lower-class investors still ﬁnd the required risky

investment too large, and so they choose not to participate.

We next examine the magnitude, drivers, and distribution of wel-

fare gains from access to the robo advisor. The reduction raises welfare

of new robo participants by 0.8%, in terms of lifetime consumption.

This increase is meaningful yet plausible.4Interestingly, we ﬁnd sub-

stantial heterogeneity by age: the reduction raises welfare three times

as much for investors over age 55 (1.7%) than for those under age 35

(0.6%). This diﬀerence arises from diﬀerences in lifetime human capi-

tal, mainly because younger investors have longer working lives. As a

result, many of them would have accumulated enough earnings to over-

come the previous minimum anyway, whereas older investors would

not have become robo participants without the reduction.

3Recent and contemporaneous studies have found that robo portfolios have

characteristics that would improve welfare in a static model of asset allocation,

such as better diversiﬁcation (e.g., D’Acunto et al. 2019, Rossi and Utkus 2021b,

Loos et al. 2020). Our approach builds on these insights by both quantifying the

gains from these static features in terms of lifetime utility and by highlighting

dynamic beneﬁts (e.g., age glide paths).

4Measuring welfare gains as a percent of lifetime consumption has a long

tradition in the literature dating back to at least Lucas (1987). A gain of greater

than 0.5% is typically considered consequential. For reference, a gain of 0.8%

lies well-within the range of gains from correcting various investment mistakes

in a workhorse model developed by Cocco et al. (2005).

Journal of Financial Economics 155 (2024) 103829

M. Reher and S. Sokolinski

Since our comparison group consists of self-managed portfolios, the

welfare gains for new robo participants must fundamentally reﬂect the

diﬀerent characteristics of robo portfolios relative to self-managed ones.

We evaluate which characteristics add the most value by calculating the

welfare gain for counterfactual robo portfolios after removing each of

their advantages, one at a time. Without improved diversiﬁcation, the

welfare gain is 70% lower. Similarly, removing exposure to bond risk

factors reduces the welfare gain by 70%. By contrast, diﬀerences in ex-

posure to the overall stock market have almost no impact on welfare.

This ﬁnding reﬂects how investors in our data attain such exposure

on their own. Interestingly, we ﬁnd that personalization by wealth im-

proves welfare as much as does personalization by age, highlighting the

importance of a “double glide path”.

Taken together, our results present a novel perspective on the im-

pact of robo advisors. First, robo advisors do not solely impact aﬄu-

ent investors, who already have access to professional managers: robo

advisors also expand access to the less-wealthy by relaxing minimum-

account constraints. Removing constraints raises participation because

less-wealthy investors have strong demand for sophisticated wealth

management. This result challenges the canonical model of a self-

suﬃcient investor. Our results also challenge models in which non-

ﬁnancial beneﬁts drive the demand for delegated management, since

we can explain this demand as a rational response to improved asset al-

location. In that vein, our evidence matches survey results from Rossi

and Utkus (2021a), who ﬁnd that investors seek robo advisors more

to improve their investment ability and performance than to achieve

peace of mind. More broadly, we exemplify how sophisticated wealth

management is not necessarily a luxury good.

The second contribution of this paper is to rigorously document the

properties of a fully-automated asset allocation rule and to assess which

of its features investors value most. We show that robo advisors do not

simply substitute for a low-cost equity index fund or a TDF. Instead,

much of their value comes from providing sophisticated features like

a double glide path and exposure to multiple risk factors. Less-wealthy

investors appreciate these features because they struggle to enact sim-

ilar features on their own. Older investors especially beneﬁt because

they rely less on non-ﬁnancial income, a ﬁnding that challenges robo

advisors’ stereotype as a product for millennials. Overall, our results

help explain the sustained growth of the robo market and its integra-

tion with the traditional ﬁnancial sector, despite the many other retail

products available during the FinTech era.

Our conclusions are based on the subset of retail investors who are

interested in fully automated wealth management, rather than the av-

erage U.S. investor. The investors in our sample are presumably more

sophisticated than the general U.S. population, since, for example, many

of them at least participated in the stock market before joining the

advisor. However, a unique feature of our setting is that the advisor’s al-

gorithm allocates the same robo portfolio to all investors with the same

demographic proﬁle and risk attitude, regardless of sophistication. By

adhering to this form of algorithmic fairness, the robo advisor plausibly

generates higher gains for less-sophisticated investors who either do not

participate in the stock market or do so with major pitfalls. Therefore,

our welfare calculations would constitute a lower bound with respect to

the gains for the general population. By the same logic, our estimated

eﬀect on participation would constitute an upper bound.5

Related literature

This article speaks to a nascent literature on FinTech intermedi-

aries, which includes robo advisors. Unlike prior and contemporaneous

5Recent evidence suggests that retail investors prone to particular behav-

ioral biases endogenously use FinTech products that perpetuate those biases

(e.g., Cookson et al. 2023; Barber et al. 2022; Ben-David et al. 2022). These re-

sults would imply that bias-prone investors prefer not to participate with robo

advisors because they dislike the “textbook” aspect of robo portfolios.

studies, we study automated wealth management in a pure form that

involves: complete portfolio delegation, as opposed to non-binding ad-

vice (e.g., D’Acunto et al. 2019; Bianchi and Briére 2022; D’Hondt et al.

2020); no option for human interaction (e.g., Rossi and Utkus 2021b);

and robo advisors unaﬃliated with the banking system (e.g., Loos et al.

2020). We also focus on less-wealthy investors in a quasi-experimental

setting, complementing the literature’s typical focus on more-aﬄuent

investors.6

Within the literature on robo advisors, our study shares common

ground with Reher and Sun (2019), who examine how deposit inﬂow

changes after the 2015 reduction in minimum. Several crucial dis-

tinctions set our work apart. Most importantly, our novel structural

approach enables both a rigorous calculation of welfare gains and an

assessment of the economic mechanisms driving the surge in account

formation after the reduction.7In particular, we can challenge and re-

fute three potential mechanisms left open by the atheoretical approach

in Reher and Sun (2019). First, we challenge the idea that investors

become robo participants purely to follow a popular trend; instead,

we quantitatively rationalize their response using diﬀerences between

robo and self-managed portfolios. Second, we challenge the notion that

they respond solely because they cannot form a diversiﬁed portfolio of

stocks; we show that a hypothetical investor in a low-cost equity in-

dex fund would also gain from access to robo portfolios, underscoring

the importance of more-sophisticated features like additional risk fac-

tors and glide paths. Third, we challenge the conventional assumption

that robo advisors only beneﬁt younger investors; while the young com-

prise a large share of new robo participants, the primary beneﬁciaries

are actually older investors with limited investable wealth. Our em-

pirical approach also diﬀers from Reher and Sun (2019)by rigorously

accounting for time-varying confounds, particularly targeted advertis-

ing, which enables us to credibly identify the reduction’s causal eﬀect.

Lastly, Reher and Sun (2019) only document a response by investors

with less than $107,500 in wealth, but we show that the reduction im-

pacts investors much farther down the U.S. wealth distribution, notably

those with $6,000 to $42,000 (third quintile) and especially those with

$1,000 to $6,000 (second quintile).

Our study also delivers a new perspective on the drivers of de-

mand for professional portfolio management. The literature has argued

that retail investors seek portfolio managers because managers actively

collect information about the underlying assets (e.g., Gârleanu and Ped-

ersen 2018) or confer peace of mind (e.g., Gennaioli et al. 2015). We

ﬁnd that many less-wealthy investors have more mundane needs: to

construct personalized and diversiﬁed portfolios, even if the underly-

ing assets are passively managed.8These mundane needs are not fully

met by another form of personalized asset management, TDFs (e.g., Bal-

duzzi and Reuter 2019; Parker et al. 2021), as robo advisors oﬀer more

personalization via a double glide path.

Finally, we contribute to the household ﬁnance literature by char-

acterizing account minimums as a friction that distorts investment in

risky assets. This ﬁnding demonstrates the importance of frictions that

arise from the supply side, like internet speed (e.g., Hvide et al. 2023),

as distinct from frictions that directly depend on investor characteristics

6Our study also relates to a broader question of how technological innovation

aﬀects ﬁnancial inclusion and wealth inequality. See, for example, the studies in

the contexts of app-based payments (e.g., Hong et al. 2022), bank deposits (e.g.,

Bachas et al. 2018; Bachas et al. 2020; Higgins 2022), and mortgage markets

(e.g., Fuster et al. 2019; Bartlett et al. 2021; Fuster et al. 2021).

7By studying a model of portfolio choice with a dynamic decision between

self-managed and professionally managed portfolios, we also contribute to the

literature on quantitative life cycle models summarized by Gomes (2020).

8This ﬁnding corroborates the Von Gaudecker (2015)result that Dutch in-

vestors diversify better when they consult a ﬁnancial professional. Insofar as

wealth correlates with ﬁnancial literacy, our ﬁndings are also consistent with

the Bianchi (2018)result that less-literate investors struggle to rebalance to

their optimal risk exposure.

研报加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载研报到电脑，查找使用更方便

免费 22人已下载

立即下载

标签： #财富管理

摘要：

JournalofFinancialEconomics155(2024)103829Availableonline22March20240304-405X/©2024ElsevierB.V.Allrightsreserved.ContentslistsavailableatScienceDirectJournalofFinancialEconomicsjournalhomepage:www.elsevier.com/locate/jfecRoboadvisorsandaccesstowealthmanagement✩MichaelRehera,∗,StanislavSokolinskibaUn...

展开>> 收起<<

智能投顾与财富管理可及性（英）-22页.pdf

共21页,预览7页

还剩页未读，继续阅读