Whom to Educate?Financial Fraud and Investor Awareness∗

Zhengqing Gui† Yangguang Huang‡ Xiaojian Zhao§

January 2018

Abstract

We study how investors are exploited by fraudulent financial products. These in-vestors purchase financial products that are inconsistent with their risk attitudes, andin turn, their behaviors provide an incentive for firms to commit financial fraud. Wethen conduct experiments and surveys in Shenzhen, China to measure investors’ riskpreferences and the effect of an eye-opening financial education program. Participatingin our education program significantly reduces investors’ tendency to invest in fraud-ulent products, especially among those who are risk-averse. Therefore, compared torandomly assigning the education program to investors, targeting risk-averse investorswill be more effective in fighting financial fraud.

Keywords: financial fraud, risk attitude, financial education program, unawareness.JEL Classification: D14, D83, G11, I20

∗The project is supported by the Qianhai Institute for Innovative Research and HKUST Institute forEmerging Market Studies (Grant No. IEMS17BM01). We have benefited from Shuaizhang Feng, Yuk-FaiFong, Spyros Galanis, Kohei Kawamura, Martin Peitz, Jane Zhang and Jidong Zhou for helpful suggestionsand stimulating discussions. We also thank participants in several seminars and conferences for valuablecomments.†Department of Economics, Hong Kong University of Science and Technology, Hong Kong. Email:

zgui@ust.hk‡Department of Economics, Hong Kong University of Science and Technology, Hong Kong. Email:

huangyg@ust.hk§Chinese University of Hong Kong (Shenzhen) and Hong Kong University of Science and Technology.

Email: xjzhao81@gmail.com

“Remember: all financial frauds have the same feature — high return.”

A police department in China

1 Introduction

With the rise of financial market liberalization, individuals have become responsible for their

wealth management through their private decisions. Consequently, financial illiteracy has

emerged as an important issue after financial markets became open to financially naive in-

vestors, especially as financial products have simultaneously become increasingly complex.

Financial literacy thus plays a crucial role in overcoming an investor’s information disad-

vantage in investment activities. After the 2008 financial crisis, the reliance on financial

literacy as an investor’s self-protection device has received a great amount of attention in

policy debates (Campbell et al., 2011).

The market implications of financial illiteracy emerge as a practical issue for policy mak-

ers with regard to investor protection. A large population of financially illiterate investors

opens the door for financial frauds that offer too-high-to-be-true return. In 2015, 220 thou-

sand Fanya Metal Exchange investors from 20 provinces in China lost a total 48 billion RMB

in investments.1 This event caused serious social consequences: thousands of Fanya investors

collectively traveled to Beijing to protest, as reported by Reuters2, The Independent3, and

BBC 4. The case of Fanya is a typical case of a Ponzi scheme that claims unrealistically high

return without providing any information about risk. The misleading product descriptions

may induce investors to over-invest in products that are not consistent with their risk atti-

tudes.5 In June 2017, the China Central Television reported a list of 350 cases of financial

fraud occurred since 2016. Based on these cases, a police department in China gave a simple

and clear warning: “all financial frauds have the same feature — high return.”6

The fact that Fanya and other fraudulent financial products attracted so much investment

suggests that some investors may be poorly informed about the possibility of financial fraud.

1See “China cracks down on alleged $7.6 billion Ponzi scheme”, http://money.cnn.com/2016/02/01/investing/china-ezubao-alleged-ponzi-scheme-arrests/index.html.

2See “The China metal exchange at center of investment scandal”, http://www.reuters.com/article/us-china-metals-fanya-insight-idUSKBN0TX00K20151214.

3See “28 Chinese investors in the Fanya Metal Exchange have gone missing after a protest”,http://www.independent.co.uk/news/business/news/28-chinese-investors-in-the-fanya-metal-

exchange-have-gone-missing-after-a-protest-a6713861.html.4See “China protests by Fanya Metal Exchange investors”, http://www.bbc.com/news/world-asia-

china-34318344.5For the Fanya product description, see http://chuansong.me/n/659162852657.6http://www.sohu.com/a/153362230_479788.

1

In this context, to what extent can some eye-opening education programs increase investors’

awareness of financial fraud and help them make better decisions when choosing financial

products?

To answer this question, we build a model to study how investors are exploited by financial

products with too-high-to-be-true return. There is a fraction of boundedly rational investors

who are unaware that a firm can commit financial fraud, such as running away with the

money as a moral hazard consequence.7 Unaware investors purchase financial products that

are inconsistent with their risk attitudes, and in turn, their behaviors provide incentives for

the firm to commit financial fraud. Reducing the proportion of unaware investors compels

the firm to behave honestly, and financial fraud disappears if this proportion drops below a

certain threshold.

With this insight, we conducted experiments and surveys in Shenzhen, China with the

support from the Qianhai Institute of Innovative Research (QIIR).8 Specifically, we recruited

1216 effective respondents distributed in 30 communities in the city of Shenzhen. We ex-

perimentally measured investors’ risk preferences by the multiple-price-list method9 and

conducted an eye-opening education program that intended to help investors make reason-

able financial investment decisions. Additionally, we surveyed these investors’ demographic

information and financial literacy. Our data suggest that risk-averse preferences are often

observed among female and elderly investors.10 Regarding financial literacy, Shenzhen resi-

dents receive similar scores as the survey results in the United States (Lusardi and Mitchell,

2014). Moreover, we do not find any strong relationship between financial literacy and risk

preferences.

Based on respondents’ hypothetical investment decisions, our eye-opening education pro-

gram reduced an investor’s tendency to be involved in financial fraud to a large extent, even

if the education takes a highly simple form. According to our experiments and surveys,

71.62% of the investors invest in the financial product with too-high-to-be-true return. Our

education program reduces this percentage to 63.89%. The efficacy of the education program

depends on investors’ risk attitudes, financial literacy measure, and demographic character-

istics. In particular, we find that risk-averse investors are most sensitive to the education

program. 71.88% of the risk-averse investors choose to buy fraudulent products. In the

presence of our education program, this figure decrease to 55%.

7See “China’s P2P lending boom: Taking flight”, https://www.economist.com/news/finance-and-

economics/21688940-allure-and-peril-chinese-fintech-companies-taking-flight.8http://www.qiir.org/.9The multiple-price-list format is one of the standard methods to elicit risk preferences of experimental

subjects. See Holt and Laury (2002) and Andersen et al. (2006).10This experimental result echoes the existing behavioral and experimental economics literature, e.g.,

Croson and Gneezy (2009) and Albert and Duffy (2012).

2

The fact that some investors make investment decisions that are inconsistent with their

risk attitudes and the effectiveness of our education program suggest that these investors are

unaware of the relationship between high return and high risk. Based on our model, investors’

unawareness of the possibility of financial fraud contributes to the spread of financial fraud.

As long as the proportion of unaware investors is significantly reduced, the firm’s incentive

to offer fraudulent products can be attenuated or even eliminated.

Our findings have some important policy implications for preventing financial fraud in

financial markets. Our simple education program has demonstrated a significant effect in

helping some unaware investors become aware of the potentially high risk associated with

a high-return product. Policy makers can further efficiently utilize limited educational re-

sources by targeting risk-averse investors, whom can be identified by survey questions as well

as the link between investors’ risk attitudes and their observed characteristics. For example,

women and the elderly are likely to demonstrate risk-averse preferences, as suggested in our

study. Thus, in promoting financial education programs, one can consider focusing on these

two groups of investors. As a policy practice, the China Security Regulatory Commission im-

poses a mandatory risk preference survey on investors before their purchase of risky financial

products.11 Adding an educational component to this survey will be beneficial to investors.

Following our paper, this mandatory educational component can be also personalized based

on investors’ answers to survey questions.

In addition to its policy relevance, our paper also contributes to two strands of literature.

Bounded rationality and its application to contracting problems has been a growing topic

in economic theory. Contracting with unforeseen contingencies belongs to this emerging

literature (see, e.g., Chapter 5.3 in Spiegler, 2011). In industrial organization, consumers

may be unaware and thus exploited by firms through deceptive products (e.g., Gabaix and

Laibson, 2006; Heidhues, Koszegi and Murooka, 2017). von Thadden and Zhao (2012),

Auster (2013) and the present paper introduce asymmetric awareness into moral hazard

problems. While the investor is unaware of the firm’s certain option in the present paper,

in von Thadden and Zhao (2012), the agent is possibly unaware of some of his own actions,

and the agent is unaware of some contingencies in Auster (2013).

Moreover, our paper contributes to the study of financial literacy and its influence on

people’s investment behaviors. Lusardi and Mitchell (2014) and Fernandes, Lynch and Nete-

meyer (2014) provide excellent reviews of the literature on financial literacy and report the

worldwide findings of financial illiteracy, but data from China have been largely underde-

veloped.12 Consistent with reports around the world, we find that Chinese investors are

11The survey asks for age, income, wealth, investment experience, and some hypothetical investmentdecision questions. See http://www.sac.net.cn/tzzyd/fzgj/201205/t20120503_15058.html.

12Song (2015) conducts a survey in rural China and finds a strong relationship between the neglect of com-

3

subject to an inadequacy of financial knowledge as those in other countries, but the Chinese

respondents exhibit a lower self-confidence about their knowledge in finance. Moreover, we

do not observe an apparent gender difference in financial literacy as in many other countries

reported in Lusardi and Mitchell (2014). On top of it, we do not find that the three standard

financial literacy questions in the literature can predict an investor’s tendency to purchase

fraudulent products. Therefore, we call for adding another question concerning the basic

high-return-high-risk theorem, making the financial literacy test more valuable in predicting

how much investors can protect themselves from financial fraud. Besides, a few papers have

attempted to theoretically study the link between financial literacy and investment decisions,

as well as the (mis-)selling behaviors of financial professionals in the market (e.g., Inderst

and Ottaviani, 2009, 2012 and 2013; and Lusardi, Michaud and Mitchell, 2017). In contrast

to these papers, we explicitly model the investor’s unawareness as a special form of financial

illiteracy.

The paper proceeds as follows. Section 2 introduces the model. Section 3 presents our

experimental and survey design. Section 4 reports our experimental and empirical results.

Section 5 concludes.

2 Model

In this section, we introduce a simple model that captures the interaction between a firm

and an investor who is possibly unaware that the firm may commit financial fraud.

A firm needs outside finance from investors to fund a risky project. The project’s outcome

θ can be publicly observed. If the project succeeds (θ = 1), it will generate a constant rate

of return R > 0 to the firm. If the project fails (θ = 0), the firm’s return will be normalized

to zero. Let the project succeed with probability p. Assume p(1 +R) > 1, implying that the

investment is profitable.

A contract specifies a rate of return r repaid to the investor when θ = 1, and zero

otherwise. Before offering the contract, the firm chooses whether to commit financial fraud

(b = 0) or behave honestly (b = 1) — moral hazard. If the firm chooses b = 0, it refuses to

repay (1 + r)I when θ = 1 and incurs a cost c > 0, which can be interpreted as the expected

reputation loss or legal punishment. If the firm chooses b = 1, it repays (1 + r)I to the

investor as stated in the contract when θ = 1.

Investors are endowed with identical initial wealth w, and are heterogeneous in their risk

attitudes: they have Bernoulli utility u = mα over the monetary payment m, where α is

pound interest and pension contributions. Moreover, Song (2015) finds that financial education contributesto a correct understanding of compound interest, and thus increases individuals’ investment in pension plans.

4

distributed over (0,+∞) with a commonly known c.d.f. F (α) and p.d.f. f(α).13

The timing of the game is as follows. The firm first chooses its action b and then offers a

contract with a rate of return r to the investor. If the investor rejects the offer, both parties

get zero payoffs from the project and the game ends. If the investor accepts the offer, he/she

decides the amount of investment I, and then the project’s cash flow is realized. The timing

specified in our model requires that the firm decides whether to commit financial fraud in

the early beginning. This assumption is based on the observation that if the firm decides to

commit financial fraud, it involves certain preparations for taking this action before actually

offering the fraudulent product to the investor and the realization of the project outcome.14

For simplicity, we will consider only pure strategies.

Given that the investor accepts the offer, his/her expected utility E(u) and the firm’s

expected profit E(π) are determined by the following:

E[u(b, r, I)] = pb(w + rI)α + (1− pb)(w − I)α,

E[π(b, r)] = pI[1 +R− b(1 + r)]− (1− b)c.

Given that the firm behaves honestly (b = 1), the investor determines his/her investment

level by maximizing

E[u(1, r, I)],

subject to the participation constraint

p(w + rI)α + (1− p)(w − I)α ≥ wα.

It is easy to see that when r = 0, I∗ = 0 is the only solution. When r ∈ (0, R], the first

order derivative is given by

∂E[u(1, r, I)]

∂I= αpr[(w + rI)α−1 − 1− p

pr(w − I)α−1].

When α < 1, E[u(1, r, I)] is concave. Therefore if r ∈ (0, 1−pp

], there is a corner solution

I∗ = 0; if r ∈ (1−pp, R], there is an interior solution

I∗(α, r) =1− (1−p

pr)

11−α

1 + r(1−ppr

)1

1−αw.

13Our utility form modifies the isoelastic utility function and thus allows for risk-seeking preferences.14For example, the firm has to prepare exaggerating and sometimes unreal marketing materials to at-

tract potential investors or provide a forged auditor’s report in order to deceive the regulators. Detaileddescriptions of these preparations can be found in many court cases. See, e.g., http://im.ft-static.com/content/images/b6eabf92-b631-11e1-8ad0-00144feabdc0.pdf.

5

When α ≥ 1, E[u(1, r, I)] is convex. Therefore if r ≤ p−1α −1, or equivalently α ≤ − log1+r p,

there is a corner solution I∗ = 0; if r ≥ p−1α − 1, or equivalently α ≥ − log1+r p, there is a

corner solution I∗ = w. In particular, when α = 1, r = 1−pp

, any I ∈ [0, w] can be a solution.

The firm chooses r by maximizing

E[π(1, r)] =

0 if r = 0

p(R− r)∫ +∞− log1+r p

wdF (α) if r ∈ (0, 1−pp

)

p(R− r)∫ +∞1

wdF (α) if r = 1−pp

p(R− r)∫ 1

0I∗(α, r)dF (α) + p(R− r)

∫ +∞1

wdF (α) if r ∈ (1−pp, R].

The first order derivative is given by

∂E[π(1, r)]

∂r=

{pw[F (− log1+r p)− 1− R−r

(1+r) ln(1+r)f(− log1+r p) log1+r p] if r ∈ (0, 1−p

p)

p∫ 1

0[(R− r)∂I

∗(α,r)∂r

− I∗(α, r)]dF (α)− p∫ +∞1

wdF (α) if r ∈ (1−pp, R].

Note that I∗(α, r)→ 0, − log1+r p→ 1 as r → 1−pp

, so E[π(1, r)] is continuous on the closed

interval [0, R], which ensures the existence of a global maximum. Let this maximum be

attained at r∗. Since E[π(1, 0)] = E[π(1, R)] = 0, r∗ ∈ (0, R). Then either r∗ = 1−pp

, or, if

r∗ ∈ (0, 1−pp

), it solves

∂E[π(1, r)]

∂r

∣∣∣r∗∈(0, 1−p

p)

= 0,

if r∗ ∈ (1−pp, R), it solves

∂E[π(1, r)]

∂r

∣∣∣r∗∈( 1−p

p,R)

= 0.

Suppose that some investors are unaware that the firm can choose b = 0. In other words,

they mistakenly believe that the firm can only choose b = 1 and do not have financial fraud

in mind.15 An aware investor, knowing the possibility of financial fraud, will reject any

contract with r 6= r∗ which is a non-justifiable contract (see, e.g., Auster, 2013), while an

unaware investor will accept a contract unless r > R.16 Suppose that a fraction λ ∈ (0, 1) of

investors is unaware, while the remaining 1 − λ is aware. Therefore, if the firm chooses to

commit financial fraud (b = 0), it will propose r = R to attract as many unaware investors

15Unlike other applied works in awareness of unawareness, e.g. Chung and Fortnow (2016), Tirole (2009,2014) and Zhao (2015), we implicitly assume that the investors are unaware that they may be unaware.

16More rigorously, unaware investors will not accept any contract that leads to a negative profit for thefirm from their perspectives. Here, we still make a justifiability assumption for unaware investors to acceptthe contract. Otherwise, the firm will propose r =∞ to attract unaware investors with all the risk attitudes,leading to a less plausible prediction in practice. Thus, even for unaware investors, a certain degree ofcognitive sophistication is required. Nonetheless, we adopt a weaker version of the justifiability conditionthan Auster (2013).

6

as possible,17 implying that its expected profit is

E[π(0, R)] = λp(1 +R)[

∫ 1

0

I∗(α,R)dF (α) +

∫ +∞

1

wdF (α)]− c.

In sum, the firm will propose r = r∗ and choose b = 1 if and only if

E[π(1, r∗)] ≥ E[π(0, R)],

or, equivalently,

c ≥ λp(1 +R)[

∫ 1

0

I∗(α,R)dF (α) +

∫ +∞

1

wdF (α)]− E[π(1, r∗)] ≡ c∗.

Proposition 1 There exists a cutoff c∗ such that: if c ≥ c∗, the firm proposes r = r∗ ∈(0, R), chooses b = 1 (to be honest), risk-averse investors accept the offer when r∗ ∈ (1−p

p, R),

risk-neutral and risk-seeking investors accept the offer when α ≥ − log1+r∗ p; if c < c∗, the

firm proposes r = R, chooses b = 0 (to commit financial fraud), and only unaware investors

accept the offer.

According to Proposition 1, the firm’s optimal choice depends on the relationship between

its own cost of default c and the cutoff point c∗. On the other hand, the amount of investment

made by the investor is determined by both his/her risk preference and awareness. Risk-

seeking unaware investors will always invest the largest possible amount.

Note that when

λ ≤ E[π(1, r∗)]

p(1 +R)[∫ 1

0I∗(α,R)dF (α) +

∫ +∞1

wdF (α)]≡ λ∗,

c∗ becomes nonpositive, meaning that the firm has no incentive to commit financial fraud at

all. We summarize our results in Corollary 1.

Corollary 1 c∗ increases with λ, i.e., a decrease in λ implies a decrease in the firm’s incen-

tive to commit financial fraud. Specifically, there exists a cutoff λ∗ such that when λ ≤ λ∗,

the firm has no incentive to commit financial fraud irrespective of c.

17If instead, the firm proposes r = r∗, since we focus on pure strategies, aware investors will accept theoffer only when they can ensure that the firm will be honest. In other words, if aware investors accept anoffer with r = r∗, it must be in the firm’s interest to choose b = 1 since aware investors are fully rational.On the other hand, suppose that aware investors reject any offer with r = r∗, the firm may propose r = Rrather than r∗ and play b = 0 thereafter since aware investors will not accept the firm’s offer anyway. In sum,given the timing specified in our model, the firm proposing r = r∗ and playing b = 0 cannot be sustainedsimultaneously unless we allow for their mixed-strategies.

7

Corollary 1 implies that in a market with a larger proportion of aware investors, a firm

is more likely to behave honestly.18 Given that education can reduce λ, educating unaware

investors not only partially prevents them from being mis-sold as a direct effect, but also

indirectly reduces or even eliminates financial fraud.

Corollary 2 (High-return-high-risk theorem) If the firm proposes r = r∗, the default risk is

1− p; if the firm proposes r = R, the default risk is 1.

In terms of Galanis (2013), a fraction λ of investors are unaware of “theorems”, as

they are not aware that the firm can commit financial fraud. In the following sections, we

will experimentally study the extent to which financial education program can reduce this

fraction. Our education program is about the message that “financial products with high

return always have high risk”. This message does not involve any uncertainty resolution.

Therefore, one can also consider the relationship between risk and return as a “theorem”

and our message as an eye-opening treatment to investors.

3 Experimental Design and Survey Data

We conducted experiments and surveys in 2016. Our experiments aimed to examine whether

education programs could improve the financial decisions associated with an investor’s risk

preference and other demographic characteristics, including financial literacy.

We randomly chose 30 communities in Shenzhen as our experimental sites as depicted in

the map in Figure 1. We went to these communities and collaborated with each community’s

residents committee to obtain a full list of all residents living in the community. Then, we

randomly chose residents from each community and conducted face-to-face experiments and

surveys. In order to raise the response rate, we prepared a USB cable as a small reward for

participation. In sum, we recruited 1216 effective participants after removing 70 respondents

who provided inconsistent responses in part 3 of our incentivized experiment.19

We randomly assigned the participants into four groups as shown in Table 1. In groups

Y1 and Y0, we gave each participant an education flyer and provided detailed instructions

18This self-reinforcing pattern is similar to Gabaix and Laibson (2006), von Thadden and Zhao (2012),and Li, Peitz and Zhao (2016) who study a monopolistic firm’s behavior of revealing adverse effects tounaware consumers, and find that a larger share of unaware consumers reduces the firm’s incentive to discloseinformation. Relatedly, Zhou (2007) suggests that naive consumers tend to overestimate the importance ofcertain attributes of a product after firms’ advertising and finds that increasing the proportion of naiveconsumers will also reduce the surplus of sophisticated consumers.

19In part 3 of our experiment, as shown in Appendix A, we let the participants indicate their preferencesbetween a lottery and a constant amount of money. If one preferred a smaller amount of money to the lotterybut again chose to play the same lottery rather than accept a higher amount of money, then the participant’sresponse is inconsistent because it violates the axiom of transitivity of risk preference if monotonicity overmoney is implicitly assumed.

8

Figure 1: Distribution of Participants in Shenzhen, China

to teach them that “financial products with high return always have high risk” before they

answered the survey and experimental questions. We did not provide any educational inter-

vention in groups N1 and N0. We introduced a fraudulent product with an annual return

of 20% to the participants in group Y1 and N1,20 and a normal product with an annual

return of 8% in group Y0 and N0.21 A complete English translation of the education flyer,

survey and experimental instruction of group Y1 can be found in Appendix A. The original

education flyer is shown in Appendix B.

Table 1: Treatments Adopted in Four Groups

Education programNo Yes

Normal product (8% return) N0 Y0Observations 304 304

Fraudulent product (20% return) N1 Y1Observations 306 306

The survey and experimental questionnaire consisted of three parts. In part 1, we adopted

20Following the Fanya scandal in China, the options we provided in the fraudulent products were exactlythe same as those in the Fanya products in practice (with minor revisions), but we did not explicitly mentionthe names of the products and the Fanya company. For more details about the Fanya products and the Fanyaevents, see the article in Chinese http://blog.sina.com.cn/s/blog_9e2674c30102wrnd.html.

21As a comparison, the average annual return of Dow Jones Industrial Average index in the past 30 yearsis approximately 8.13%.

9

the four questions in Lusardi and Mitchell (2014) to measure participants’ objective and

self-assessed financial literacy, followed by a hypothetical investment decision question. The

participants were required to report whether they would invest in the listed (fraudulent or

normal) product, and if so, how much to invest.

In part 2, we asked the participants a series of peripheral questions to collect some demo-

graphic information, including gender, age, education, income, and their previous experience

in financial activities.

In part 3, we used the multiple-price-list format in an incentivized experiment to elicit

each participant’s risk attitude. In particular, we designed ten questions, asking the partic-

ipants to choose between a constant reward amount ranging from 5 RMB to 95 RMB and a

lottery with a 50% probability of rewarding 100 RMB and a 50% probability of rewarding

nothing.

In sum, the questionnaire collects information on participants’ demographics, financial

literacy, and risk preferences. By comparing the results in the groups in the dimension of

1 and 0, we can investigate the difference in participants’ purchasing behaviors between a

fraudulent product and a normal product. Similarly, by comparing groups in the dimension

of Y and N, we can analyze the effect of education on their purchasing behaviors. The results

are reported as follows.

4 Experimental and Empirical Results

Table 2 shows the summary statistics of all the variables analyzed in this section. Two

dummy variables, Education treatment and Fraudulent product, indicate whether the re-

spondent receives our education program and whether a fraudulent product is offered to the

respondent, respectively. Edu*Fraud is their product. Literacy Q1 -Q3 indicate whether

the respondent correctly answers the 1st, 2nd, and 3rd questions regarding his/her objec-

tive financial literacy. Confidence measures respondents’ self-assessed knowledge of finance.

Investment indicates the respondent’s choice in the investment decision question, where a

higher choice of Investment represents a higher level of investment, i.e., Investment = 5 if

the respondent chooses to “invest 2,600 yuan or more” and Investment = 0 if the respon-

dent does not invest at all. In future analyses, we will also use a binary variable, Investment

dummy, to indicate whether Investment is strictly positive. Male, Single, Schooling, Income,

Past involvement, and Past experience are categorical demographic variables that correspond

to the respondent’s answer in question 1 to 6, and question 9 in part 2 of the questionnaire.

10

The alphabetical choices in these questions are transformed into positive integers in order.22

Risk preference measures our respondents’ risk attitudes. Risk preference = J means that

the respondent answers “Disagree” only in the first J questions in part 3 of the questionnaire,

so a larger J implies a more risk-seeking preference.

Table 2: Summary Statistics

Variable Mean Std. Dev. Min. Max. NEducation treatment 0.498 0.5 0 1 1216Fraudulent product 0.503 0.5 0 1 1216Edu*Fraud 0.252 0.434 0 1 1216Literacy Q1 0.693 0.462 0 1 1214Literacy Q2 0.615 0.487 0 1 1210Literacy Q3 0.55 0.498 0 1 1210Confidence 2.655 1.792 1 7 1152Investment 3.358 2.049 1 6 1156Investment dummy 0.687 0.464 0 1 1156Age 37.04 12.99 12 95 1161Male 0.342 0.474 0 1 1215Single 0.258 0.438 0 1 1210Schooling 4.094 0.972 1 6 1212Income 1.992 0.881 1 5 1195Past experience 0.538 0.499 0 1 1198Past involvement 1.619 0.756 1 5 1206Risk preference 5.535 2.943 0 10 1158

Figure 2 illustrates an overview of our education program’s effectiveness. It compares the

fraction of investors in the presence of education who choose each of the six choices in our

hypothetical investment decision question with fraudulent products with that in the absence

of education, i.e., it tells us the difference between group N1 and group Y1, using Investment

as the horizontal axis. It shows that our education program reduces the fraction of investors

purchasing fraudulent products at each level of investment.

4.1 Risk Preference

Figure 3 depicts the distribution of the respondents’ risk attitudes. The horizontal axis shows

the value of α calculated from each respondent’s answer in part 3 of the questionnaire, based

on the utility functions in our model. Specifically, if the respondent switches from “Disagree”

to “Agree” in the J + 1th question, i.e., Risk preference = J , we take the midpoint of

22We transform the multiple-choice question in part 2 of the questionnaire on respondents’ marital statusinto a dummy variable Single because the size of the sample with “Divorced or separated” or “Widowed” istoo small. Therefore, Single equals 1 if the respondent never gets married, and 0 otherwise.

11

Figure 2: Distribution of Investment in Group N1 and Y1

the two constant return in the Jth and J + 1th questions as the respondent’s certainty

equivalent. Therefore, if Risk preference = 5, the respondent is identified as risk-neutral.

If Risk preference is higher or lower than 5, the respondent is classified as risk-seeking or

risk-averse, respectively.

Table 3 shows the two-sample t-test results for Risk preference for each of the three

dummy variables of financial literacy, i.e., Literacy Q1 -Q3. In none of the three tests can

we reject the null hypothesis that an investor’s financial literacy and risk attitude are inde-

pendent.23

Table 3: Relationship between Risk Preference and Financial Literacy

Incorrect Correct T-test p-valueLiteracy Q1 5.32 5.62 0.12

(0.17) (0.10)Literacy Q2 5.41 5.61 0.27

(0.15) (0.11)Literacy Q3 5.50 5.55 0.75

(0.13) (0.12)

Regarding demographic information, 55.01% of the respondents are risk-averse or risk-

neutral, among which 69.34% are female, 41.35% have a college education or equivalent, and

76.76% have an annual income of no more than 50,000 RMB (7,250 USD). Table 4 presents

the results of regressing Risk preference over other demographic variables using an ordered

probit model. From the regression results, we find that Risk preference is negatively corre-

23For this reason, we assume λ and α are independent in the model.

12

Figure 3: Distribution of Risk preference

Table 4: Ordered Probit Model

Dependent variable: Risk preference(1) (2) (3) (4) (5)

Age -0.00714∗∗∗ -0.00604∗∗ -0.00613∗∗

(0.00242) (0.00289) (0.00289)Male 0.181∗∗∗ 0.168∗∗

(0.0635) (0.0667)Single 0.147∗∗ 0.0670 0.0258

(0.0692) (0.0838) (0.0854)Observations 1106 1157 1152 1102 1102Pseudo R2 0.002 0.002 0.001 0.002 0.003∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.

13

lated with Age and positively correlated with Male. The correlation between Risk preference

and Single becomes insignificant after controlling for Age, thus it could be conjectured that

the vanishing correlation is due to the collinearity between Single and Age. In sum, men

and youth people are more risk-seeking than women and the elderly.

4.2 Eye-opening Education Program

In this subsection, we apply the Difference-in-Difference technique (DID) to estimate the

net effect of our eye-opening financial education program on eliminating investors’ over-

investment in fraudulent products. While we use random assignment in groups Y1 and

N1, comparing the results in groups Y0 and N0 still plays a key role in controlling for the

placebo effect. It is possible that our action of offering education, rather than the content

of education per se, affects participants’ willingness to invest. Therefore, by introducing two

control groups, Y0 and N0, and applying the DID technique, we can potentially control the

factors that may simply raise one’s tendency to invest, irrespective of the product’s return.

In Table 5, we conduct t-tests between four groups to compare their risk preferences

and the proportion of respondents who choose a positive investment level in the investment

decision question. We do not observe a significant difference in risk preferences among the

four groups, while the education program does make a difference in investment decision. This

suggests that the education program affects investors’ awareness, but not their preferences.24

Table 6 illustrates the DID regression results using probit and ordered probit models

with Investment dummy and Investment as dependent variables, respectively. Both t-tests

in Table 5 and regression results in Table 6 suggest that our education program significantly

reduces respondents’ tendency to invest in fraudulent products, even after controlling for a

series of demographic variables.

To further study the association between education program and respondents’ risk at-

titudes, we conduct DID regressions over respondents with different risk attitudes. Table

7 and Table 8 present the estimation results. We find that risk-averse investors are most

sensitive to our education program.

Following the theory in Section 2, λ represents the proportion of investors who are un-

aware of the link between the too-high-to-be-true return on a financial product and its

underlying high risk due to moral hazard. Therefore, we estimate λ using the proportion of

24Change in behavior is driven by change in either preferences or beliefs. Since risk preferences are notaltered after education, the program only changes the subjects’ beliefs. Moreover, telling them a theoremdoes not give any uncertainty resolution about the financial product. Instead, the program is an eye-openinginformation provision expanding their awareness. Relatedly, Kawaguchi, Uetake and Watanabe (2016) showthat advertisements may influence both consumers’ attention and preferences, which can be identified bytheir product availability approach.

14

Table 5: Differences in Risk preference and Investment dummy

Education programNo Yes Difference

Risk preference Mean 5.562 5.560 0.002(0.182) (0.164) (0.245)

Normal Observations 290 291product Investment dummy Mean 0.682 0.711 -0.029

(0.027) (0.027) (0.039)Observations 292 280

Risk preference Mean 5.463 5.552 -0.088(0.178) (0.168) (0.245)

Fraudulent Observations 287 290product Investment dummy Mean 0.716 0.639 0.077**

(0.026) (0.028) (0.039)Observations 296 288

Risk preference Mean 0.090DID (0.346)

Investment dummy Mean -0.107*(0.055)

participants who choose to invest a positive amount in the fraudulent product in groups N1

and Y1. Overall, we obtain that λ = 71.62% for N1 and λ = 63.89% for Y1, which means

that our education program helps 7.73% investors become aware of financial fraud.

Regarding risk attitude, Figure 4 illustrates λ for participants with different Risk prefer-

ence. The proportion of risk-averse participants who are exploited by the fraudulent product

decreases from 71.88% in group N1 to 55% in group Y1, which indicates that the education

program targeting risk-averse participants is twice as effective as a randomized education

program. Therefore, if policy makers can identify risk-averse investors, the efficacy of an eye-

opening education program can be greatly improved. As previously mentioned, the China

Security Regulatory Commission has already imposed a mandatory risk preference survey on

investors.25 Based on this practice, policy makers can further improve the survey and add a

personalized education program to investors before they make their investment decisions.

4.3 Financial Literacy

Question 1-3 in part 1 in Appendix A are the standard questions as shown in Lusardi and

Mitchell (2014) to objectively test an individual’s (i) numeracy and capacity to calculate

compound interest rates, (ii) comprehension of inflation, and (iii) the sense of risk diversi-

fication, respectively. Table 9 describes the financial literacy patterns in China. Our study

25The survey asks ten questions, and then, each investor is told that he or she belongs to one of three cate-gories: conservative, prudent, or proactive. See http://www.sac.net.cn/tzzyd/fzgj/201205/t20120503_

15058.html.

15

Table 6: Probit and Ordered Probit Model (whole sample)

Dependent variable: Dependent variable:Investment dummy Investment

(1) (2) (3) (4)Education treatment 0.0836 0.0830 -0.0718 -0.0569

(0.110) (0.136) (0.0907) (0.106)Fraudulent product 0.0997 0.0777 -0.165∗ -0.170

(0.109) (0.133) (0.0899) (0.105)Edu*Fraud -0.300∗ -0.391∗∗ 0.249∗ 0.291∗

(0.154) (0.190) (0.128) (0.151)Literacy Q1 -0.0988 -0.0163

(0.113) (0.0871)Literacy Q2 -0.152 0.0494

(0.109) (0.0839)Literacy Q3 0.258∗∗ -0.198∗∗

(0.106) (0.0833)Confidence 0.0341 -0.0553∗∗

(0.0293) (0.0233)Age -0.00759 0.00690∗

(0.00517) (0.00417)Male -0.0661 -0.0633

(0.111) (0.0878)Single -0.0737 0.121

(0.142) (0.112)Past experience -0.239∗∗ 0.267∗∗∗

(0.110) (0.0867)Risk preference No Yes No YesSchooling No Yes No YesIncome No Yes No YesPast involvement No Yes No YesCommunity fixed effect No Yes No YesObservations 1156 917 1156 917Pseudo R2 0.003 0.112 0.001 0.045∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.

16

Table 7: Probit Model (by risk preference)

Dependent variable: Investment dummyRisk-averse Risk-neutral Risk-seeking

(1) (2) (3) (4) (5) (6)Education treatment 0.142 0.214 -0.125 0.124 0.101 0.150

(0.202) (0.295) (0.237) (0.365) (0.164) (0.221)Fraudulent product 0.126 0.279 -0.153 0.359 0.135 0.102

(0.197) (0.268) (0.244) (0.401) (0.163) (0.213)Edu*Fraud -0.596∗∗ -0.773∗ 0.191 -0.257 -0.265 -0.346

(0.281) (0.402) (0.330) (0.513) (0.235) (0.307)Literacy Q1 0.272 -0.395 -0.337∗

(0.223) (0.319) (0.186)Literacy Q2 -0.291 -0.223 -0.00664

(0.221) (0.300) (0.163)Literacy Q3 -0.00143 -0.201 0.400∗∗

(0.220) (0.274) (0.174)Age -0.0147 -0.000422 -0.0104

(0.0107) (0.0157) (0.00821)Male 0.102 0.0800 -0.217

(0.238) (0.314) (0.173)Confidence 0.0941 0.0690 0.0289

(0.0655) (0.0872) (0.0447)Single -0.365 0.354 -0.0736

(0.312) (0.389) (0.214)Past experience -0.677∗∗∗ -0.636∗∗ 0.0223

(0.231) (0.311) (0.185)Schooling No Yes No Yes No YesIncome No Yes No Yes No YesPast involvement No Yes No Yes No YesCommunity fixed effect No Yes No Yes No YesObservations 346 268 259 172 500 392Pseudo R2 0.016 0.201 0.001 0.221 0.002 0.149∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.

17

Table 8: Ordered Probit Model (by risk preference)

Dependent variable: InvestmentRisk-averse Risk-neutral Risk-seeking

(1) (2) (3) (4) (5) (6)Education treatment -0.0353 0.0189 0.0225 -0.0791 -0.114 -0.157

(0.166) (0.217) (0.193) (0.252) (0.136) (0.163)Fraudulent product -0.107 -0.183 -0.108 -0.498∗ -0.220 -0.214

(0.163) (0.207) (0.201) (0.264) (0.135) (0.162)Edu*Fraud 0.395∗ 0.383 0.0431 0.242 0.260 0.367

(0.236) (0.293) (0.272) (0.357) (0.196) (0.232)Literacy Q1 -0.301∗ 0.0966 0.137

(0.167) (0.228) (0.133)Literacy Q2 0.114 0.0366 -0.0518

(0.168) (0.213) (0.122)Literacy Q3 -0.0346 0.151 -0.435∗∗∗

(0.168) (0.194) (0.127)Age 0.00157 0.0157 0.00349

(0.00806) (0.0112) (0.00630)Male -0.0377 -0.175 -0.000977

(0.177) (0.214) (0.131)Confidence -0.0680 -0.0940 -0.0610∗

(0.0473) (0.0603) (0.0347)Single 0.159 -0.0992 0.132

(0.223) (0.268) (0.164)Past experience 0.747∗∗∗ 0.561∗∗∗ -0.0423

(0.173) (0.208) (0.137)Schooling No Yes No Yes No YesIncome No Yes No Yes No YesPast involvement No Yes No Yes No YesCommunity fixed effect No Yes No Yes No YesObservations 346 283 259 211 500 423Pseudo R2 0.005 0.101 0.001 0.126 0.002 0.076∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.

18

Figure 4: λ and Risk preference in Group N1 and Y1

suggests that compared to the patterns in the United States (Table 10), Chinese participants

appear to have a higher tendency to choose “don’t know” and “refuse” in response to all

three questions. While Chinese participants are more likely to mark a correct answer in

question 1 and 3, this is not the case for question 2. As suggested in the international evi-

dence documented in Lusardi and Mitchell (2014), national historical experiences may play

a role. In particular, countries that had planned economies in the past, such as Romania

and Russia, exhibit lower knowledge of inflation.

Table 9: Financial Literacy Patterns in Shenzhen, China

Responses ObservationsCorrect Incorrect DK Refuse

Compound interest 69.28% 17.88% 11.12% 1.73% 1214Inflation 61.49% 19.91% 16.94% 1.65% 1210

Stock risk 54.96% 12.23% 28.76% 4.05% 1210

Number of correct responses Observations3 2 1 0

Proportion 30.60% 36.65% 21.06% 11.69% 1206

Note: DK = respondent indicated “don’t know.”

Table 11 summarizes the respondents’ self-assessed financial knowledge. We find that

the average score is lower than that in the US data reported in Table 12, implying that the

Chinese respondents had lower self-confidence about their financial knowledge. Relatedly,

Lusardi and Mitchell (2014) find that Japanese respondents report the lowest self-assessed

financial knowledge. This confirms the literature in psychology and economics literature on

19

Table 10: Financial Literacy Patterns in the US

ResponsesCorrect Incorrect DK Refuse

Compound interest 67.1% 22.2% 9.4% 1.3%Inflation 75.2% 13.4% 9.9% 1.5%

Stock risk 52.3% 13.2% 33.7% 0.9%

Number of correct responses3 2 1 0

Proportion 34.3% 35.8% 16.3% 9.9%

Note: DK = respondent indicated “don’t know.”Source: Lusardi and Mitchell (2014)

studying the differences in self-confidence across cultures and its implications for investment

behaviors.26

Table 11: Self-Reported Financial Literacy in Shenzhen, China

Scores 1 2 3 4 5 6 7 AveragePercentage 39.50% 15.36% 16.93% 9.72% 10.50% 2.86% 5.12% 2.65Observations: 1152

Table 12: Self-Reported Financial Literacy in the US

Scores 1-2 3 4 5 6 7 AveragePercentage 3.9% 5.2% 14.9% 33.2% 26.1% 13.6% 5.1Source: Lusardi and Mitchell (2014)

Figure 5 depicts the financial literacy patterns by age. Consistent with Lusardi and

Mitchell (2014), we find that financial literacy is lower among the young and the old than

middle-aged people.

Figure 6 depicts gender differences in financial literacy. Interestingly, we do not observe

an obvious gender difference, unlike the differences found in many other countries reported

in Figure 1 in Lusardi and Mitchell (2014). Based on evidence from many countries, Hsu

(2016) proposes a hypothesis of the specialization of labor and knowledge within households.

Our study finds that this hypothesis does not hold in China.

Figure 7 depicts the differences in financial literacy according to schooling. One may find

that participants with postgraduate degrees have a lower percentage of completely correct

responses than those with college degrees. This may be a result of our small sample size

(only 16) of participants with postgraduate degrees.

26See, e.g., Heine et al. (1999), Schmitt and Allik (2005), and Dessi and Zhao (2015).

20

Lastly, and more importantly, based on results in Table 6, 7, and 8, we do not find that

the three standard financial literacy questions can predict an investor’s awareness of financial

fraud. Therefore, adding another question on the basic high-return-high-risk theorem may

make the financial literacy test more valuable in predicting the extent to which investors can

protect themselves from financial fraud. Based on our model, the awareness of high-return-

high-risk theorem can also indicate the overall potential of financial fraud in an economy.

For instance, we may consider adding the following question.

Suppose you had invested in two financial products: one with 8% annual return and theother with 20% annual return. Which product is more likely to be repaid to you?A. The 8%-annual-return one;B. Equally likely;C. The 20%-annual-return one;D. Do not know;E. Refuse to answer.

While we take a first step in the direction of designing a question to elicit respondents’

awareness of the basic high-return-high-risk theorem as our modest aim here, we leave a

more elaborate design of the question for future work on this.

5 Concluding Remarks

Widespread of financial fraud has emerged as a pressing problem in China and other de-

veloping countries. Investors are attracted by high return but some of them may not have

a proper awareness of the underlying high risk and possibility of financial fraud. In this

paper, we experimentally investigate the role of a financial education program in investors’

investment decisions. We find that, by receiving a simple education flyer showing the high-

return-high-risk theorem, a significant number of investors will not purchase a fraudulent

financial product. This effect is especially strong for risk-averse investors.

Our findings have important policy implications. Based on our model of boundedly ra-

tional investors, reducing the proportion of unaware investors not only directly helps these

investors, but also attenuates the incentive to commit financial fraud. Therefore, educating

investors to be aware that high return always come with high risk is essential for developing

a healthy financial market. From our experiments and surveys, we find that risk-averse in-

vestors are most sensitive to the education program, which provides the possibility of targeted

education. Based on existing policy practices, policy makers can require investors to com-

plete a mandatory survey to identify vulnerable investors and thereby provide personalized

education. Moreover, we find that the current standard financial literacy test cannot reflect

21

investors’ likelihood of being exploited by financial fraud. We suggest adding a question

regarding investors’ knowledge of the high-return-high-risk theorem.

22

Figure 5: Correct Answers to All Three Questions (by Age)

Figure 6: Correct Answers to All Three Questions (by Gender)

Figure 7: Correct Answers to All Three Questions (by Schooling)

23

Appendix A: English Translation of Experimental In-

struction of Group Y1

Survey Introduction

Greetings! We are researchers from the Qianhai Institute for Innovative Research and are

currently conducting a survey with the Hong Kong University of Science and Technology on

the purchasing behavior of financial products. We are not employees of government or other

for-profit organizations. All your answers will be used only for academic research and will

remain confidential unless you authorize otherwise.

If you agree to participate in this survey, we will give you a small gift. Moreover, one of

ten participants will randomly be selected as lucky participants and will be rewarded based

on their choices in this survey. RMB is used as the monetary unit throughout this survey.

Please do not talk to other participants and please keep your mobile phones, computers,

and other electronic devices off during the survey. You can raise your hand if you have any

questions regarding the survey.

Thank you very much for your cooperation!

24

Before choosing your answers to the questions in the survey, please review the following

pictures very carefully. You can raise your hand if you have any questions regarding the

pictures.

Table 13: A Summary of Return/Risk for Different Financial Products

Annual Rate of Return: 3%Access: Fixed depositRisk: Almost zeroAnnual Rate of Return: 5%Access: Short-term financial products sold by banksRisk: Almost zeroAnnual Rate of Return: 6%Access: Long-term (6 months or longer) financial products sold by banksRisk: Almost zeroAnnual Rate of Return: 7%Access: Fixed return funds; bonds of listed companiesRisk: Bankruptcy of mutual fund companies or listed companies (never happened be-fore)Annual Rate of Return: 8-9%Access: Financial products or P2P products sold by famous companiesRisk: Bankruptcy of companies such as Pingan, Baidu, Alibaba, Sina, or Netease, andno re-organization is performed

Annual Rate of Return: 10-11%Access: Trust; first tier P2P productsRisk: Breach of rigid payment; the P2P company goes bankrupt, and no re-organizationis performedAnnual Rate of Return: 12%Access: Private loans between friendsRisk: Your friend has no money to repay debtsAnnual Rate of Return: 13-15%Access: Second tier P2P productsRisk: The P2P company goes bankrupt, and no re-organization is performed

25

Annual Rate of Return: 15-20%Access: Third tier P2P products; Applying for IPOsRisk: The P2P company goes bankrupt, and no re-organization is performed; unex-pected fall of stock pricesAnnual Rate of Return: 20%Access: Fourth tier P2P products; junk bondsRisk: The P2P company goes bankrupt, and no re-organization is performed; the issuerof junk bonds usually faces financial difficulties and needs a helping hand from thegovernment

Annual Rate of Return: 24%Access: Private loans between friendsRisk: The borrower runs awayAnnual Rate of Return: 25%Access: Fifth tier P2P productsRisk: You are simply unluckyAnnual Rate of Return: 30%Access: UsuryRisk: Hard to get your money back unless you are a gang memberAnnual Rate of Return: 40%Access: Predatory lendingRisk: Hard to get your money back even if you are a gang member

Generally speaking, financial products with high return always have high risk.

26

Part 1

1. Suppose you had 100 yuan in a savings account and the interest rate was 2 percent per

year. After 5 years, you would have in the account if you left the money grow.

A. More than 102 yuan;

B. Exactly 102 yuan;

C. Less than 102 yuan;

D. Do not know;

E. Refuse to answer.

2. Imagine that the interest rate on your savings account was 1 percent per year and inflation

was 2 percent per year. After 1 year, you would be able to buy today with the money in

this account.

A. More than;

B. Exactly the same as;

C. Less than;

D. Do not know;

E. Refuse to answer.

3. “Buying a single company stock usually provides a safer return than a stock mutual

fund.” You think the above statement is

A. True;

B. False;

C. Do not know;

D. Refuse to answer.

4. On a scale from 1 to 7, where 1 means “very low” and 7 means “very high”, how do you

assess your overall financial knowledge?

5. “Now there is a newly innovated financial product that could give you a 20% annual return

at most.27 The highest return for this product is 70 times of the demand deposit interest

rate and 15 times of the one-year deposit interest rate. Additionally, it can be withdrawn

any time after the second day of your investment and no transaction fees are charged for

this product.”

The description above is about a financial product. Based on the description, which of

the options you are most likely to choose?28

A. Invest 2,600 yuan or more in this product;

B. Invest more than 1,300 yuan but no more than 2,600 yuan in this product;

C. Invest more than 180 yuan but no more than 1,300 yuan in this product;

27In groups N0 and Y0, the annual return of the product provided in this question is 8%.28The six choices offered here come from the product descriptions designed by the Fanya Metal Exchange.

27

D. Invest more than 60 yuan but no more than 180 yuan in this product;

E. Invest no more than 60 yuan in this product;

F. Do not invest in this product.

28

Part 2

1. Your year of birth

2. You are

A. Male;

B. Female.

3. You are

A. Single;

B. Married;

C. Divorced or separated;

D. Widowed.

4. Your education level is

A. No formal schooling;

B. Preliminary school;

C. Junior high school;

D. Senior high school or specialized secondary schools;

E. Undergraduate or poly technique college;

F. Graduate school or above.

5. You have an annual income level

A. No more than 20,000 RMB;

B. Higher than 20,000 but no more than 50,000;

C. Higher than 50,000 but no more than 100,000;

D. Higher than 100,000 but no more than 500,000;

E. Higher than 500,000.

6. Have you ever had experience with investing in financial products before?

A. Never (skip questions 7 and 8);

B. Yes.

7. What kinds of financial products have you ever invested in? (multiple choices)

A. Financial products sold by banks;

B. Government bonds;

C. Corporate bonds;

D. Money market funds;

E. Public funds;

F. Private equity funds;

G. Investment trusts;

H. Stocks;

I. Commercial insurance;

29

J. Others, please specify .

8. From where have you obtained information when making financial investment decisions?

(multiple choices)

A. Friends/family members;

B. Newspapers/books/TV;

C. Clerks or advisors in banks/funds/securities company;

D. Internet;

E. Others, please specify .

9. Have you ever heard of or invested in the recent financial scams such as “Fanya” or

“e-zubao”?

A. Never heard of or invested in any of them;

B. Never invested in any of them, but have a basic knowledge of some of them;

C. Never invested in any of them, but have a detailed understanding of some of them;

D. Have invested in some of them, but did not lose any money.

E. Have invested in some of them and lost money.

30

Part 3

Now you are invited to play a game. Flip a fair coin, if it is numbers, you get 100 yuan;

otherwise, you get nothing.

1. If we give you 5 yuan, will you agree to give up playing this game?

A. Agree B. Disagree

2. If we give you 15 yuan, will you agree to give up playing this game?

A. Agree B. Disagree

3. If we give you 25 yuan, will you agree to give up playing this game?

A. Agree B. Disagree

4. If we give you 35 yuan, will you agree to give up playing this game?

A. Agree B. Disagree

5. If we give you 45 yuan, will you agree to give up playing this game?

A. Agree B. Disagree

6. If we give you 55 yuan, will you agree to give up playing this game?

A. Agree B. Disagree

7. If we give you 65 yuan, will you agree to give up playing this game?

A. Agree B. Disagree

8. If we give you 75 yuan, will you agree to give up playing this game?

A. Agree B. Disagree

9. If we give you 85 yuan, will you agree to give up playing this game?

A. Agree B. Disagree

10. If we give you 95 yuan, will you agree to give up playing this game?

A. Agree B. Disagree

Now we will randomly pick one lucky participant of ten participants and randomly choose

one out of ten questions listed above. For example, if you are selected as a lucky participant,

and the 3rd question is chosen, and your answer to this question is “Agree”, you will be

rewarded 25 yuan immediately; if your answer is “Disagree”, we will flip a coin, and if it is

numbers, you get 100 yuan, otherwise you get nothing.

Our survey ends here. Thank you again for your participation!

31

Appendix B: Original Education in Chinese29

29Our education program is designed in accordance with the slides available at http://mp.weixin.

qq.com/s?__biz=MjM5OTYwOTgxOA==&mid=203894079&idx=1&sn=0d758195d4a7f395d0513f93c6ad73e1&

scene=1&srcid=0710LOzz68JjvJ8fWTRz7asQ#wechat_redirect, with minor modifications.

32

References

[1] Albert, S. M., and J. Duffy (2012). “Differences in Risk Aversion Between Young and

Older Adults”, Neuroscience and Neuroeconomics, 2012(1).

[2] Andersen, S., G. W. Harrison, M. I. Lau, and E. E. Rutstrom (2006). “Elicitation Using

Multiple Price List Formats”, Experimental Economics, 9(4), 383-405.

[3] Auster, S. (2013). “Asymmetric Awareness and Moral Hazard”, Games and Economic

Behavior, 82, 503-521.

[4] Campbell, J. Y., H. E. Jackson, B. C. Madrian, and P. Tufano (2011). “Consumer

Financial Protection”, Journal of Economic Perspectives, 25 (1), 91-114.

[5] Croson, R., and U. Gneezy (2009). “Gender Differences in Preferences”, Journal of

Economic Literature, 47(2), 448-474.

[6] Chung, K. S., and L. Fortnow (2016). “Loopholes”, Economic Journal, 126(595), 1774-

1797.

[7] Dessi, R. and X. Zhao (2015). “Overconfidence, Stability and Investments”, Toulouse

School of Economics Working Paper, n. 15-580.

[8] Fernandes, D., J. G. Lynch Jr., and R. G. Netemeyer (2014). “Financial Literacy, Fi-

nancial Education, and Downstream Financial Behaviors”, Management Science, 60(8),

1861-1883.

[9] Gabaix, X., and D. Laibson (2006). “Shrouded Attributes, Consumer Myopia, and

Information Suppression in Competitive Markets”, Quarterly Journal of Economics,

121(2), 505-540.

[10] Galanis, S. (2013). “Unawareness of Theorems”, Economic Theory, 52(1), 41-73.

[11] Heidhues, P., B. Koszegi, and T. Murooka (2016). “Inferior Products and Profitable

Deception”, Review of Economic Studies, 84(1), 323-356.

[12] Heine, S. J., D.R. Lehman, H. R. Markus, and S. Kitayama (1999). “Is There a Universal

Need for Positive Self-Regard?”Psychological Review, 106, 766-794.

[13] Holt, C. A., and S. K. Laury (2002). “Risk Aversion and Incentive Effects”, American

Economic Review, 92(5), 1644-1655.

33

[14] Hsu, J. W. (2016). “Aging and Strategic Learning: The Impact of Spousal Incentives

on Financial Literacy”, Journal of Human Resources, 51(4), 1036-1067.

[15] Inderst, R., and M. Ottaviani (2009). “Misselling through Agents”, American Economic

Review, 99(3), 883-908.

[16] Inderst, R., and M. Ottaviani (2012). “How (Not) to Pay for Advice: A Framework for

Consumer Financial Protection”, Journal of Financial Economics, 105, 393-411.

[17] Inderst, R., and M. Ottaviani (2013). “Sales Talk, Cancellation Terms and the Role of

Consumer Protection”, Review of Economic Studies, 80(3), 1002-1026.

[18] Kawaguchi, K., K. Uetake, and Y. Watanabe (2016). “Identifying Consumer Attention:

A Product-Availability Approach”, mimeo, University of Tokyo.

[19] Li, S., M. Peitz, and X. Zhao (2016). “Information Disclosure and Consumer Aware-

ness”, Journal of Economic Behavior & Organization, 128, 209-230.

[20] Lusardi, A. and O. S. Mitchell (2014). “The Economic Importance of Financial Literacy:

Theory and Evidence”, Journal of Economic Literature, 52(1), 5-44.

[21] Lusardi, A., P. Michaud, and O. S. Mitchell (2017). “Optimal Financial Knowledge and

Wealth Inequality”, Journal of Political Economy, 125(2), 431-477.

[22] Schmitt, D. and J. Allik (2005). “Simultaneous Administration of the Rosenberg Self-

Esteem Scale in 53 Nations: Exploring the Universal and Culture-Specific Features of

Global Self-Esteem”, Journal of Personality and Social Psychology, 89(4), 623–642.

[23] Song, C. (2015). “Financial Illiteracy and Pension Contributions: A Field Experiment

on Compound Interest in China”, mimeo, National University of Singapore.

[24] Spiegler, R. (2011). Bounded Rationality and Industrial Organization, Oxford University

Press.

[25] Tirole, J. (2009). “Cognition and Incomplete Contracts”, American Economic Review,

99(1), 265-294.

[26] Tirole, J. (2014). “Cognitive Games and Cognitive Traps”, mimeo, Toulouse School of

Economics.

[27] von Thadden, E. L. and X. Zhao (2012). “Incentives for Unaware Agents”, Review of

Economic Studies, 79(3), 1151-1174.

34

[28] Zhao, X. (2015). “Strategic Mis-selling and Pre-purchase Deliberation”, mimeo, Hong

Kong University of Science and Technology.

[29] Zhou, J. (2007). “Advertising, Misperceived Preferences, and Product Design”, mimeo,

NYU Stern School of Business.

35