How Do Investment Ideas Spread through Social Interaction? Evidence from a Ponzi Scheme

• DOI: http://doi.org/10.1111/jofi.12822
• Author: VILLE RANTALA
• Published date: 01 October 2019
• Date: 01 October 2019

THE JOURNAL OF FINANCE •VOL. LXXIV, NO. 5 •OCTOBER 2019

How Do Investment Ideas Spread through Social

Interaction? Evidence from a Ponzi Scheme

VILLE RANTALA∗

ABSTRACT

A unique data set from a large Ponzi scheme allows me to study word-of-mouth

diffusion of investment information. Investors could join the scheme only by invitation

from an existing member, which allows me to observe how the idea spreads from one

person to the next based on inviter-invitee relationships. I find that the observed

social network has a scale-free connectivity structure, which significantly facilitates

the diffusion of the investment idea and contributes to the growth and survival of

the socially spreading Ponzi scheme. I further find that investors invest more if their

inviter has comparatively higher age, education, and income.

SHILLER (2000,2014)ARGUES THAT INVESTMENT ideas can spread like epidemics

among investors, with asset prices influenced by social dynamics. In this paper,

I examine whether large-scale social contagion effects among investors exist,

and if so, what the information diffusion process looks like.

To do so, I employ a data set in which I directly observe the diffusion of an

investment idea from one person to the next. The data set consists of partici-

pants of a large investment Ponzi scheme, whereby investors could join only by

personal invitation from an existing member, who was referred to as a sponsor.

This feature allows me to study the effect of word-of-mouth information at an

individual level. Information about the scheme was not publicly available, so

∗Ville Rantala is with the Miami Business School, University of Miami. I am grateful to the

Finnish National Bureau of Investigations for providing the source documents. Parts of this paper

were previously circulated as a separate working paper entitled “Keeping Up with the Ponzis.” I

thank Nick Barberis, Peter Bossaerts, James Choi, Nicholas Christakis, Henrik Cronqvist, Florian

Ederer, Ester Faia, David Hirshleifer,Harrison Hong, Hans Hvide, Markku Kaustia, Matti Kelo-

harju, Samuli Kn¨

upfer, Timo Korkeam¨

aki, George Korniotis, Alok Kumar, Juhani Linnainmaa,

Steffen Meyer, Raghu Rau, Sophie Shive, Scott Weisbenner, Charlotte Østergaard, participants at

the Helsinki Finance Summit 2014, European Finance Association Doctoral Tutorial 2015, Euro-

pean Conference on Household Finance 2015, Conference on Behavioral Aspects of Macroeconomics

and Finance 2015, Miami Behavioral Finance Conference 2015, SFS Cavalcade 2016, American

Economic Association 2017, European Retail Investment Conference 2017; as well as seminar par-

ticipants at BI Norwegian Business School, Cass Business School, Copenhagen Business School,

Georgetown University, HEC Paris, Indiana University, University of Melbourne, University of

Miami, University of New South Wales, Washington University in St. Louis, and Yale University

for helpful comments. I also thank Antti Lehtinen for assistance with geographic information anal-

ysis. I acknowledge financial support from the OP Group Research Foundation. The author does

not have any potential conflicts of interest to disclose, as identified in the Journal of Finance’s

disclosure policy.

DOI: 10.1111/jofi.12822

2349

2350 The Journal of Finance R

when a new investor joins, I know that he has learned about the opportunity

from the inviter. By observing the inviter-invitee social network, I am able

to examine the process through which an investment idea spreads across a

population.

The investment scheme, Wincapita, was a Finnish investment operation that

was active from 2003 to 2008. Wincapita offered its investors large returns, ini-

tially claiming that the profits were generated by sports betting and later by

currency trading. In reality,it was a classic Ponzi scheme whereby all incoming

cash flows came from new and existing investors, with none of the profits gener-

ated by actual trading or investments. The scheme grew very large, ultimately

growing to over 10,000 members, approximately 0.2% of the total population of

Finland.

The collapse of Wincapita in 2008 led to one of the largest criminal inves-

tigations in Finnish history.1The data used in this study come from the in-

vestigation documents of the Finnish National Bureau of Investigation.2The

documents allow me to identify over 5,000 Wincapita investors and contain

detailed information for over 3,000 investors who were questioned by police. In

addition to details about their investments and withdrawals, these data contain

information on investors’ personal characteristics, such as age, income, loca-

tion, and education. I combine these data with information on their sponsoring

relationships.3Because the data were collected from investors in a formal police

interview, they do not suffer from many of the typical reporting and selection

biases that exist in survey data on social relationships. The interviewing offi-

cer’s responsibility was to collect facts that could be used as evidence in a court

of law.

Several papers provide theoretical models of information diffusion and the

structure of information networks among investors.4I present empirical evi-

dence on these phenomena. In particular, the data allow me to shed light on

the social network structure of information diffusion among investors. Empir-

ical social networks that connect people through different kinds of personal

ties exhibit strong structural regularities and differ significantly from random

graph networks in which the connections between nodes are evenly distributed

1Reported by YLE News on May 5, 2008 (“Wincapitasta tulossa maan laajimpia rikostutkin-

toja.”)

2Finnish National Bureau of Investigation, 2010, Esitutkintap¨

oyt¨

akirja 2400/R/61/10.

3There was a financial incentive to sponsor others. Sponsors received €200 of (virtual) money

for each sponsored investor and 20% of the virtual profits earned by the sponsored investor’s

investments (I discuss this further in Section Ibelow). A sponsor’s sponsor would not receive any

portion of these profits, so Wincapita was not a traditional pyramid scheme whereby the profits

are determined by the investor’s level in the “pyramid.”

4Stein (2008), Han and Yang (2013), and Andrei and Cujean (2017) model the transmission

of information through personal communication between investors. Ozsoylev and Walden (2011)

study the asset pricing implications of large information networks. Shiller and Pound (1989),

Shiller (2000), and Shive (2010) argue that epidemic models can be used to characterize the

diffusion of social interest among investors. For general models of word-of-mouth communication,

see Ellison and Fudenberg (1995), Banerjee and Fudenberg (2004), and Cao, Han and Hirshleifer

(2011).

How Do Investment Ideas Spread through Social Interaction? 2351

(see Jackson and Rogers (2007) for a review). However, because the diffusion

of word-of-mouth information is typically unobservable, there is little evidence

on whether these structures play a role in the diffusion of information within

the networks. In social networks, the distribution of connections per node

typically has a heavy right tail, and the average node to node distances are

short. I find that both of these characteristics exist in the diffusion process of

Wincapita.

In particular, the distribution of the number of connections per node in Win-

capita’s sponsoring network decays approximately as a power law. The empir-

ical probability of sponsoring kinvestors is proportional to the power of k,so

that P(k)∼k−γwhere γis a constant. The power-law characteristic is visu-

ally apparent in a log-log plot, and Kolmogorov-Smirnov test statistics based

on a fitted power-law model strongly support the view that the data follow a

power law. Networks with this structure are known as scale-free networks,

and power-law distributions are very common in empirical social networks

(Barab´

asi (2009)). Here, the power law indicates that a small minority of in-

vestors are responsible for most of the social effect.

I compare Wincapita’s actual sponsoring network to a simulated random

network and find that the power-law topology has a dramatic effect on the rate

at which information spreads. The random network has the same number of

investors, the same percentage of sponsors, and the same average number of

sponsored investors per sponsor. The only difference compared to the actual

network is that the distribution of the number of sponsored investors follows

a Poisson distribution instead of the power law, as in the model of Erd˝

os and

Renyi (1959). The results show that an epidemic that spreads through the

Wincapita network one step at a time reaches all investors in 15 steps. In

comparison, in the simulated network, it takes an average of 161 network steps

to reach the same number of investors. When I calibrate a simple Ponzi scheme

model to both networks, I find that the actual network can sustain significantly

higher payouts relative to investors’ investments without collapsing.

The network structure also provides information about the diffusion dynam-

ics of Wincapita. I find that the cumulative number of investors as a function

of network distance from the originator of the scheme follows an S-shaped

curve. The S-curve implies that information diffusion within social networks

progresses in a nonlinear fashion.

I next analyze how social connections within the scheme are related to par-

ticipants’ investment decisions. After controlling for personal characteristics

and inviter fixed effects, I find that investors invest statistically significantly

more if their inviter has comparatively higher income, age, and education.

The marginal effect of the inviter’s income is highest when it is just above

the invitee’s income, suggesting that it may be a reference point in decision

making. I also find that sponsors invest more than nonsponsors, which in-

dicates that social behavior and investment behavior are correlated among

investors.

This paper contributes to the literature in several ways. First, I show that an

investment idea can spread across the population through social interaction,