ArticlePDF Available

Abstract and Figures

By the time the Colombian government closed DMG and DRFE, two Ponzi schemes that were operating in Colombia until 2008, over half a million customers had deposited funds corresponding to 1.2% of Colombia’s annual GDP. We show that the individuals who invested in DMG and DRFE obtained close to 40% more loans in the formal financial sector prior to the government closing these firms, compared to similar individuals who did not invest in these pyramids. Moreover, deposits in the formal financial sector fell in those municipalities affected by these two pyramids: a one-standard deviation increase in the municipal presence of the pyramid schemes reduced municipal saving deposits by 2.9% and Certificate Deposits by close to 10%. After the firms were shut down, the proportion of nonperforming loans of investors rose 35% above non-investors’ loans; two years later, investors’ deposits had not yet fully recovered.
Content may be subject to copyright.
Ponzi Schemes and the Financial
Sector: DMG and DRFE in Colombia
Marc Hofstetter
Daniel Mejía
José Nicolás Rosas
Miguel Urrutia
Documentos
CEDE
ISSN 1657-7191 Edición electrónica.
No.35
MAYO DE 2017
Serie Documentos Cede, 2017-35
ISSN 1657-7191 Edición electrónica.
Mayo de 2017
© 2017, Universidad de los Andes, Facultad de Economía,
CEDE. Calle 19A No. 1 – 37 Este, Bloque W.
Bogotá, D. C., Colombia Teléfonos: 3394949- 3394999,
extensiones 2400, 2049, 3233
infocede@uniandes.edu.co
http://economia.uniandes.edu.co
Impreso en Colombia – Printed in Colombia
La serie de Documentos de Trabajo CEDE se circula con
propósitos de discusión y divulgación. Los artículos no
han sido evaluados por pares ni sujetos a ningún tipo de
evaluación formal por parte del equipo de trabajo del CEDE.
El contenido de la presente publicación se encuentra
protegido por las normas internacionales y nacionales
vigentes sobre propiedad intelectual, por tanto su
utilización, reproducción, comunicación pública,
transformación, distribución, alquiler, préstamo
público e importación, total o parcial, en todo o en parte, en
formato impreso, digital o en cualquier formato conocido o por
conocer, se encuentran prohibidos, y sólo serán lícitos en
la medida en que se cuente con la autorización previa y
expresa por escrito del autor o titular. Las limitaciones y
excepciones al Derecho de Autor, sólo serán aplicables
en la medida en que se den dentro de los denominados
Usos Honrados (Fair use), estén previa y expresamente
establecidas, no causen un grave e injusticado perjuicio
a los intereses legítimos del autor o titular, y no atenten
contra la normal explotación de la obra.
Universidad de los Andes | Vigilada Mineducación
Reconocimiento como Universidad: Decreto 1297 del 30 de mayo de
1964. Reconocimiento personería jurídica: Resolución 28 del 23 de
febrero de 1949 Minjusticia.
Ponzi Schemes and the Financial Sector:
DMG and DRFE in Colombia*
Marc Hofstetter, Daniel Mejía,
José Nicolás Rosas§ and Miguel Urrutia**
Abstract
By the time the Colombian government closed DMG and DRFE, two
Ponzi schemes that were operating in Colombia until 2008, over half a
million customers had deposited funds corresponding to 1.2% of
Colombia’s annual GDP. We show that the individuals who invested
in DMG and DRFE obtained close to 40% more loans in the formal
financial sector prior to the government closing these firms, compared
to similar individuals who did not invest in these pyramids. Moreover,
deposits in the formal financial sector fell in those municipalities
affected by these two pyramids: a one-standard deviation increase in
the municipal presence of the pyramid schemes reduced municipal
saving deposits by 2.9% and Certificate Deposits by close to 10%. After
the firms were shut down, the proportion of nonperforming loans of
investors rose 35% above non-investors’ loans; two years later,
investors’ deposits had not yet fully recovered.
Keywords: Ponzi Schemes; Pyramids; Colombia; Financial Sector; Savings and
Loans; Loan Ratings.
JEL codes: E 21, E 44, G 11.

*We would like to thank Román David Zárate for his excellent research assistance and the
Superfinanciera for providing us with financial microdata. We would also like to thank the
participants of the Lacea 2016 conference, and participants at the seminars at Universidad de los
Andes and Universidad Nacional of Colombia, for their comments and suggestions. All
remaining errors are ours.
Department of Economics and CEDE, Universidad de los Andes, Bogotá; email:
mahofste@uniandes.edu.co. Homepage: https://economia.uniandes.edu.co/hofstetter.
Secretary of Security in Bogotá; email: dmejia@uniandes.edu.co.
§ The Inter-American Development Bank; email: josero@iadb.org. The opinions expressed in this
work are those of the author and do not necessarily reflect the views of the Inter-American
Development Bank, its Board of Directors, or the countries they represent.
** Department of Economics and CEDE, Universidad de los Andes, Bogotá; email:
murrutia@uniandes.edu.co. Homepage: http://economia.uniandes.edu.co/urrutia
1
Las pirámides y el sector financiero:
DMG y DRFE en Colombia*
Marc Hofstetter1, Daniel Mejía2,
José Nicolás Rosas3 y Miguel Urrutia4
Resumen
Para cuando el gobierno colombiano logró cerrar DMG y DRFE, dos
pirámides que operaron en el país hasta finales de 2008, más de medio
millón de personas habían depositado allí recursos equivalentes a 1.2%
del PIB. Mostramos que los individuos que invirtieron en DMG y
DRFE tenían 40% más créditos con el sector financiero formal justo
antes del cierre de las pirámides, comparados con individuos similares
pero que no invirtieron allí. Adicionalmente, en los municipios con
mayor presencia de las pirámides, los depósitos en el sector financiero
formal cayeron. Un aumento de una desviación estándar en la
presencia municipal de las pirámides redujo los depósitos municipales
en cuentas de ahorros en 2.9% y de los CDTs en casi 10%. Tras la
desaparición de las firmas, la proporción de cartera mala de los
inversionistas se incrementó en 35% en comparación a la de los que no
invirtieron en pirámides.
Palabras Clave: Ponzi; Pirámides; Colombia; Sector financiero; Ahorro y
crédito, Calificación crediticia.
Códigos JEL: E 21, E 44, G11.


*Nuestros agradecimientos a Román David Zárate por su excelente trabajo como asistente de
investigación y a la Superfinanciera por proveernos de los microdatos financieros. Agradecemos
también los comentarios y sugerencias de los participantes de la Lacea 2016, y de seminarios en
la Universidad de los Andes y la Universidad Nacional of Colombia. Los errores son nuestra
responsabilidad.
1 Facultad de Economía y CEDE, Universidad de los Andes, Bogotá; email:
mahofste@uniandes.edu.co. Página web: https://economia.uniandes.edu.co/hofstetter.
2 Secretario de Seguridad de Bogotá; email: dmejia@uniandes.edu.co.
3 Banco Interamericano de Desarrollo; email: josero@iadb.org. Las opiniones expresadas en este
trabajo son las del autor y no necesariamente reflejan la visión del Banco Interamericano de
Desarrollo, su junta directiva o la de los países que representan.
4 Facultad de Economía y CEDE, Universidad de los Andes, Bogotá; email:
murrutia@uniandes.edu.co. Página web: http://economia.uniandes.edu.co/urrutia
2
1. Introduction
As of the end of 2008, the Colombian government had closed two firms, DMG
and DRFE, both of which were accused of running Ponzi schemes. By the time
these firms where put out of business, over half a million people had invested in
them. 533,560 costumers deposited resources equivalent to 1.2% of Colombia’s
GDP, an amount corresponding to 3.9% of total deposits in the financial sector in
2008, or 22% of the total deposits Bancolombia—the largest bank in the country—
reported as of the end of that year. 80% of investors lost their deposits. The
amount of the average deposit reached 93% of annual per capita GDP for 2008.
While a few infamous schemes—like the Madoff fraud in the U.S. or the schemes
in Albania that collapsed in the late-90s, sending that country into chaos and
bringing down its government (Jarvis, 2000)—have drawn the concentrated
attention of the media, Ponzi schemes are more common than is generally
recognized. Deason et al. (2015) described 376 Ponzi schemes prosecuted by the
SEC (Securities & Exchange Commission) between 1988 and 2012 in the U.S. A
study conducted by the Caribbean Policy Research Institute (CaPRI, 2008),
identified 21 schemes operating in Jamaica as of January 2008. Carvajal et al.
(2009) reported that in 2008, over 200 schemes promising returns of up to 300
percent within 6 months were operating in Colombia. In 2017, Germán
Cardona—also known as the Spanish Madoff—was sentenced to 13 years in
prison for leading a pyramid scheme that defrauded more than 180.000 people.
While the scheme was initiated in Spain, it attracted investors from over 100
countries.
The literature on unregulated investment schemes has identified several of their
negative consequences. Carvajal et al. (2009) summarize them in seven points: i.
They divert deposits from banks and increase non-performing loans if loan
proceeds are diverted to these schemes; ii. They divert savings from productive
to unproductive uses and, in some cases, from the domestic economy to foreign
destinations; iii. They cause swings in consumption driven by paper profits or
early withdrawals; iv. They undermine confidence in financial markets; v. They
3
imply fiscal costs if bailouts occur; vi. They cause socio-economic strife if a
sufficiently large number of households are suddenly exposed to losses; and vii.
They undermine the reputation of political authorities, regulators, and law
enforcers on account of their failure to prevent open fraud and to address money
laundering or the support of other illegal enterprises by the schemes’ operators.
More recently, Cortés et al. (2016) found that the breakdown of Ponzi Schemes
also increases shoplifting and robbery in places with weak law enforcement
institutions and lower access to credit.
Despite the long list of grave consequences, the nature of the schemes and the
lack of systematic data has precluded researchers from pinning down the size
and scope of some of these consequences. What we find in the literature on Ponzi
schemes remains, to a large extent, anecdotal.5 Hence, Deason et al. (2015) claim
that the “[e]xtant knowledge of Ponzi schemes in the … literature is mainly
anecdotal”. Similarly, Jarvis (2000), in his study of the infamous Albanian
schemes, acknowledges that when assessing the impact of their collapse on the
economy, the evidence is mostly anecdotal. Finally, and arguably more relevant
to the focus of our paper, Carvajal et al. (2009)—who describe several Ponzi
schemes in the Caribbean, Colombia, the U.S. and Africa—point out that there is
“anecdotal evidence that some of the schemes … diverted deposits and increased
NPLs [nonperforming loans]”.
There is also abundant anecdotal evidence on the impact of DMG and DRFE on
the financial sector. David Murcia—the founder of DMG—repeatedly claimed
that his business was legal and that he was subjected to a conspiracy by the
banking sector because the high yields offered by his company were diverting
deposits from the formal banking sector to his firm.6 In fact, as reported by
Carvajal et al. “commercial banks had expressed concern that their depositors
were withdrawing funds to invest in schemes.” Other anecdotal evidence

5TherearealsosometheoreticalpapersstudyingPonzischemes,e.g.,Artzrouni(2009)andBhattacharaya
(2003).
6Inthisvideoforinstance,heaccusesLuisCarlosSarmiento,themainbankerinColombia,ofconspiring
againsthisfirm:http://www.youtube.com/watch?v=LdWdgjnbFeE

4
suggests that investors in pyramids often obtained loans from the formal
financial sector to invest in these scams.
Using a unique dataset containing the universe of individual investments and
profits or losses in these two schemes, we are able to estimate—for the first time,
as far as we can tell, in the literature studying Ponzi schemes—the impact they
had on the formal banking system. In particular, we are able to pinpoint by how
much deposits fell in the banking system while the schemes were operating, and
how persistent this effect was after the schemes were shut down. Moreover, on
the loans’ side, we estimate by how much the individuals that participated in the
schemes increased their loans within the banking sector prior to the schemes
ending, and the effect this had on non-performing loans both during the life-cycle
of the pyramids and after they were put out of business.
Our findings show that individuals that invested in the Ponzi schemes had (just
before the schemes went out of business) 39% more loans within the formal
financial sector than similar individuals who did not invest in pyramids. We also
find that, prior to DMG and DRFE ending, the proportion of loans within the tiers
with the best ratings was 33% higher for investors in the pyramids: while they
were making money, they maintained better ex-post loan ratings than similar
individuals who did not participate in DMG or DRFE. Later, nevertheless, once
the firms were no longer operating, the proportion of nonperforming loans was
as much as 35% higher for individuals that had invested in the pyramids.
We also provide evidence showing that deposits in the financial sector were
affected by DMG and DRFE. While the Financial Supervisory Agency of
Colombia (Superfinanciera) has no information concerning deposits at the
individual level, it does record data on deposits at the municipal level. By
exploiting the variation across municipalities for the presence of Ponzi schemes
as well as municipal deposits, we are able to estimate the effects of the pyramids
on deposits. We find that a one standard deviation increase in the presence of
pyramids (as defined below) at the municipal level reduced total deposits in the
financial sector by between 2.4% and 2.7%. The same effect is much greater on
5
certificate deposits (CDs): they fell by between 9.2% and 10.2%. Moreover, the
latter effect was long-lived, as two years after the government put the schemes
out of business, CDs had not fully recovered.
The rest of the paper is organized as follows. Section 2 summarizes the story of
the two Ponzi schemes. Section 3 describes the datasets and stylized facts. In
sections 4 and 5, we explore the impact of these schemes on loans and deposits,
respectively. Section 6 concludes.
2. Some background: The Rise and fall of DMG and DRFE7
David Murcia Guzman, the founder of DMG, arrived to La Hormiga, Putumayo,
a small town in the southwest of the country, in 2003. With a high school diploma
and previous experience as a door-to-door salesman in a multi-level marketing
company, he began DMG. A couple of years later, he opened offices in the states
of Meta and Nariño. By the time his empire was put out of business in November
2008, DMG had expanded to 62 (out of 1103) municipalities in Colombia and had
investments in Panama, Venezuela and Ecuador. Before being shut down by the
government, the company had diversified into businesses ranging from freight
companies to television channels.
The modus operandi was a masked Ponzi scheme where high returns were paid
via the deposits of new customers. To avoid drawing the attention of the
authorities for illegally taking deposits without being a financially supervised
firm, it sold “prepaid cards” that promised high yields or granted the right to
buy appliances and other goods at below-market prices in the future.
DRFE (Dinero Rápido, Fácil y Efectivo—Money in Cash, Fast and Easy) also
attracted thousands of customers with similar strategies, although fewer than
DMG, as is reported in the next section. The company was started by a Pastor,
Carlos Alfredo Suarez, in San Juan de Pasto, capital of the state of Nariño, also in

7Thissectionisbasedonnewspaperreports,decreesbygovernmentagencies,andjudiciarysentences.
6
the southwest of Colombia. By the end of 2008, when the authorities finally shut
it down, it had offices in 69 municipalities.
An ostentatious ceremony of DMG’s official launching of a television channel
(called the Body Channel) in 2006—one attended by several celebrities—caught
the attention of the national media and the authorities. Soon after, the
Superfinanciera published a public warning in the newspapers explaining that
DMG was not a financially supervised firm and was not authorized to take
deposits from the public. Several government agencies began tracking the firm’s
activities. The Superfinanciera sent a commission to the south of the country to
investigate DMG’s operations, while the Intelligence and Financial Analysis Unit
(UIAF)—a government agency that tracks money laundering activities—began
investigating DMG’s activities. The latter discovered suspicious transactions at
the "Body Channel" company, and over the ensuing year, two additional visits
by the Superfinanciera ended with a resolution in 2007, ordering DMG to cease
taking deposits from the public and to return the COP $18,145 million (US$ 8
million) invested by nearly 12,000 people. After the ban was imposed by the
Superintendent, David Murcia Guzmán began a public relations campaign,
began financing various political campaigns for mayors and governors in the
states of Putumayo and Nariño, and lobbied Congress to approve a law making
his firm’s activities legal. Further attempts by the Superfinanciera to close the
company received harsh critiques by the House representative of Putumayo,
Orlando Guerra. Another representative of the State, Guillermo Rivera,
encouraged the government to intensify its investigation of the origins of the
firm’s wealth. A mob attacked Rivera’s family business in Putumayo following
his statements.
At the end of 2007, the Superfinanciera ordered the company’s liquidation for the
first time. Murcia and his lawyers appealed in small local courts and succeeded
in reversing the decision. The operations resumed under a new company and
some changes in the modus operandi so as to circumvent the Superfinanciera’s
orders. In February 2008, the DIAN (Colombia’s tax agency) together with the
7
Superintendencia de Sociedades (the supervisor of large corporations) joined in
the investigations against DMG. From June through November 2008, DMG
continued to attract the money of thousands of Colombians and used the media
attention and Murcia Guzman’s extravagant lifestyle—inclusive of a private jet
and a fleet of luxury cars—to promote the firm. Finally, in November 2008 the
Colombian government shut down both DMG and DRFE, using the legal
authority of a decree declaring a State of Social Emergency and extending the
powers of the Supersociedades.
A series of protests and riots followed, with crowds expressing support for David
Murcia Guzman and Carlos Suarez, but this time their luck had run out.
Shareholders and legal representatives received arrest orders, and as
investigations progressed, it turned out that several politicians including two
governors and several congressmen, journalists and officials had been involved
in the schemes. Additionally, the authorities also investigated the links these
schemes had with guerrillas, paramilitaries and drug trafficking organizations.
After all, DMG was born during the years of the coca boom in Putumayo.
Prison sentences were awarded to the leaders of DMG and DRFE: David Murcia
Guzman, arrested in Panama City in 2009 and extradited to the U.S., was given a
nine year sentence in the U.S. and has a pending 22 year long one in Colombia.
Several other accomplices were also captured and sentenced in Colombia or
extradited to the U.S. In 2011, Carlos Alfredo Suárez, head of DRFE, was
sentenced to seven years in jail and ordered to pay a heavy fine.
Regarding the reimbursing of funds to investors, the government appointed legal
auditors to liquidate the assets seized from DMG and DRFE. Reimbursement of
the money from DMG began in July 2009, with each investor receiving COP
$275,000 (US$118)—a fraction of their actual investments. In 2012, the US
government returned USD $2'183,000 from Murcia´s confiscated bank accounts
to be distributed among DMG’s victims. The reimbursement of the money from
DRFE began in 2009, with each investor receiving $350,000 (US$ 150)—again a
fraction of their investments.
8
Since 2011, the Superfinanciera has given advice to citizens regarding the
appearance of at least 100 new schemes similar to DMG and DRFE across the
country. In September 2016, the Superfinanciera was investigating flyers
announcing the return of DMG.
3. Data and stylized facts
The main database contains a list of DMG and DRFE’s costumers, along with
their investments in the firms and the profits or losses they made. The
information does not, however, tell us anything about the timing of the
investments: we do not know when costumers deposited money or when they
received the proceeds, if any, of their investments. We only know the final
balance at the moment the government shut the two firms down as of the end of
2008, along with total investments. Additionally, the dataset does not provide
any information regarding the characteristics of the investors beyond their
identification.
In Table 1, we describe the main stylized facts from this dataset. Over half a
million people participated in the pyramids and 80% lost some or all of the
money they invested. While the mean investment was over US$4,600 (close to the
per capita annual GDP of 2008) there is great variation across investors: investors
in the 10th percentile invested, on average, just over US$400, while those in the
90th percentile invested more than US$12,000. The average loss was US$2,570,
while the average net profit (of the winners) reached US$3,417. These figures are
reported in dollars, using the average exchange rate in November 2008, when the
government seized the two businesses.
9
Table 1. Descriptive statistics, DMG and DRFE’s costumers. $ corresponds to figures
in US$ dollars converted from Colombian pesos at the exchange rate for November
2008.
Investors: who and where
The dataset just described provides no information on the characteristics of the
investors. To gain some insight into who the investors were and their geographic
locations, we match the DMG/DRFE dataset with the SISBEN survey run by the
Colombian government to gather information on individuals’ socioeconomic
characteristics. We use the second wave of SISBEN, conducted between 2003 and
2007, which collected information on 32.5 million individuals nationwide. (The
total population in 2008 was 44.5 million.) This allows us to obtain socioeconomic
characteristics of investors prior to the end of the pyramids.
The government uses the SISBEN survey to target the recipients of many of its
social programs. Thus, for the most part, it does not include individuals from the
most affluent portions of the population. Using the IDs in both datasets, we were
DMG DRFE Both Total
Number of investors 356.631 153.878 23.051 533.560
% of losers 0,79 0,83 0,78 0,80
Total $ 1.191.261.625 $ 865.592.979 $ 340.238.032 $ 2.395.378.591
Mean $ 3.559 $ 5.656 $ 14.741 $ 4.671
Median $ 1.714 $ 3.514 $ 10.284 $ 2.143
10th percentile $ 321 $ 557 $ 2.485 $ 429
90th percentile $ 9.170 $ 13.155 $ 32.096 $ 12.213
Total $ 192.401.608 $ 72.418.423 $ 35.052.231 $ 299.957.963
Mean $ 3.356 $ 2.825 $ 6.813 $ 3.417
Median $ 1.276 $ 1.500 $ 3.447 $ 1.474
10th percentile $ 93 $ 159 $ 433 $ 122
90th percentile $ 9.213 $ 6.042 $ 15.341 $ 8.399
Total $ 677.047.974 $ 312.813.304 $ 105.413.798 $ 1.096.989.122
Mean $ 2.414 $ 2.449 $ 5.871 $ 2.570
Median $ 1.071 $ 1.713 $ 4.028 $ 1.286
10th percentile $ 96 $ 381 $ 888 $ 150
90th percentile $ 6.256 $ 5.228 $ 13.027 $ 6.299
Losers' losses
Winner's profits
Deposits
10
able to match 51% (269,855 individuals) of all investors in DMG and DRFE with
individuals in the SISBEN. Beyond the fact that the SISBEN is not a Census, typos
in the IDs in either the SISBEN or the pyramids explain why the merged sample
is not larger. As we emphasize in sections 4 and 5, the fact that the matched
sample does not include individuals in the higher brackets of the income
distribution, along with the observation that we did not find a match for roughly
half the investors, imply that our estimates on the impact of the pyramids on
loans and deposits should be interpreted as constituting the lower bound of their
actual impact. There are two reasons for this: first, individuals in the highest
brackets of the income distribution have greater access to financial services than
those in the lowest brackets. For instance, according to the 2010 ELCA survey in
Colombia, more than 80% of individuals with outstanding loans and belonging
to the highest quintile of the income distribution obtained their respective loans
from the banking sector. This proportion is less than 40% for individuals in the
lowest quintile. On the other hand, once we turn to econometric estimations,
some individuals who invested in pyramids but who we were unable to match
to the SISBEN due to typos in the IDs, could end up in our control groups, thus
attenuating the effects.
In Table 2, we report the descriptive statistics of the matched individuals. The
sample is almost evenly split between males and females, with the latter holding
a slight advantage. More than a third of investors were over 44 years old and less
that 8% were under the age of 25. Almost 13% had an education beyond high
school and 40% had at least a high school diploma. This figure, to put it in
perspective, is similar to that at the national level in 2008: 38.5% of the population
of 25+ years had attained an upper secondary educational level. This supports
the idea that, as in other pyramid cases (Madoff, Spain, Albania) the costumers
were not necessarily uneducated or financially illiterate.
Moreover, 56% were married or living with a partner. Investors reported an
average income of $82 per month, with a large standard deviation ($443). Recall,
11
however, that the SISBEN does not survey individuals in the highest deciles of
the income distribution, so this figure should be read as a lower bound.
The two pyramids had costumers with different profiles. Those in DRFE were
much poorer—their average income was less than half of those in DMG. They
were also less educated, with only about 5% of them having studied beyond high
school, a statistic that is almost 17% for customers of DMG. Those in DRFE were
also part of larger households and had lower SISBEN scores (which are used by
the government to prioritize subsidies).
Table 2. Socioeconomic characteristics of matched investors. $ corresponds to US$
dollars (at average exchange rate as of November 2008)
What about their location? We proxy for location by assigning to each investor
the location reported in the SISBEN survey. Then, for each municipality, we
Invested in
Variables Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev.
Demographics&Income, investors
Male 47,8% 0,50 51,2% 0,50 46,5% 0,50 48,5% 0,50
Income (montlhy) $ 82 $ 443 $ 43 $ 102 $ 104 $ 542 $ 50 $ 108
Age<18 5,1% 0,22 8,0% 0,27 3,9% 0,19 8,9% 0,28
Age 18-24 2,5% 0,16 1,8% 0,13 1,4% 0,12 2,2% 0,15
Age 25-34 26,8% 0,44 27,9% 0,45 27,0% 0,44 22,6% 0,42
Age 35-44 28,3% 0,45 28,5% 0,45 29,3% 0,46 30,0% 0,46
Age>44 36,0% 0,48 33,8% 0,47 38,3% 0,49 36,1% 0,48
Educa tion, inv estors
No education 5,3% 0,22 8,1% 0,27 4,0% 0,20 9,3% 0,29
Incomplete elementary 19,7% 0,40 28,1% 0,45 15,9% 0,37 28,3% 0,45
Complete elementary 18,9% 0,39 22,8% 0,42 17,1% 0,38 22,4% 0,42
Incomplete high school 16,2% 0,37 13,6% 0,34 16,6% 0,37 13,1% 0,34
Complete high school 27,0% 0,44 21,9% 0,41 29,5% 0,46 21,4% 0,41
More than complete high school 12,8% 0,33 5,5% 0,23 16,8% 0,37 5,6% 0,23
Marital status, investors
Cohabitation 23,7% 0,43 22,1% 0,42 24,8% 0,43 24,9% 0,43
Married 32,1% 0,47 31,9% 0,47 33,3% 0,47 32,8% 0,47
Widowed 2,5% 0,15 2,1% 0,14 2,7% 0,16 2,6% 0,16
Single/Divorced 41,7% 0,49 43,9% 0,50 39,1% 0,49 39,7% 0,49
Household variables
Household size 3,8 1,74 4,0 1,89 3,7 1,66 4,0 1,87
Kids' proportion 13,4% 0,18 15,0% 0,19 12,6% 0,18 15,0% 0,19
Household head's years of education 3,6 1,7 3,1 1,6 3,9 1,8 3,2 1,6
Household head's earnings (monthly) $ 126 $ 182 $ 74 $ 123 $ 151 $ 211 $ 86 $ 126
Household's per capita income (monthly) $ 58 $ 372 $ 29 $ 52 $ 71 $ 527 $ 34 $ 60
Sisben score 18,9 11,4 14,0 9,0 21,2 12,5 14,3 9,3
Ponzi schemes
Deposits $ 4.206 $ 6.428 $ 5.271 $ 6.471 $ 3.168 $ 4.928 $ 13.841 $ 14.098
Observations
DMG or DRFE
269.885 11.668181.36076.827
DRFE DMG DMG and DRFE
12
calculate the ratio of investors to the population with 14+ years.8 Of course, the
ratios are underestimated given that we did not find a match for 49% of investors.
We report the distribution of per capita investors in Map 1, along with markers
indicating municipalities where the pyramids had at least one office.
Reassuringly, the municipalities that the anecdotal evidence suggests should be
heavily affected by the pyramids show up as such on our maps.
Both pyramids were particularly strong in the southwest of the country, in
municipalities belonging to the states of Nariño and Putumayo where they were
born. Out of the top 100 municipalities in terms of per capita investors, 55 belong
to these two states. Cundinamarca—the state in the middle of the country, where
Bogota is located—was also hit hard, with 21 of its municipalities making it into
the top 100. The five municipalities with large numbers of per capita (14+)
investors—all in Nariño and Putumayo—have figures between 18% and 21%.9
Considering that our count of investors (that is, as matched with SISBEN)
underestimates the actual figures, and that conceivably only one individual per
household invested resources in the pyramids, these figures imply that virtually
all households in these municipalities invested in DMG or DRFE.
Only municipalities in the states of Amazonas, Guainia and Vaupes report zero
investors. While the pyramids’ stronghold was in the southwest, they also
prospered in other parts of the country. For instance, in Bogotá, our matching
strategy identifies close to 70,000 investors in the pyramids, the vast majority of
them in DMG. As a matter of fact, out of the individuals investing in DMG whose
origin we were able to proxy, close to a third were from Bogota. Other smaller
municipalities like Suesca, Sopó and Tocancipá—all in Cundinamarca and near
Bogotá—made it into the top twenty, with large numbers of per capita investors
corresponding to figures above 10%.

8Thepopulation14+foreachmunicipalityistakenfromtheDANE,Colombia’sstatisticalagency.
9ThetopfivemunicipalitiesareValledelGuamuez,Nariño,Orito,Mocoa,andSibundoy.
13
Map 1: Investors per capita (14+). DMG + DRFE.
We also use the location of investors and information on their investments and
losses or profits to further explore the size and scope of the pyramids. In
particular, we calculate the losses and revenues in each municipality by adding
all resources invested by individuals matched to the municipality and
subtracting the sum of the resources they received back from the pyramids. We
express the outcome as a percentage of annual municipal public expenditures.10

10AnnualmunicipalpublicexpendituresaretakenfromDNP—NationalPlanningDepartment.
#
#
#
#
#
#
#
#
#
###
#
##
#
#
##
#
#
#
#
#
#
#
#
#
#
#
#
##
##
#
##
#
###
#
##
##
#
#
##
##
#
###
#
#
#
##
##
###
#
#
#
#
#
##
#
##
#
#
No Data
Investors pc 14 0.4%
0.4% < Investor s pc 14 1%
1% < Investors pc 14 2.5%
2.5% < Investor s pc 14 5%
5% < Investors pc 14 10%
10% < Investors pc 14 15%
Investors pc 14 > 15%
DMG or DRFE office
14
The list of municipalities with heavy losses is long. There are 110 municipalities
where losses were above 10% of the corresponding annual municipal
expenditures. Six of them had aggregate losses (again, likely underestimated, as
we could not match all investors to a municipality) larger than the annual
municipal public expenditures. Five out of these six are located in Nariño, in the
southwest of the country. Aggregate net revenues are only positive in a few
municipalities. Expressed as a percentage of municipal annual expenditures, five
municipalities had in terms of profits a net balance above 10%. With the exception
of Zipaquirá (which is located near Bogotá, in the center of the country), the top
five winners are also located in the southwest of the country, where the two
pyramids were born.
4. Impact on loans
In this section, we test whether the individuals who invested in the pyramids
obtained more loans within the financial sector, presumably to leverage their
investments. Moreover, we would also like to know whether they were paying
their loans back on time. Our prior assumption is that they should have paid the
loans back on time while the pyramids were functioning, but afterwards, once
the pyramids had been shut down, they may have had trouble paying them back.
To address these questions, we first built a control group of investors by
identifying individuals in the SISBEN who did not participate in the pyramids,
but who shared similar characteristics with the investors in the merged datasets
of DMG/DRFE and SISBEN. We do this by using a propensity score matching
technique (the details of which are in the appendix) aimed at providing us with
one similar individual surveyed by the SISBEN for each investor in the matched
SISBEN/Pyramids sample. Then, for all individuals in the treatment and control
groups, the Superfinanciera provides us with individual semiannual loan stocks
and (ex-post) loan ratings from 2006 until 2010.
Since to comply with the habeas data regulation, the dataset had to be merged by
staff from the Superfinanciera, we agreed with them that the dataset would only
have one control individual per investor. The controls were chosen with a
15
matching algorithm of the nearest neighbor according to the PSM described in
the appendix.11 The anonymized sample returned to us by the Superfinanciera—
which contained information on the treatment (investors) and control groups
(similar individuals according to the propensity score)—comprised 269.855
investors and the same number of controls. The descriptive statistics of this
sample are also provided in the appendix.
Using this sample, we begin our empirical analysis by looking at the impact the
pyramids had on loans and on ex-post loan ratings. We run panel data fixed
effects regressions of the following kind, with standard errors clustered at the
municipal level:
 
∙
∙



 , (1)
where  is the variable of interest—for instance, consumer loans for individual
i in semester t. The regression has time effects,, and a time trend,. The data is
semiannual from 2006 through 2010.12 is a dummy variable that takes the
value of 1 if the individual participated in the pyramids, and 0 otherwise. The
coefficient of interest is, which denotes whether respective variables differ
between the treatment and control groups over time.
As mentioned in the previous section, the effects that we estimate should be
interpreted as a lower bound of actual impacts for two reasons. First, the SISBEN
sample does not survey individuals in the upper brackets of the income
distribution, which also happen to comprise those with the greatest access to
credit from the financial sector. Second, it is conceivable that a few of the
investors we were unable to match due to errors in the ID ended up in our control
group. This would also attenuate the size of our coefficients.
To summarize the results, we report plots of over time with 90% confidence
intervals. Figure 1 reports four subplots. The top left one reports the coefficients

11TocomplywiththehabeasdataregulationofColombia,theSuperfinancierasuppressedtheIDsand
replacedthemwithfictitiousonesinthedatasettheyreturnedtous.Inaddition,theyroundedsomeof
thevariablestoavoidreidentificationoftheindividuals.
12Toavoidcollinearity,wedropthe20061coefficients.
16
when  is the stock (in logs) of consumer loans (including credit cards); the top
right one looks at the flows of consumer loans, that is, the change in the stock of
loans for each individual from one semester to the next. At the bottom, we look
at the results when the variable of interest is either the proportion of ex-post good
quality consumer loans (the left hand, labeled “A rated loans”) and
nonperforming loans (the right hand, labeled “D and E rated loans”).The
Superfinanciera classifies consumer loans as belonging in category A (good
loans) if they are not overdue or—if they are—they are not overdue by more than
a month. Nonperforming loans, that is, loans in categories D or E, are overdue by
more than three months.
Figure 1. for consumer loans - complete sample. Fixed-effects panel data regression
results with 90% confidence bands with errors clustered at municipal level. The vertical
lines indicate when the pyramids were shut down.
We begin the analysis by looking first at the results before the pyramids were shut
down, that is, to the left of the vertical line. The top panels show that while the
schemes were operating, the loans acquired by investors from the formal financial
sector were significantly and increasingly higher than for non-investors. The size
0
.1
.2
.3
.4
.5
06h2 07h1 07h2 08h1 08h2 09h1 09h2 10h1 10h2
Natural logarithm of loan stocks
-100
0
100
200
300
06h2 07h1 07h2 08h1 08h2 09h1 09h2 10h1 10h2
Loan flows (thousands of COP)
.01
.02
.03
.04
.05
06h2 07h1 07h2 08h1 08h2 09h1 09h2 10h1 10h2
Percentage of A-rated loans
-.004
-.002
0
.002
.004
06h2 07h1 07h2 08h1 08h2 09h1 09h2 10h1 10h2
Percentage of D and E-rated loans
17
of the effects is considerable. For instance, by the end of 2008, the loans (stocks)
of investors where 39.3% higher than for non-investors. Consistent with the
timing described when discussing the history of the pyramids, at the beginning
of our sample in 2006-2, investors already had higher credit stocks than the
control group, though certainly by a much smaller magnitude: the size of the
difference is 7.6%.
The bottom panels show that, relative to non-investors, the loan ratings of
investors improved steadily while the pyramids were operating: the proportion
of investors’ A-rated loans rose while that of D- and E-rated loans fell. Regarding
the size of the coefficient, for good quality loans—prior to DMG and DRFE being
shut down—the largest coefficient is 4.43%. This is a large figure: in our control
group, 13.6% of credits are A-rated.13 The coefficient in the regression therefore
implies that for investors, the proportion of good quality loans rose by close to
33% while the pyramids were functioning. As for bad quality loans, the last
coefficient while the firms were still operating is 0.279%. Since the stock of bad
quality consumer loans for the control group is 0.9%, the proportion of bad
quality loans fell by 31% for the treatment group prior to the pyramids being shut
down. Moreover, the fact that the coefficients are statistically different from zero
at the beginning of the sample (2006-2) suggests that by that time, the pyramids
had already affected loans’ ratings.
After the government shut down DMG and DRFE, the flow of loans and their ex-
post ratings declined. As for the flow, it fell immediately after the pyramids
ended to negative figures, close to 100,000 pesos (US$43) per loan and period: the
increasing trend in the difference in loan stocks between investors and non-
investors prior to the vertical line, therefore, was reversed. In terms of bad loans
(D- and E-rated) the largest coefficient after the pyramids ended is 0.311
percentage points—that is, an increase in bad loans of 35% relative to the control

13Wecodethecreditratingsofindividualswithoutcreditinthefinancialsectorwithzeros.
18
group. Consistent with the latter, the proportion of investors’ good quality loans
began a steady decline after the pyramids were shut down.
Further results-Loans
Are these results different if we focus on large compared to small investors? To
address this question, we estimate equation (1) for the highest and lowest
quintiles in terms of the amounts invested in the pyramids. In each case, we use
the individuals in the respective quintile as the treatment, with individuals
sharing similar characteristics according to the PSM constituting the
corresponding control group. We expect larger effects for those in the highest
quintile of investments. The results are reported in Figure 2: the panels on the left
hand focus on the highest quintile, those on the right hand, on the lowest. The
scales on the axes are identical in order to facilitate the comparison.
The qualitative patterns when we split the sample between high and low
investment quintiles are very similar to those uncovered when looking at the
whole sample. Pyramid costumers’ obtained more loans from the financial sector
and their credit standings where better than those in the respective control
groups while the schemes were in business—that is, to the left of the vertical lines.
Afterwards, once the schemes were shut down, their loan stocks started to
decrease and their ratings with the banking sector deteriorated.
The qualitative similarities between the two quintiles hide important quantitative
differences. Prior to the pyramids being shut down—and relative to the
respective control groups—the increase in the loan stocks of those in the highest
quintile of investments is almost 11 percentage points above those in the lowest
quintile. Once they were put out of business, the loan flows fell by 360,000 pesos
(US$ 162) in the highest quintile, 20 times more than in the lowest. The non-
performing loans increased by 38% for the highest quintile at their peak after the
pyramids were shut down, compared to 21% for the lowest quintile (relative to
the respective control groups).
19
Figure 2: for consumer loans. Left hand plots: highest quintile of investments; right
hand plots: lowest quintile of investments. Fixed-effects panel data regression results
with 90% confidence bands and clustered standard errors (municipalities). The vertical
line indicates when the pyramids were shut down.
Beyond the quantitative comparison across quintiles, a remarkable result of these
estimations is that even for those individuals in the lowest quintile of
investments, we detect sizeable and statistically relevant effects on their financial
behavior as a consequence of having invested in the pyramids, both before and
after they were shut down.
0
.1
.2
.3
.4
06h2 07h1 07h2 08h1 08h2 09h1 09 h2 10h1 10h2
Natural logarithm of loan stocks
0
.1
.2
.3
.4
06h2 07h1 07h2 08h1 08h2 09h1 09h2 10h1 10h2
Natural logarith m of loan sto cks
-500
0
500
06h2 07h1 07h2 08 h1 08h2 09h1 09h2 10h1 10h2
Loan flows (thousands of C OP)
-500
0
500
06h2 07h1 07 h2 08h1 08h2 09h1 09h2 10h1 10 h2
Loan flows (thousands of C OP)
-.02
0
.02
.04
06h2 07h1 07h2 08 h1 08h2 09h1 09h2 10h1 10h2
Percentage of A-rated loans
-.02
0
.02
.04
06h2 07h1 07h2 08h1 08 h2 09h1 09h2 10h1 10h2
Percentage of A-rated loans
-.005
0
.005
06h2 07h1 07h2 08h1 08 h2 09h1 09h2 10h1 10h2
Percentage of D and E-rated loans
-.005
0
.005
06h2 07h1 07h2 08h1 08 h2 09h1 09h2 10h1 10h2
Percentage o f D and E-rated loans
20
We also study whether there are differences in financial behavior after the
pyramids were shut down, comparing investors who lost money and those who
actually made a profit. Our prior hypothesis is that, once the pyramids were
gone, winners would use the proceeds of their investments to reduce the extra
debt they had acquired investing in them. We thus expect their loan stocks to
have fallen at a faster rate. We also expect that the winners would have better ex-
post credit ratings after the pyramids were shut down. As for differences prior to
November 2008, it is unclear in what direction the results should go. 80% of
investors lost money. It seems reasonable then to suspect that they did not know
the pyramids were illegal or that they would collapse at that time. As for the 20%
that made a profit, it is unclear whether they were lucky, simply happening to
withdraw on time by chance, or whether they had more serious reasons for doing
so.
The results are reported in Figure 3, with plots on the left for the winners and on
the right for the losers. We observe that the winners’ stock of loans decreased at
a faster pace after the pyramids were shut down than the losers’ stock of loans,
as we anticipated. We also find that the ex-post loan ratings deteriorated more
for the winners, a result that is at odds with our prior assumptions: we were
expecting that once the pyramids were shut down, the winners’ loan ratings
would improve relative to that of the losers. Nevertheless, this latter result should
be taken with a grain of salt. The sample gets significantly smaller on the winners
side, inasmuch as only 17% of the investors actually made a profit (in the merged
sample) and of them, fewer than a quarter actually had loan ratings in the dataset.
21
Figure 3: for consumer loans - winners and losers. Fixed-effects panel data
regression results with 90% confidence bands and clustered standard errors
(municipalities). The regression is estimated for individuals who could be correlated
with the SISBEN data. The vertical line indicates the time when the pyramids were shut
down. Plots on the left are for winners, those on the right, for losers.
5. Impact on deposits
One popular story told by the founder of DMG around the time the government
took control and shut the firm down was that he was being pursued by a bank-
conspiracy. His rationale was that the higher yields offered by his business
0
.1
.2
.3
.4
.5
06h2 07h1 07h2 08h1 08h2 09h1 09 h2 10h1 10h2
Natural logarithm of loan stocks
0
.1
.2
.3
.4
.5
06h2 07h1 07h2 08h1 08h2 09h1 09h2 10h1 10 h2
Natural logarith m of loan sto cks
-200
0
200
400
600
06h2 07h1 07h2 08 h1 08h2 09h1 09h2 10h1 10h2
Loan flows (thousands of C OP)
-200
0
200
400
600
06h2 07h1 07 h2 08h1 08h2 09h1 09h2 10h1 10 h2
Loan flows (thousands of C OP)
0
.02
.04
.06
06h2 07h1 07h2 08 h1 08h2 09h1 09h2 10h1 10h2
Percentage of A-rated loans
0
.02
.04
.06
06h2 07h1 07h2 08h1 08 h2 09h1 09h2 10h1 10h2
Percentage of A-rated loans
-.005
0
.005
.01
06h2 07h1 07h2 08h1 08 h2 09h1 09h2 10h1 10h2
Percentage of D and E-rated loans
-.005
0
.005
.01
06h2 07h1 07h2 08h1 08h2 09h1 09h2 10h1 10h2
Percentage o f D and E-rated loans
22
reduced the deposits that banks were able to attract from the public. Consistent
with this claim, Carvajal et al. (2009) mention that there is anecdotal evidence that
some of the schemes they analyzed diverted deposits from the financial sector.
We estimate—again as far as we are able to tell, for the first time in the
literature—whether and by how much the pyramids impacted deposits. We find
that the pyramids had significant effects on deposits.
As mentioned, the Superfinanciera does not record data on deposits at the
individual level. To study the pyramids’ impact on deposits, we exploit the
municipal variability in deposits (which the Superfinanciera does report) and the
intensity of the pyramids’ presence in respective municipalities (we build
municipal variables that work as proxies of the relative importance of the
pyramids.)
We use two definitions of the municipal intensity of the pyramids: first, using the
location of investors based on the data matched with the SISBEN survey, we
calculate the number of investors per capita (14+) across municipalities. We call
this variable Ipc. Second, again using the location of individuals, we first sum the
total amount invested by individuals matched to each municipality, and then
calculate the ratio of this statistic to the respective municipal government’s total
annual expenditures. We label this variable K/E. In Table 3, we report the
summary statistics of these ratios.
Table 3. The summary statistics of the intensity of pyramids in municipalities.
Investors per capita, 14+: Ipc. Total municipal investments/municipal expenditures: K/E.
Variables Obs Mean
SD Min
Max p10 Median p90
I
pc 1089 0,0100
0,0247
0 0,2074
0,0009
0,0026 0,0196
K / E 1082 0,1555
0,4810
0 5,4565
0,0079
0,0381 0,2087
Finally, we estimate the pyramids’ effects on municipal deposits in the formal
financial sector by running regressions of the following type:
 
 

 
 

 , (2)
23
where  refers to the deposits of type j at time t in municipality i in constant
pesos and expressed in logs. We estimate this for three types of deposits: total
deposits, deposits in savings accounts, and certificate deposits (CDs), which the
Superfinanciera records at a quarterly frequency by municipality.14 βj0 is a
constant, T is a time trend, and λt are quarterly time effects. is either Ipc or K/E
in municipality i. i is a municipal fixed effect and  is the error term.15 We
report the results with errors clustered at the municipal level. The results without
clusters and with clusters at the state (departamento) level yield similar
conclusions.16
The parameter of interest,, represents the differences over time in deposits of
type j if the “intensity” of the pyramids is marginally increased in a municipality.
If the pyramids happened to have shifted deposits away from the financial sector,
 should be negative. To summarize our findings, we plot  with 90%
confidence bands in Figure 4 for three types of deposits (j) and two proxies of the
intensity of Ponzi schemes in the municipalities (X). The left columns use Ipc, the
right ones, K/E to identify the effects.
The two top plots—that is, those that look at the pyramids’ effects on total
deposits—show that during the last quarter of 2006, two years prior to the
pyramids being shut down, the coefficients become negative; that is, deposits in
municipalities most affected by the pyramids fell below those least affected (or
not affected at all). This downward trend reaches a trough during the second
quarter of 2008—that is, a quarter and a half prior to the pyramids being shut
down. At that point, the gap between the two series is statistically significant.
After the pyramids ended, the coefficients again become insignificant.

14TotaldepositsarethesumofCDs,savingaccounts,andcurrentaccounts.In2006,51%oftotaldeposits
weresavingsaccountsand29%wereCDs.Wedonotfocusoncurrentaccounts,astheirbalancesare
mostlydrivenbycorporateandgovernmenttransactions.
15Wedroptheinteractionof2006Itoavoidmulticollinearity.
16Availablefromtheauthorsuponrequest.
24
Figure 4:  and 90% confidence bands. The left hand panels are estimated with j = Ipc;
the right hand panels are estimated with j =K/E; confidence bands use errors clustered
at the municipality level.
What does the size of the coefficients tell us? At the trough for the top left plot,
the coefficient is -1.08. That means that a 1% increase in the ratio of the population
investing in the pyramids reduces total deposits in the banking sector by 1.08%.
To give further intuition to the size of the effects, we report in Table 4 two
additional results for each case. On the one hand, we report by how much
deposits j fall if Ipc or K/E change by one standard deviation; on the other hand,
the table reports the predicted effect on deposit j in the top five municipalities
-2 -1 012
06q2 06q4 07q2 07q4 08q2 08q4 09 q2 09q4 10q2 10q4
Total Deposits
-.1 -.05 0.05 .1
06q2 06q4 07q2 07q4 08q2 08q4 09q2 09q4 10 q2 10q4
Total Deposits
-2 -1 0 1 2 3
06q2 06q4 07q2 07q4 08q2 08q4 09 q2 09q4 10q2 10q4
Saving Accounts
-.1 -.05 0.05 .1
06q2 06q4 07q2 07q4 08q2 08q4 09q2 09q4 10 q2 10q4
Saving Accounts
-6 -4 -2 0 2
06q2 06q4 07q2 07q4 08q2 08q4 09 q2 09q4 10q2 10q4
CDs
-.3 -.2 -.1 0.1
06q2 06q4 07q2 07q4 08q2 08q4 09q2 09q4 10 q2 10q4
CDs
25
with a stronger average presence of pyramids, in terms of Ipc or K/E, respectively.
These effects are reported for the respective troughs identified in Figure 4—that
is, for either the second or third quarters of 2008, the quarters immediately prior
to the pyramids being shut down.
As reported in Table 4, a one standard deviation increase in Ipc reduces total
deposits in a municipality by 2.7%. In the top five municipalities for per capita
investors, total deposits fell on average by 20% (relative to municipalities without
pyramids). Reassuringly, if we use K/E to identify the effects, the interpretation
of the coefficients yields similar results—see the right hand panels: one standard
deviation increase in K/E reduces total deposits in the municipalities by 2.4%, and
the predicted reduction in deposits in the five municipalities with the highest K/E
is 22%.
In the middle panels of Figure 4, which report the pyramids’ impact on savings
accounts, we see a significant drop in the coefficients, especially during the final
two quarters prior to the end of the pyramids. According to Table 4, the size of
these coefficients at the trough indicate that a one standard deviation increase in
either Ipc or K/E decreased the deposits in savings accounts by 2.9%. Moreover,
the pyramids caused a decrease of 22% to 27% in savings accounts in the top five
most affected municipalities.
Table 4. The effects on total deposits, saving accounts and CDs of changes in Ipc and K/E.
The effects in the left panel are identified with Ipc, those in the right panel, with K/E. The
numbers reported correspond to the effects at the respective troughs (see Figure 4).
Total Deposits
β2, 08q2*SD -2,7%
β2, 08q2*SD -2,4%
β2, 08q2*Mean Top 5 -20,4%
β2, 08q2*Mean Top 5 -22,3%
Saving accounts
β2, 08q2*SD -2,9%
β2, 08q3*SD -2,9%
β2, 08q2*Mean Top 5 -22,3%
β2, 08q3*Mean Top 5 -26,8%
CDs
β2, 08q3*SD -9,2%
β2, 08q3*SD -10,2%
β2, 08q3*Mean Top 5 -70,1%
β2, 08q3*Mean Top 5 -94,6%
26
The bottom panels in both Figure 4 and Table 4 show the evolution of certificate
deposits, CDs, a popular financial vehicle for saving money (almost a third of
deposits in the banking sector consist of CDs). CDs pay higher interest rates than
savings accounts, but are also less liquid and thus are a better substitute for
investment in pyramids. That is indeed what the results suggest. Both the size
and the persistence of the effects are great. In terms of persistence—unlike with
savings and total deposits—the decline in the value of CDs continued well
beyond the end of DMG and DRFE. Even by the end of 2010, they had not
recovered completely. The size is also very large. A one standard deviation
increase in Ipc reduced CDs in a municipality (at the trough) by 9.2%; the same
statistic, using K/E to identify the effect, suggests a reduction of 10.2% in CDs.
For the top five municipalities vis-à-vis the presence of pyramids, CDs are
predicted to have fallen between 70% and 95%.
If we extrapolate the figures using investors per capita at a national scale to get
back-of-the-envelope estimates of the pyramids’ impact on nationwide deposits
in the formal financial sector, we find that they caused total deposits to fall by
0.6% and CDs by 2%. The impact of DMG and DRFE on deposits in the financial
sector is significant and economically relevant.
6. Conclusions
DMG and DRFE, two Ponzi schemes that operated in Colombia through 2008,
attracted over half a million customers who invested resources equivalent to 1.2%
of the country’s GDP, an amount corresponding to 22% of the total deposits of
the country’s largest bank.
While Ponzi schemes are more common that thought even in developed
countries, our understanding of their consequences is mostly anecdotal. Using a
unique dataset—merging the universe of investors in the pyramids with their
loan records in the formal financial sector—we show that before the government
shut down these firms, the individuals who invested in DMG and DRFE acquired
close to 40% more loans in the financial sector compared to similar individuals
who did not invest in the pyramids. Moreover, we also find that deposits in the
27
formal financial sector fell in municipalities heavily affected by these two
pyramids: individuals pulled resources away from the financial sector in order
to invest in Ponzi schemes. A one standard deviation increase in the municipal
presence of the pyramids decreased deposits in saving accounts by 2.9% and CDs
by more than 9%. After the firms were shut down, the ex-post loan ratings of
investors deteriorated: nonperforming loans increased by 35% compared to
similar individuals who did not participate in the schemes. Moreover, we show
that their loan stocks started falling and that two years later, deposits had not yet
fully recovered.
Beyond being able to pinpoint the effects of Ponzi schemes on the financial
behavior of households—as far as we can tell, for the first time in the literature—
we believe the lessons learned here extend beyond them. Indeed, some of the
effects we were able to estimate may have a parallel in episodes we generically
call “bubbles.” Samuelson (1957) used “Ponzi schemes” interchangeably with
“chain letters” and “bubbles.” Charles Kindelberger, in his comprehensive
history of financial crises (Kindelberger and Aliber, 2005), describes bubbles as
euphoric periods during which “an increasing number of investors seek short-
term capital gains from the increases in the prices of real estate and of stocks
rather than from the (…) income based on the productive use of these assets.”
That is analogous to what people seem to be doing when investing in Ponzi
schemes. Thus, our results might hint at how financial consumers behave during
episodes we identify as bubbles. This opens up interesting avenues of research
related to the studying of the loans, loan ratings and deposits in the financial
sector of individuals who invested in assets, and whose behavior we later
associate with bubble-like patterns.
28
References
Artzrouni, M. (2009). The Mathematics of Ponzi Schemes. Mathematical Social
Sciences, 58(2), 190-201.
Bhattacharya, U. (2003). The Optimal Design of Ponzi Schemes in Finite
Economies. Journal of Financial Intermediation, 12(1), 2-24.
Caribbean Policy Research Institute (CaPRI). (2008). Investigating Informal
Investment Schemes in Jamaica. Kingston, Jamaica: Caribbean Policy Research
Institute.
Carvajal, A., Monroe, H. K., Wynter, B., & Pattillo, C. A. (2009). Ponzi Schemes in
the Caribbean (No. 9-95). International Monetary Fund.
Cortés, D., Santamaría, J., & Vargas, J. (2016). Economic Shocks and Crime:
Evidence from the Crash of Ponzi Schemes. Journal of Economic Behavior &
Organization, Volume 131, Part A, 263-275.
Deason, S., Rajgopal, S., & Waymire, G. B. (2015). Who Gets Swindled in Ponzi
schemes? Available at SSRN 2586490.
ELCA, 2010. Longitudinal survey of Colombia: Universidad de los Andes.
Jarvis, C. (2000). The Rise and Fall of the Pyramid Schemes in Albania. IMF Staff
Papers, 47(1), 1-29.
Kindleberger, Charles Poor, & Aliber, Robert Z. 2005. Manias, Panics, and Crashes:
a History of Financial Crises. Hoboken, N.J.: John Wiley & Sons.
Lewis, M. K. (2012, December). New Dogs, Old Tricks. Why do Ponzi Schemes
Succeed? Accounting Forum (Vol. 36, No. 4, 294-309). Elsevier.
Samuelson, P. A. (1957). Intertemporal Price Equilibrium: A Prologue to the
Theory of Speculation. Weltwirtschaftliches Archiv, 181-221.
Smith, F. (2010). Madoff Ponzi Scheme Exposes the Myth of the Sophisticated
Investor. U. Balt. L. Rev., 40, 215.
Tennant, D. (2011). Why do People Risk Exposure to Ponzi Schemes?
Econometric Evidence from Jamaica. Journal of International Financial Markets,
Institutions and Money, 21(3), 328-346.
29
Appendix
One of the uses of the merged dataset was to obtain a control group for the
investors, by identifying individuals in the SISBEN who did not participate in the
pyramids but who shared similar characteristics with those who did. For that
purpose, we implemented a propensity score-matching technique, which
estimated a propensity score for each individual in the sample, and paired
individuals in the control and treatment groups following the nearest neighbor
algorithm.
The main purpose of propensity score matching is to balance the distribution of
observed covariates (Lee, 2013), so that there are no systematic differences in the
distribution of covariates between the two groups. In this work, the equality of
means of observed characteristics in the treatment and control groups was
examined using the Imbens Statistic, which controls by the size of the sample.
Table 5 summarizes the results of this analysis, which allow us to conclude that,
effectively, no significant differences in observable characteristics exist between
the treatment and control groups.
Table 5. Treatment and control groups, descriptive statistics.
Invested in DRFE
Variables
Mean
Control
Mean
Treatment
Mean
Difference
Imbens
Statistic
Significance
Imbens
Male 0,529 0,513 0,016 0,032
Age 39,565 39,675 -0,110 -0,008
Income $ 93.767 $ 99.846 -$ 6.079 -0,026
No education 0,072 0,078 -0,007 -0,025
Incomplete elementary 0,339 0,283 0,056 0,122
Complete elementary 0,216 0,229 -0,013 -0,032
Incomplete high school 0,132 0,138 -0,006 -0,018
Complete high school 0,181 0,218 -0,037 -0,092
Secondary/post education 0,060 0,053 0,007 0,030
Cohabitation 0,213 0,224 -0,010 -0,025
Married 0,333 0,321 0,012 0,025
Widowed 0,021 0,021 0,000 0,001
Single/divorced 0,432 0,434 -0,002 -0,003
Household size 4,253 4,048 0,204 0,107
Proportion of kids 0,145 0,149 -0,004 -0,024
Household head's years of
education 3,072 3,101 -0,029 -0,018
30
Household head's earnings $ 160.703 $ 172.144 -$ 11.441 -0,036
Household's per capita income $ 64.334 $ 68.106 -$ 3.772 -0,028
SISBEN score 13,987 14,079 -0,092 -0,010
Invested in DMG
Male 0,474 0,465 0,009 0,018
Age 41,765 41,405 0,360 0,027
Income $ 179.992 $ 242.996 -$ 63.004 -0,069
No education 0,048 0,040 0,008 0,039
Incomplete elementary 0,215 0,159 0,055 0,142
Complete elementary 0,211 0,171 0,040 0,102
Incomplete high school 0,200 0,166 0,033 0,086
Complete high school 0,223 0,295 -0,072 -0,165
Secondary/post education 0,104 0,168 -0,064 -0,189
Cohabitation 0,280 0,248 0,032 0,072
Married 0,307 0,333 -0,026 -0,057
Widowed 0,029 0,027 0,002 0,015
Single/divorced 0,384 0,391 -0,008 -0,016
Household size 3,947 3,723 0,224 0,132
Proportion of kids 0,128 0,126 0,002 0,011
Household head's years of
education 3,522 3,882 -0,360 -0,204
Household head's earnings $ 275.459 $ 352.992 -$ 77.533 -0,179
Household's per capita income $ 124.556 $ 166.102 -$ 41.546 -0,047
SISBEN score 18,348 21,212 -2,864 -0,240
Invested in DRFE & DMG
Male 0,502 0,490 0,011 0,023
Age 39,755 40,114 -0,359 -0,025
Income $ 113.774 $ 120.852 -$ 7.078 -0,025
No education 0,076 0,090 -0,014 -0,050
Incomplete elementary 0,336 0,288 0,047 0,103
Complete elementary 0,205 0,221 -0,017 -0,041
Incomplete high school 0,138 0,134 0,004 0,010
Complete high school 0,174 0,213 -0,039 -0,100
Secondary/post education 0,072 0,053 0,019 0,078
Cohabitation 0,246 0,255 -0,009 -0,021
Married 0,332 0,327 0,005 0,011
Widowed 0,026 0,028 -0,002 -0,013
Single/divorced 0,397 0,391 0,006 0,012
Household size 4,190 3,997 0,193 0,101
Proportion of kids 0,145 0,149 -0,004 -0,023
Household head's years of
education 3,156 3,144 0,012 0,007
Household head's earnings $ 189.895 $ 202.264 -$ 12.369 -0,038
Household's per capita income $ 76.675 $ 80.887 -$ 4.212 -0,029
SISBEN score 14,676 14,453 0,223 0,023
31
The propensity score is estimated using a probit model, which defines the
probability of investing in Ponzi schemes as a function of individual
characteristics such as age, income, years of education, gender, marital status,
location and Sisben score, as well as household characteristics, like the earnings
of the head of the household, the proportion of kids in a household, household
size, the household head’s years of education, the household head’s mean
earnings, and the household's per capita income. All these variables were
measured during the second wave of SISBEN, conducted between 2003 and 2007,
and which collected information on 32.5 million individuals nationwide. (The
total population in 2008 was 44.5 million.) This allows us to obtain investors’
socioeconomic characteristics prior to the end of the pyramids. Estimates of the
probit model are given in Table 6.
Table 6. Probit estimates
Variable Coefficient SE
Age -0.003*** 0.000
Income 0.000*** 0.000
Sisben score 0.004*** 0.000
Years of education 0.029*** 0.001
Married -0.012*** 0.002
Widowed 0.055*** 0.002
Single/Divorced 0.038*** 0.005
Male -0.011*** 0.001
Household head's years of education 0.011*** 0.001
Proportion of kids -0.018*** 0.002
Household size -0.005*** 0.001
Household head's earnings 0.000*** 0.000
Household's per capita income -0.000 0.000
State dummy 1 0.251*** 0.045
State dummy 2 0.278*** 0.044
State dummy 3 0.322*** 0.044
State dummy 4 0.305*** 0.042
State dummy 5 0.328*** 0.041
State dummy 6 0.185*** 0.048
State dummy 7 0.283*** 0.043
State dummy 8 0.344*** 0.041
State dummy 9 0.246*** 0.046
State dummy 10 0.376*** 0.036
State dummy 11 0.378*** 0.040
State dummy 12 0.285*** 0.044
State dummy 13 0.350*** 0.040
State dummy 14 0.254*** 0.047
State dummy 15 0.350*** 0.039
32
State dummy 16 0.358*** 0.038
State dummy 17 0.362*** 0.043
State dummy 18 0.201*** 0.047
State dummy 19 0.215*** 0.047
State dummy 20 0.260*** 0.045
State dummy 21 0.187*** 0.048
State dummy 22 0.279*** 0.045
State dummy 23 0.172*** 0.048
State dummy 24 0.090* 0.049
State dummy 25 0.204*** 0.050
State dummy 26 0.290*** 0.043
State dummy 27 0.478*** 0.030
State dummy 29 0.088 0.060
State dummy 30 0.106 0.068
State dummy 31 0.127** 0.053
State dummy 32 0.196** 0.100
State dummy 33 0.211*** 0.059
Notes: The dependent variable takes the value of 1 if the individual invested in Ponzi schemes,
and 0 otherwise. The dummy variable for state 28 (San Andres) is dropped to avoid
collinearity. Likewise, the comparison category for marital status is Cohabitation. *** p<0.01, **
p<0.05, * p<0.1.
The estimation results in Table 6 indicate that, according to our model, the
probability of investing in Ponzi schemes negatively correlates with age, being
married, being male, household size, and larger proportions of kids. Larger
households with larger proportions of kids and lower per capita incomes
probably cannot spare any money to invest in Ponzi schemes, and thus definitely
act against the probability of investing. On the other hand, the probability of
investing positively correlates with being widowed, single or divorced. Likewise,
it seems to relate positively to a household head's earnings, and with being
located in the states of Putumayo (State dummy 27), Cundinamarca (State
dummy 11) and Nariño (State dummy 17), something corroborated by our results
in the previous sections.
The probit estimates are used to calculate the propensity score for all individuals.
It is crucial for the validity of the matching that there is a common support. Figure
6 depicts the kernel densities of the propensity scores for both investors (treat)
and non-investors (control) in the SISBEN survey. The results allow us to
conclude that there is sufficient overlap between the propensity scores of the
treatment and control group.
33
The purpose of this propensity score matching is to obtain a control group of
investors by identifying individuals in the SISBEN who did not participate in the
pyramids, but shared similar characteristics with those who did. For this we
finally implement a matching algorithm of nearest neighbor with no
replacement, in order to ensure that there is one control for each individual in the
treatment group. The nearest neighbor matching matches a subject from the
control group to a subject in the treatment group based on the closest propensity
score. With the no replacement property, if for a treated unit, forward and
backward matches happen to be equally good, the program randomly draws
either the forward or backward match (Cox-Edwards & Rodriguez-Oreggia,
2009). This leaves us with 269,855 investors and the same number of controls.
Figure 6. Kernel densities of propensity scores for investors and non-investors in the
SISBEN survey.
01234
Density
0.2 .4 .6 .8 1
Investment Probability
Treat Density
Control Density
Investment Probability Distribution
... In the occurrence of the failure of this scheme, the funds were lost and that could lead to loss of capital formation which may hamper economic growth in the country. Hofstetter et al (2017) found that the operation of ponzi schemes reduced municipal savings deposits by 2.9% and certificate of deposit by about 10% in Colombia as participant in Ponzi schemes withdraw their deposit to invest in Ponzi schemes. According to Simon (2017), prior to system collapse, Nigerians lost N18 billion to MMM in 2016 through about three million participants in the scheme. ...
... Furthermore, the collapse of ponzi schemes may have increased bad debt for banks in Nigeria. Hofstetter et al (2017) observed that operation of Ponzi schemes has the tendency of increasing loan and non-performing loans created by the formal financial sector. Using DMG and DRFE schemes in Colombia, they found that people who invested in Ponzi schemes have 39% more loans with formal financial system than those who did not. ...
Article
Full-text available
ABSTRACT This study examined the factors influencing Ponzi scheme participation in Nigeria. A multi-stage sampling procedure was used to select 384 participants for the study. Descriptive statistic and One-Way ANOVA was employed to analyze the data. The result of the analysis revealed that friends’ recommendation, expected benefits, current economic situations, ease of obtaining funds and the get rich quick syndrome were the most influencing factors on Ponzi scheme participation and these factors were significant with the gender, age, marital status, employment status and educational level at 5% level of significance. The study concluded that ponzi scheme participation was mostly influenced by friend’s recommendation, expected benefit, current economicsituations, ease of obtaining funds and the Get rich quick syndrome. Therefore, the study recommended that the Nigerian government should integrate Ponzi schemes into the Nigerian financial system and subject them to the country’s regulatory environment to minimize the consequences of future failure of the schemes. Keywords: Ponzi scheme participation, Friends’ recommendation, Expected benefits, Economic Situations, Get rich quick syndrome
... En la medida en que la colectividad afectada aumenta, en esa misma medida se siente el efecto en el orden público económico, situación que menoscaba la confianza en el mercado y distorsiona la economía. Máxime cuando normalmente los afectados son personas en estratos socioeconómicos bajos, con lo que el efecto de la pérdida es bastante mayor (Hofstetter, Rosas y Urrutia, 2017). ...
Article
Full-text available
La presente investigación se ha propuesto establecer los alcances de la Ley 1700 de 2013, y los efectos que puede tener sobre los participantes en dichas empresas. Especialmente si se tiene en cuenta que son los consumidores quienes están en el centro de dicha relación comercial, y que, como consumidores requieren de especial protección del Estado. Para ello, se usó el método inductivo, partiendo del análisis de la norma en concreto, y de su posterior comparación con otras normas. Tal ejercicio nos permitió concluir, que la manera como se regularon las empresas multinivel, parece no ser la mejor para frenar los abusos que con esta figura jurídica se realizan. Se requiere que la Ley esté acompañada de medidas que al tiempo favorezcan la protección de las personas que se vinculan a las mismas; para evitar que se puedan afectar los intereses de los consumidores, y facilitar la supervisión y vigilancia de la Superintendencia de Sociedades.
Article
Apart from a few shorter papers inspired by the nomination of a new crime prohibiting the organization of ‘pyramid games’ by the Hungarian Criminal Code in 1996, the topic of ‘pyramid and Ponzi schemes’ remained of little interest to Hungarian legal scholars. Internationally, the topic has garnered increased attention due to the grave socio-economic effects of ever newer scheme-collapses, from the high-profile American Madoff (2009) to the myriad less-known cases from emerging systems like the fiasco of the Albanian pyramid schemes in the mid-1990s, pyramid schemes camouflaged as multi-level marketing (MLM) ventures, or their online versions more recently. Comparative works that would juxtapose the pertaining laws and experiences of the United States with those of Hungary are lacking. To fill the void, this article contrasts a select number of differing regulatory approaches. At one end of the spectrum is the United States (US), which, instead of passing sector-specific laws, mobilized and adapted the enforcement tools of all utilizable branches of law to combat the schemes. While in the US this has been uniquely primed by securities laws, in Hungary the task remains limited to what criminal law and the criminal justice system could offer, coupled with the dominantly ad hoc reactions of the Hungarian Securities and Exchange Commission (SEC). Development of tests to distinguish legitimate Multi-Level Marketing (MLM) companies from pyramid schemes disguised as such represents the only segment where significant rapprochement occurred between the US, the European Union (EU) and therefore also Hungary. For contrast and illustration of the other end of the spectrum, the systems that were forced to react to risks corollary to the schemes by enacting sector-specific laws, the most recent regulatory reactions of India, Myanmar, and Sri Lanka had to be resorted to. As the latter two imposed complete bans on all MLMs, it is only the Indian 2019 comprehensive act that attempts to combat the schemes and akin forms of financial fraud relying on a new comprehensive regulatory model. The Philippines is an interesting mixture made of local and transplants that readily proves that the solutions of the most developed US system could successfully be transplanted into a significantly different socio-economic environment.
Chapter
The Impact of Collapsed Pyramid and Ponzi Schemes on Access to Finance by Vulnerable Consumers by Prof. Tibor Tajti (Thaythy) © ABSTRACT The paper aims to analyze one of the rarely discussed paradoxes corollary to making affordable financial accessible to the neediest: the paralyzing effects the collapse of pyramid and Ponzi schemes have on vulnerable consumers’ access to finance. As the schemes – save a few notable exceptions – collapse normally within a year or two making investors lose all their investments with minimal or no chances of any recovery, that may hinder and otherwise negatively affect the efforts aimed at making affordable finance available to the poorest segments of the society. Although no country is immune to these, they appear in fledgling financial and legal systems regularly well-before the paraphernalia of a robust financial system get established and stabilized. This makes the topic of special importance to the latter category of countries. The negative effects may take various forms from the ensuing ‘trust shock’ that may survive for more generations. The collapse may cause financial exclusion as well because the investors having lost their savings would not be in the position to make the required downpayments, contributions, or pay the minimal but required fees that would allow them to qualify for projects targeting vulnerable consumers. Last but not least, as the collapse of some high-profile schemes showed, this can endanger even the entire financial system. Schemes of such magnitudes present a systemic risk that could negatively impact policies aimed at helping vulnerable consumers, too. To make things worse, the mushrooming of digital Ponzi schemes (or Unlicensed Digital Investment Schemes - UDIS) thanks to the spreading of such new technologies as Blockchain, or novel methods of exploiting mobile telecommunication networks, make the problem even more acute. This because now the promoters can not only more easily and cheaply reach vulnerable investors through their computers, but the horizons of their activities have as well radically been expanded. Considering the above, the article will first briefly introduce pyramid and Ponzi schemes by canvassing the saga of several collapsed pyramid and Ponzi schemes in fledgling economies where vulnerable consumers inherently form the dominant percentage of the population and thus the true dimensions of the devastating impact of collapsed schemes are most visible. These include examples from Ethiopia, Nigeria, Myanmar, and Sri Lanka, as well as from the Albania and Russian of the 1990s, the first years of transition towards market economy. Given the focus on countries with fledgling regulatory frameworks, special focus is devoted to multi-level marketing schemes (MLM) that routinely surface in these countries often luring the poorest segments of the society. They present a problem many of the MLMs are disguised pyramid schemes. As distinguishing legitimate and fraudulent MLM schemes remains a hurdle even in the United States (US) as the system possessing the most sophisticated securities regulatory system, the freshly appointed regulators of the fledgling systems face insurmountable problems dealing with these preventing vulnerable consumers.
Article
Full-text available
The global financial crisis has led to the unearthing of numerous Ponzi schemes around the world. This paper examines the factors determining individuals' extent of exposure to Ponzi schemes. The theoretical literature on investor gullibility and risk tolerance was used to develop an economic model to explain such exposure. The model was empirically tested by analyzing the results of a survey of 402 investors in Jamaican Ponzi schemes. A profile of persons most likely to be highly exposed to Ponzi schemes was created, along with a delineation of the most critical factors accounting for the popularity of such schemes.
Article
Full-text available
As no rational agent would be willing to take part in the last round in a finite economy, it is difficult to design Ponzi schemes that are certain to explode. This paper argues that if agents correctly believe in the possibility of a partial bailout when a gigantic Ponzi scheme collapses, and they recognize that a bailout is tantamount to a redistribution of wealth from non-participants to participants, it may be rational for agents to participate, even if they know that it is the last round. We model a political economy where an unscrupulous profit-maximizing promoter can design gigantic Ponzi schemes to cynically exploit this “too big to fail” doctrine. We point to the fact that some of the spectacular Ponzi schemes in history occurred at times where and when such political economies existed—France (1719), Britain (1720), Russia (1994), and Albania (1997).
Article
Full-text available
A first order linear differential equation is used to describe the dynamics of an investment fund that promises more than it can deliver, also known as a Ponzi scheme. The model is based on a promised, unrealistic interest rate; on the actual, realized nominal interest rate; on the rate at which new deposits are accumulated and on the withdrawal rate. Conditions on these parameters are given for the fund to be solvent or to collapse. The model is fitted to data available on Charles Ponzi's 1920 eponymous scheme and illustrated with a philanthropic version of the scheme.
Article
Full-text available
This paper investigates the impact of international migration on technical efficiency, resource allocation and income from agricultural production of family farming in Albania. The results suggest that migration is used by rural households as a pathway out of agriculture: migration is negatively associated with both labour and non-labour input allocation in agriculture, while no significant differences can be detected in terms of farm technical efficiency or agricultural income. Whether the rapid demographic changes in rural areas triggered by massive migration, possibly combined with propitious land and rural development policies, will ultimately produce the conditions for a more viable, high-return agriculture attracting larger investments remains to be seen.
Article
In November 2008, Colombian authorities dismantled a network of Ponzi schemes, making hundreds of thousands of investors lose tens of millions of dollars throughout the country. Using original data on the geographical incidence of the Ponzi schemes, this paper estimates the impact of their breakdown on crime. We find that the crash of Ponzi schemes differentially exacerbated crime in affected districts. Confirming the intuition of the standard economic model of crime, this effect is only present in places with relatively weak judicial and law enforcement institutions, and with little access to consumption smoothing mechanisms such as microcredit. In addition, we show that, with the exception of economically-motivated felonies such as robbery, violent crime is not affected by the negative shock.
Article
What lessons can be drawn from the unprecedented growth and spectacular collapse of financial pyramid schemes in Albania? This paper discusses the origins of the pyramid schemes and the way the authorities handled them. It also analyzes the economic effects of the pyramid schemes, concluding that despite the descent into anarchy triggered by the schemes' collapse, their direct effects on the economy are difficult to specify and appear to have been limited. Finally, the paper argues that prevention of pyramid schemes is better than cure and that governments and international financial institutions should be vigilant in clamping down on frauds. Copyright 2000, International Monetary Fund
Madoff Ponzi Scheme Exposes the Myth of the Sophisticated Investor
  • F Smith
Smith, F. (2010). Madoff Ponzi Scheme Exposes the Myth of the Sophisticated Investor. U. Balt. L. Rev., 40, 215.
Longitudinal survey of Colombia
  • Elca
ELCA, 2010. Longitudinal survey of Colombia: Universidad de los Andes.