Lost in the mail: a field experiment on crime.
Castillo, Marco ; Petrie, Ragan ; Torero, Maximo 等
I. INTRODUCTION
Crime and opportunistic behavior, e.g., shirking on the job or
pilfering from the office when no one is looking, impose high costs on
firms and individuals. Even the time spent in protecting oneself from
theft or others trying to take advantage of a situation can be onerous.
While the costs associated with each event may not be large, and may be
compensated for by a large number of positive transactions, small losses
can add up. Indeed, some markets might not even emerge because losses
eat up too much of potential profits. This is especially important in
developing countries where markets are weak or missing. Measuring these
types of costs is difficult because observational data is not available
for missed transactions (those that had been a priori assessed to be too
risky to enter into). Costs due to corruption, such as at ports,
borders, and police stops, have been estimated, but these measures might
reflect local monopoly power (Olken and Barron 2009) rather than the
distribution of costs faced by the population at large. In this paper,
we address these issues with a field experiment. We examine how an
activity that is important for trade, the delivery of mail, is affected
by crime, who in the population suffers the most, and the underlying
motivations.
We show that the costs of crime are nontrivial and that not only
shirking but strategic crime explain their existence. While sending
something in the mail would seem to be an ordinary task that need not be
given much thought, the reliability of delivery has important effects on
the ability of a firm or an individual to safely transport goods at a
low cost. Our experiment suggests that engaging in exchange in such an
environment can be costly and that certain cues can make mail more
likely to be "lost." Clearly, when these crimes of opportunity
are part of daily life, trade is less efficient. Importantly, we show
that these costs are not distributed equally. Middle-class households
are disproportionally affected.
Our empirical strategy is novel and simple, and it allows us to
measure the probability that a piece of mail arrives at its destination,
even in areas where the use of mail services is low. Our design provides
counterfactual information on what would have happened if the service
had been used. We send identical envelopes to different households in
Lima, Peru, from two American cities and record arrivals. Peru is an
interesting case study because it is representative of middle-level
developing countries struggling to create market-based institutions and
integrate with the global economy. The experiment includes a large
population of volunteer households across neighborhoods of different
socioeconomic backgrounds. To better understand the motivation behind
the commission of crime, we manipulate the contents and the sender of
the mail. In particular, every household was sent four envelopes over
the course of a year. Two envelopes had a sender with a foreign name and
two had the last name of the sender and recipient matched (to indicate
the letter came from a family member). Finally, one of each of the two
envelopes contained something inside the enclosed card (a small amount
of money) that could not be easily detected without careful attention.
(1) The other envelope contained just the enclosed card. All these
modifications were as subtle as possible, and the order in which each
different envelope was sent was random.
Our design allows us to develop a behavioral measure of crime. We
can directly, measure crime and its differential impacts. Because we
send mail across all neighborhoods, we can measure the level of crime
even in places where the population may not normally use these services
for fear of mail not arriving. Specifically, the design identifies
whether crime has occurred or not, which segments of the population are
more likely to suffer from crime, and whether this conforms with
economic rationality broadly understood. (2)
We concentrate on the delivery of mail for several reasons. First,
the existence of a reliable mail sector is considered to be instrumental
in the growth of electronic trade (World Bank 2009). Second, mail
services are used by all segments of the population, both rich and poor.
Third, as we will describe, mail delivery is amenable to field
experimentation with little or no intrusion. This is important as the
study of crime could be constrained by ethical considerations and
measurement problems. Fourth, mail delivery is a highly decentralized
activity and likely suffers from moral hazard problems regardless of
firm ownership. For instance, sources of lost noncertified mail are
nearly impossible to detect. Fifth, in our study environment, mail
service, which is normally provided by a public firm, is done by a
private entity. Finally, crime in the mail sector is expected to be
highly correlated with the expected gains and losses of committing a
crime and much less with social pathologies. That is, it is a crime of
opportunity that can help us understand economic motivations.
By manipulating the information made available to the person
handling the mail, we can test several hypotheses behind the commission
of crime. First, mail can be lost because the cost of delivery is larger
than the cost of being caught shirking. Lost mail might be a reflection
of apathy rather than crime. Therefore, comparing rates of lost mail
containing money with those containing no money permits us to detect if
crime, rather than apathy, is taking place. Second, if those handling
mail behave strategically, one would expect that they will make use of
information on the socioeconomic characteristics of the recipient and
the social distance between the recipient and the sender. Therefore,
comparing similar pieces of mail across subgroups can potentially reveal
the expectations of those handling the mail. For instance, if a letter
from a family member is more likely to contain something of value (i.e.,
money), then letters from family members would be lost at a higher rate.
The experiments show that the mail service in Peru is highly
inefficient. The overall rate of mail lost is 18%. (3) The loss rate,
however, hides the fact that mail containing money is lost 21% of the
time while mail containing no money is lost 15% of the time. That is, we
find evidence of shirking as well as crime. Also, crime is targeted
primarily at letters coming from family members, and the quality of
service is not independent of socioeconomic status. Mail is lost at the
same rate (roughly 18%), whether it contains money or not, when sent to
a poor neighborhood. When sent to a more affluent neighborhood, however,
mail without money is lost only 10% of the time and mail with money is
lost 17% of the time. Households in middle-income neighborhoods have the
highest loss rates. This suggests two things. Crime is strategic, and
not happening randomly, and loss is occurring within Peru rather than
the United States. We ran several robustness checks and confirmed that
our main results still hold.
Our research makes several important contributions. First, it shows
that an ordinary task (sending a piece of mail) that is important for
inexpensively transporting goods or information has hidden costs. This
implies that there still remain barriers to expand commerce that relies
on the mail sector to transport goods and services. Second, it
highlights the problems that developing countries face when trying to
solve inefficiencies through privatization of public services. Private
firms suffer the same asymmetric information that state-owned
enterprises do. Third, our research presents new evidence that crime is
not shared equally. The middle class seem to be taxed more heavily.
Finally, we confirm that crime is strategic and depends on expectations
and the probability of being caught.
Our results highlight several literatures. We examine crime in a
traditionally public sector activity: the delivery of mail (however, in
our setting, this service is privatized). In many developing countries,
corrupt behavior in the provision of public services is not only
widespread but can also create important inefficiencies and inequities
(Hunt and Laszlo 2008; Reinikka and Svensson 2005), such as in obtaining
a driver's license (Bertrand et al. 2007), state asset sales and
taxes (Fisman and Wang 2010; Fisman and Wei 2004), port transactions
(Sequeira and Djankov 2010), and traffic tickets (Fried, Lagunes, and
Venkataramani 2010). Public services may be privatized, yet our results
suggest that a private firm can face the same moral hazard problems that
a public firm would, as well as significant levels of corrupt and
strategic behavior.
The findings also touch on the large literature on crime. Theory
and casual observation would suggest that people may be affected
differentially by crime and that crime may be strategic (Becker 1968).
Therefore, neither the participation in illegal activities nor the
diseconomies caused by crime are expected to be uniformly distributed
across the population. Indeed, empirical studies have shown that crime
negatively affects economic activity (see Alesina and Perotti 1996;
Abadie and Gardezabal 2003; Barro 1991; Gaviria 2002; Pshisva and Suarez
2010, for some examples). And, there is ample evidence that people
respond to economic incentives when committing crimes. (4) Deterrence,
the risk of being caught, and social norms all seem to be important
factors in deciding whether to commit a crime or not. Our results add to
this literature by confirming crime to be strategic in this setting and
also providing evidence that not all income groups are equally affected.
The paper is organized as follows. The next section presents a
model of crime to provide a framework for our hypotheses. Section III
presents a description of the Peruvian postal system. Section IV
presents the experimental design and Section V the results. In Section
VI, we run robustness checks on our results, and in Section VII, we
check that our results are not due to response bias. Section VIII
concludes.
II. A MODEL OF CRIME
Our experiment allows us to look at crime in equilibrium. By
varying the content and information available on each piece of mail, we
can better understand the strategy of crime and who is affected the most
by it. Below we present a model of crime (borrowed heavily from Anwar
and Fang's 2006 model of discrimination). The aim of this section
is to derive comparative statistics from the model to build our
hypotheses on how the experimental manipulations will affect mail loss.
We start from the premise that the incidence of crime in
equilibrium is a function of the probability of being caught stealing
and the probability that the victim has something of value to steal. (5)
In the context of the mail sector, this implies that mail loss will be a
function of the probability that the postal worker will be fired or
punished if caught stealing mail and the probability, or the
expectation, that the sender of a piece of mail includes something of
value.
We assume that individuals choose the mail they send from a set of
possible types of mail, x [member of] {1,2,..., K}. x represents the
physical characteristics of the mail, i.e., the thickness of the
envelope or whether the mail is a letter, greeting card, manila
envelope, or package. Individuals also have to decide whether to place
something of value inside the mail. This is represented by a binary
variable d that takes the value of 1 if something of value is in the
mail and 0 otherwise. To model the expectation that there is something
of value (or the risk to the sender of mailing something of value), we
assume there is a signal [theta] associated with each piece of mail. The
mailman observes [theta]. For instance, mail with valuables might
require more packaging or might have some irregularities that are
observable to a careful handler. Signal [theta] is distributed according
to a continuous density function f([theta]|x, d) over interval [0, 1].
To address the fact that pieces of mail containing valuables might
be more likely to be found out, we assume that the ratio of f([theta]|x,
1) to f([theta]|x,0) is increasing in [theta]. As enough inspection
might uncover whether a piece of mail has something of value or not, we
assume that [lim.sub.[theta][right arrow]1] [f([theta]|x,
1)]/[f([theta]|x, 0)] = [infinity] and [lim.sub.[theta][right arrow]0]
[f([theta]|x, 1)]/[f([theta]|x, 0)] = 0. Finally, we denote by z the
social (nonphysical) characteristics associated with the mail. For
instance, z might be the neighborhood of the recipient, where the mail
was sent from or the relationship between the sender and the recipient.
A person with social characteristic z sends a piece of mail with
characteristic x and something of value with probability [[pi].sub.zx]
and without valuables ith probability 1 - [[pi].sub.zx.] Note that the
same distribution of signals will be interpreted differently depending
on [[pi].sub.zx]. In other words, the same evidence might be taken more
seriously in a population where [[pi].sub.zx] is higher.
The mailman decides whether to steal or deliver the piece of mail.
The expected return of stealing a piece of mail is
(1) P(x,z)Pr(d = 1|x, z, [theta]) - q(x,z)t.
P(x, z) represents the expected value of the content of the mail of
type x with characteristic z. q(x,z) represents the likelihood that the
mailman is caught stealing. This might happen because the post office
monitors more closely mail of type x with characteristic z or because
those with characteristic z are more likely to follow up on the mail
sent. The punishment a mailman faces when caught stealing or found to be
at fault is t > 0.
We can use Bayes' rule to determine that the probability that
a piece of mail contains something of value when signal [theta] is
observed is:
(2) Pr(d = 1|x, z, [theta]) = [f([theta]|x, 1)[[pi].sub.zx]]/
f([theta]|x, 1)[[pi].sub.zx] + f([theta]|x, 0)(1 - [[pi].sub.zx]).
The expected return to delivering the mail is normalized to 0.
Given the assumptions, we have that there is a unique value [??](x, z)
such that the expected payoff of committing a crime (Equation (1))
equals the payoffs of delivering the mail, or P(x, z) Pr(d = 1|x, z,
[??](x, z)) - q(x, z)t = 0. As Pr(d = 1 |x, z, [theta]) is increasing in
[theta], we have a risk that a neutral mailman will steal the mail if he
observes a signal [theta] such that [theta] [greater than or equal to]
[??](x, z).
Figure 1 shows the potential effect of a change in x on the
behavior of the mailman. (6) If a change in x affects only the
distribution of signals, we would expect that the conditional
probability that a piece of mail contains something of value will shift
for all values of [theta]. This can be seen in the figure where Pr(d =
1|[theta], x, z) shifts up as x changes to x'. All things constant,
this change in expectations will affect the critical value [??](x, z)
and, therefore, the probability of search (i.e., the likelihood a
mailman tampers with the mail). In the figure, the critical value
changes from that associated with point A to that associated with point
B, and the probability of search increases.
[FIGURE 1 OMITTED]
However, it is possible that a particular piece of mail of type x
is monitored more closely as well (e.g., certified mail) and so also
affects q(x, z). In this case, it is possible that enhanced incentives
to commit a crime are completely offset by a change in the probability
of being caught committing the crime (e.g., moving from point A to B to
C). For instance, a piece of mail sent by a family member is more likely
to contain something of value (a shift out of Pr(d = 1|[theta], x,z)),
but it is also more likely to be monitored (an increase in q(x, z)). Our
experimental design includes some changes in x that are unlikely to
affect q(x, z) (e.g., slightly increasing the thickness of the mail by
placing money inside).
Increases in [[pi].sub.zx] will also shift the curve Pr(d = 1|x, z,
[theta]) upwards. Changes in [[pi].sub.zx] might reflect the options
available to those with characteristics z. Those who have safer ways to
send valuables will be less likely to send them through the mail. For
example, wealthier households might be able to afford to send items
through costlier, yet more secure, services.
This model is partial in the sense that we assume that
[[pi].sub.zx] and q(z, x) are determined exogenously. While the model
can be modified to make these variables part of an equilibrium, our
experimental design does not require for them to be exogenous. Some of
the treatments are subtle enough so as to consider the effect to be
minimal. Finally, without explicitly modeling q(x, z), we cannot
determine if areas with higher levels of monitoring are also areas where
more crime is detected. For instance, a monopolist firm might find it
profitable to secure better services in some areas by increasing the
cost of committing a crime.
In our experiment, we manipulate information that we expect to
affect the benefits and costs of search, as we have outlined in the
model above. For example, by making the envelope slightly thicker with
something inserted inside (changing x), we expect mail loss to increase
as postal workers think the envelope signals something of value.
Similarly, by decreasing the social distance between the sender and the
recipient, such as having the sender be a family member (changing z), we
expect mail loss to increase. By sending mail to different
neighborhoods, we can see if mail loss decreases in wealthier
neighborhoods, relative to poorer neighborhoods, because we would expect
that the rich are more likely to complain if mail does not arrive
(probability of being caught rises). The effects, however, may be
nonlinear. While we expect service to be better as neighborhood income
rises, this will interact with the expectation that recipients will
receive something of value. People who live in wealthier neighborhoods
may be more likely to receive something of value, but they may also be
more likely to pay a higher price for more secure services (and
potentially avoid the post office altogether). Recipients in poor
neighborhoods may not be expected to receive anything of value, so
mailmen may not bother to look carefully. (7)
In sum, our hypotheses are as follows: (1) we expect mail loss to
increase as the probability of being caught stealing declines and the
probability of valuables increases; (2) the relationship across
neighborhoods may be nonlinear because the probability that something of
value is being sent will depend on the recipient's likelihood of
receiving something of value, the likelihood that the recipient has
alternative means for receiving valuables, and the probability of being
caught.
III. MAIL IN PERU
To better understand how and where crime and opportunistic behavior
may be happening when mail is sent from the United States to Peru, in
this section we describe the Peruvian postal system and how mail is
delivered.
The postal system in Peru is a private concession of the Peruvian
government and was privatized in 1991 (Legislative Decree 685, 1991).
The company does not have exclusive rights to deliver letters, as is the
case in the United States, but practically speaking, they are the only
provider of low-cost, nonpackage mail service in Peru. Indeed, they are
bound by law to provide service to isolated areas of the country. There
are alternative means for sending mail, including certified services
offered by the post office, but they are very expensive, costing about
100%-200% more, depending on the destination. Mail tracking services are
ten times more expensive than regular mail and the cost (in dollars) is
higher than the cost of a similar service in the United States.
Mail sent from overseas arrives in Peru at a central processing
facility located in the capital city of Lima. The facility sorts all
mail, domestic and international, and sends it to large district or
regional administration offices in Lima or other regions in Peru for
further sorting and delivery. There are also small post office branches
where further sorting of mail may occur. There are nine administration
offices in Lima and Callao with an average of 72 employees per office,
and there are 39 branches with an average of 2.5 employees per branch.
The administration offices employ mail workers, mail carriers, and
management. The post office branches employ workers and carriers.
Mail carriers are paid a fixed salary that is about two-thirds
above the minimum wage, and many have steady employment. However, about
one-fourth of the work force is not permanently employed. According to
the firm's annual reports, several of the main distribution offices
in Lima have installed cameras to monitor the handling of mail,
suggesting that mail is likely not lost there. (8)
Table 1 shows the structure of the labor force in the mail sector
in Lima and Callao in 2005 according to the company's publicly
available data. (9) The data is disaggregated by job category and
neighborhood income. (10) The table shows that the distribution of
people across tasks is remarkably similar across neighborhoods. Around
half of the work force deals with the delivery of mail, a quarter with
mail in post offices, and another quarter in administrative tasks.
However, we find that wages and years on the job are different across
neighborhoods. For instance, postmen and messengers earn less in
middle-income neighborhoods. This difference is significant when we
compare high-income neighborhoods to middle-income neighborhoods (t-test
p value =.0183). This difference amounts to a 5% reduction in earnings
across these two neighborhoods. Also, the earnings of postal employees
is slightly lower in low-income neighborhoods relative to other
neighborhoods (t-test p value =.0489). This difference is about 1%.
Finally, the number of years on the job of postmen and messengers in
richer neighborhoods is significantly larger than those in other
neighborhoods (t-test p value =.0129). Postmen and messengers in richer
neighborhoods have tenures that are 25% longer than those in other
neighborhoods. This is consistent with the presence of efficiency wages
in richer neighborhoods as a way to enforce higher quality in delivery
in these areas, but it is also consistent with the firm not paying
enough to retain its work force in more difficult areas or different
costs of living or local market conditions. In any case, the data
suggests that the firm discriminates across neighborhoods.
In terms of mail loss, there is anecdotal evidence that it does
occur. Casual observation of ten post offices in Lima revealed that
service is slow and cumbersome and that mailing a letter or package
requires interaction with unresponsive tellers. Customers willing to
send mail of intrinsic value went through great lengths to secure the
mail, repeatedly taping the mail closed after postage had been attached.
Conversations between customers standing in line tended to be about the
reliability and risks of the mail arriving at its destination. An
attendant in a small branch post office revealed to us that, in that
particular office, mail from abroad almost never arrived at its intended
final destination or, if it did, it would arrive tampered with.
Readers writing to the newspaper El Comercio (June 06, 2007, June
23, and July 07, 2007) expressed their frustration for not receiving
magazines and/or letters from family members. The newspaper reported
that many additional letters of similar tenor are received. Indeed,
commenting on a letter sent by a reader, the newspaper noted that many
informal magazine vendors offer magazines that show "strange
addresses." As there is a market for used magazines, there is no
way to verify if these magazines are stolen or not. Similar comments are
reproduced in ForosPeru.net, a Peruvian blogging site. Interestingly,
comments are not always negative. Some people state that they have never
experienced problems while others do. This suggests that problems might
not be generalized and therefore unlikely to occur in the central
office.
Finally, in terms of discipline or penalties for being caught
stealing, there is evidence to suggest that employees do get fired. The
company reports are not complete enough for us to determine if firings
are precisely as a result of stealing, however, there is a lot of
evidence of the company disciplining employees for not doing their job.
For example, according to the company's 2005 annual report, 25
employees were fired owing to major offenses. (11) Another 23 were
separated for arbitrary reasons while 20 either retired voluntarily or
ended their trial period at the company. In addition, there have been
lawsuits brought by former employees claiming to have been wrongfully
fired because of stealing. (12)
IV. EXPERIMENTAL DESIGN
We send envelopes from the United States to Peru through normal
mail services in both countries (U.S. Postal Service and the Peruvian
postal service, respectively). We use a list of residential addresses in
metropolitan Lima, Peru that are geographically representative of low-,
middle-, and high-income neighborhoods. A resident of each address is
the recipient of the envelope and reports to us if the envelope arrives
or not.
The 2 x 2 design we employ varies the contents of the envelope and
the sender's name. The contents of the envelope is a card and
either two $1 bills folded in half or no money. The sender's name
is either a foreign name (i.e., J. Tucker, M. Scott) or the same family
name as the recipient (i.e., M. Sosa, L. Cordova). (13) Varying the
sender's name allows us to test if names signal that something of
value is in the envelope (i.e., money). The design is outlined in Table
2 and includes the number of envelopes sent in each treatment. (14)
To get a valid estimate of crime, it is important that the envelope
look realistic and like something that would normally be sent in the
mail. So, we chose an opaque solid-colored envelope and card (of the
same color). The envelope looks like one that would be sent for a
birthday or other special occasion. Keeping with that idea, on the
inside of each card, we handwrite "Happy Birthday" or
"Feliz Cumpleanos"--depending on the return addressee's
name--and sign Josh or Mike or Marco or Luis. We do this because if the
card is stolen or opened, we want it to appear, to the postal worker,
like it was actually sent by the person whose name appears on the front
of the card. Figure 2 gives examples of two of the envelopes that were
sent to the same address in the course of the study. To preserve
confidentiality, we have blackened out the addresses of the sender and
recipient and the recipient's first name. The first envelope gives
an example of mail sent by a foreigner to a recipient and the second
envelope gives an example of mail from a family member to a recipient.
As can be seen in the Figure, the treatment manipulation of family
member is subtle.
Because the envelope is opaque, the greeting inside the card cannot
be seen. If the card contains money, this also cannot be seen, even if
held up to the light. One can feel that there is something in the
envelope because the folded two $1 bills make a very slight bump.
However, it is impossible to determine what exactly is in the envelope
without opening it. (15) But, there is a hint that the envelope contains
something other than the card. We chose this subtle manipulation so that
anyone looking for something to steal would need to pay careful
attention for signs that the envelope contained something that might be
worth stealing.
All envelopes have handwritten addresses, stamps for postage, and
an airmail stamp on the front of the envelope. There are two return
addresses in Atlanta and two in Washington, DC. All addresses are real
so that we could monitor if the card was returned to the United States
for any reason. The envelopes are glued shut, making it difficult to
steam open, reseal and deliver. Envelopes are always mailed from one of
two locations. Envelopes with a return address from Atlanta were mailed
from the main post office in downtown Atlanta, and envelopes with a
return address of Washington, DC were mailed from a post office mailbox
in Washington, DC. The color of the envelope, the return address, and
the handwriting on the envelope are randomized across the four
treatments. (16) Envelopes were sent during the period November
2006-November 2007. (17)
[FIGURE 2 OMITTED]
Mailboxes in Peru are secure and not exposed to theft from people
passing by on the street. Typically, mail is placed in a locked mail box
inside a locked gate or entryway. Or, it is placed under the door of the
locked residence. Mail is not left in post boxes on the streets, as is
the case in the United States.
To find recipient addresses, we tapped into two networks of people
who engage in research to recruit volunteers willing to receive the
cards and report to us. These people are trained in data analysis and
data collection and are aware of the importance of honest reporting. In
Section VII, we describe steps taken to insure data quality. The two
networks include people from a variety of demographic and income groups.
The important design element for us was that the addresses where the
mail was sent were geographically diverse. So, even though the mail
recipients might know one another, the addresses are dispersed across
locations. To minimize the number of addresses in the study for any
given post office, no more than four households were within a 1-km
radius of each other (i.e., 0.62 miles or the equivalent of ten blocks).
We mapped all the recipient addresses on a GIS updated street map of
Metropolitan Lima to minimize agglomeration and also to verify that the
addresses were correct and active. This ensures that nonarrival of mail
is not due to an incorrect address.
Recipients of the mail reported the arrival or nonarrival of each
envelope and kept any money if an envelope with money arrived. This
removed any incentive to misreport the contents. They were instructed to
not ask the mailman about the card or go to the post office to inquire.
After the envelope was put in the mail in the United States, we sent an
e-mail to the recipients telling them that an envelope was sent. They
were instructed to inform us when the mail arrived, who it was from, the
color, and the contents. They were not told ahead of time the
characteristics of the envelope or if the envelope contained money. This
was done to ensure no a priori bias in reporting and to allow us to
check that reporting was accurate. In addition to collecting information
on whether envelopes were received or not using several methods, we also
asked for replies by e-mail. This provided a simple way for us to check
the responsiveness of the recipients. We use these data later in Section
VII to evaluate potential nonresponse biases. We also had ten
supervisors who contacted various recipients to check if the envelopes
were received or not.
To compensate recipients for their time and help, at the end of the
experiment, we conducted a lottery with cash prizes for recipients who
reported. Recipients knew of the lottery before we began sending
envelopes. Also, to verify mail receipt responses, in December 2007 we
conducted a follow-up survey and collected more individual data on mail
recipients. This also allowed us to verify for a second time that
addresses were correct. All addresses were verified, and all previous
responses were confirmed. This gives us confidence that our data are
accurate. (18)
V. RESULTS
In this section, we explore the patterns of mail loss
geographically, across various demographic characteristics and across
our treatments. First, we turn to a description of the sample, then main
findings, and finally evidence of strategic behavior.
A. Sample Characteristics
Table 3 shows descriptive statistics on the individual and
geographical characteristics of the mail recipients. The sample is split
roughly half and half between male and female recipients. The
distribution of residents across low-, middle-, and high-income
neighborhoods is not even, with more people living in middle-income
neighborhoods. (19) Most recipients have a university education, are
married or living with their partner and have a family member that lives
in the United States. (20) This latter result is important as it makes
receipt of a card from the United States not seem strange and also
attests to the degree of mail that could potentially come from the
United States. Recipients have lived in their current residence for an
average of 16.5 years, and the nearest post office is 3 minutes away.
An important component of our experimental design, in addition to a
diverse and representative distribution of individual mail recipient
characteristics, is that the distribution of recipient addresses is
geographically dispersed across neighborhoods and post offices and is
representative of metropolitan Lima. Figure 3 shows the geographical
distribution of residents in our study. The residents cover the majority
of the city. There are fewer residents in some of the peri-urban areas
of the city, but the addresses are nicely distributed across
neighborhoods. This gives us observations across most areas of Lima and
confidence that our results apply to the larger, city-wide mail sector.
B. Main Findings
Turning to loss rates, we see that mail service in Lima is
inefficient and subject to crime. Table 4 shows loss rates overall and
by experimental treatments. Overall, 18% of all envelopes sent through
the mail never arrived at their destination. (21) Envelopes with money
were less likely to arrive than envelopes without money, so it does not
appear that mail loss is solely due to bad service. This confirms our
first hypothesis and hints more of criminal activity. Over 21% of
envelopes with money did not arrive, whereas 14.8% of envelopes without
money did not arrive. This 50% increase in loss is statistically
significant (t-test p value = .047).
[FIGURE 3 OMITTED]
How was mail lost across our four treatments? The bottom panel of
Table 4 shows loss rates by the contents of the envelope and the
sender's last name. Again, envelopes with money were more likely to
be lost than those without money, whether the sender's last name
was foreign or a family name. The difference between money and no money
envelopes for envelopes with a foreign sender is not significantly
larger, but the almost 10 percentage point difference for envelopes with
a family last name is (t-test for difference in mean p-value = .047).
Indeed, most loss happened for envelopes with money sent by a family
member. Almost one in four of those envelopes never arrived.
Across low-, middle-, and high-income neighborhoods, mail is lost
at different rates. Table 5 shows that residents in middle-income
neighborhoods lose mail at the highest rate, 20.4%, and those in
high-income neighborhoods lose mail at the lowest rate, 13.5%. The loss
rate in middle-income income neighborhoods is significantly larger than
in high-income neighborhoods. (22)
One might wonder if mail loss can be attributed to the Peruvian
mail service or to the U.S. Postal Service. The results in Table 5
suggest that loss is happening on the Peruvian side. While it may be
reasonable to think that envelopes with money might be lost on the U.S.
side, it is highly unlikely that the significantly different loss rates
we see across middle- and high-income neighborhoods is due to the U.S.
Postal Service. Such loss rates cannot exist without knowledge of
neighborhoods in Lima. The next section provides further evidence of
this.
Looking at the content of the envelopes, mail with money is
significantly more likely to be lost than that without money in
middle-income neighborhoods. In middle-income neighborhoods, the loss
rate of envelopes with money is over 10 percentage points larger than
for envelopes without money (t-test p value = .057). The loss rate in
poor neighborhoods is around 18% and is similar for envelopes with and
without money. High-income neighborhoods have an almost 7 percentage
point increase for envelopes with money, but this is not significantly
different (t-test p value = .232).
This pattern of loss confirms our second hypothesis and is
consistent with an expectation that the poor are not receiving valuables
by mail, so loss rates are no different with and without money. It seems
to be more a reflection of bad service than crime. Loss rates in
middle-income and high-income neighborhoods, however, are consistent
with the expectation that these populations have valuable items to
receive through the mail. Search is relatively larger in middle-income
neighborhoods, and this may reflect an expectation that people in
middle-income neighborhoods have few alternatives for receiving and
sending mail. The positive but insignificant effect of money in
envelopes in rich neighborhoods may reflect the expectation that
wealthier households have more alternatives.
The last two rows in Table 5 show that the pattern of loss across
neighborhoods is primarily driven by envelopes where the sender and
recipient share the same last name. Mail from a family member appears to
be given special attention as loss rates vary across neighborhoods. This
result is important, and consistent with our first hypothesis, because
it suggests that those handling the mail attribute a similar probability
of being caught across neighborhoods when disposing of mail sent by
nonfamily members. It is also consistent with the expectation that mail
from nonfamily members is unlikely to contain anything of value.
C. Evidence of Strategic Behavior
The main findings in the previous section show that mail with money
is lost more frequently and that crime is not distributed equally across
the population. The mechanism for loss seems to be that envelopes coming
from family members are scrutinized more closely than those from a
foreigner. In this section, we look more closely at the patterns of loss
and why they might exist.
Table 6 shows the joint effect of income and money on mail loss.
The numbers in the table also allow us to calculate a
difference-in-difference estimate of the effect of income on crime using
our treatment variables. The presence of money in envelopes sent by a
family member increases the rate of mail lost by 16.1 percentage points
(30.9%-14.8%) in middle-income neighborhoods and decreases it by 1.4
percentage points (18.2%-19.6%) in poor neighborhoods. In other words,
mail is 17.5 percentage points less likely to arrive when sent to
middle-income neighborhoods when there is suspicion of valuable content.
A comparison of the richer neighborhoods and poorer neighborhoods gives
a similar estimate (15.0). This increase in the likelihood of loss from
poor to middle and poor to rich neighborhoods could be due to
expectations that something of value might be sent in the mail or the
perceived smaller risk of being caught.
In order to test whether the differential impact of crime across
income groups is explained by differences in the quality of service and
not expectations, in Table 7, we compare loss rates across types of
envelopes using regression analysis. This analysis assumes that quality
of service by neighborhood affects loss rates uniformly independent of
who sends the mail. The results show that both expectations that the
envelope contains something of value (money, no money) and the
probability of being caught (which varies by the socioeconomic level of
the neighborhood) explain loss rates. The table presents results from
fixed-effects linear probability (OLS) regressions of loss on whether
the envelope contained money, if it was sent by a family member, and
interactions with neighborhood income (percent classified as poor). The
dependent variable equals 1 if the mail did not arrive at its
destination and 0 otherwise. (23)
It is important to note that the analysis includes recipient fixed
effects, to control for any idiosyncrasies of our small sample, and our
main results still hold. (24) The first column in Table 7 illustrates
this. Envelopes with money are more likely to be lost. The second and
third columns show results for the subsamples of mail sent by a family
member and by a foreigner. By looking only at envelopes from family
members or from foreigners, we attempt to keep constant the expected
cost of committing a crime, so that we can focus on expectations that
the envelope contains something of value. (25) We see that the effect of
money for envelopes coming from family members, controlling for
recipient fixed effects, is stronger as neighborhoods become wealthier.
There is no significant effect of envelopes coming from foreigners. This
confirms the results in Table 6 and suggests that expectations that an
envelope contains something of value matter. This is so because we
expect not only that both the rich and the poor care about receiving
mail but that the cost of being caught stealing should not decrease with
the wealth of the neighborhood. The fact that envelopes with money are
lost at a higher rate as neighborhood income goes up says there is an
expectation that mail from family members contains something of value
when sent to wealthier neighborhoods.
The results in the fourth and fifth columns of Table 7 of the
regressions on money and no-money envelopes are also indicative of
incentives. The previous regression on family envelopes suggests that
family envelopes sent to richer neighborhoods are perceived to contain
valuables. So, we would expect that the effect of family on money
envelopes to be stronger in richer neighborhoods, not weaker. The fact
that we do not find this suggests that there is a countervailing force
limiting the incentive to commit a crime. Since larger expected costs of
being caught reduce the incentives to steal, this result is consistent
with the belief that the probability of being caught stealing is larger
in richer neighborhoods. (26) This result is also consistent with our
model predictions that changes in social characteristics, z, can affect
both the probability of something of value being there in the mail and
the probability of being caught stealing.
All together, these results are consistent with a story that, as
neighborhood income rises, expectations that valuables are sent through
the mail increase and interact with an increased probability of being
caught.
VI. ROBUSTNESS CHECKS ON MAIN FINDINGS
This section presents regression analysis of mail loss rates to
test the robustness of our main results to omitted variables and
specification assumptions. We check that our results are not due to
recipient-level fixed effects, our definition of neighborhood grouping
by income, correlation between neighborhood income and the way mail is
processed, misreporting, or other socioeconomic variables. Table 8
presents OLS regressions of mail loss on covariates, including treatment
variables, nonlinear effects of neighborhood income, distance to the
closest post office branch, whether there is an administrative center
located in the neighborhood, number of post office branches in the
neighborhood, and the time effect of when the mail was delivered. (27)
The first column in Table 8 (and the first column in Table 7)
confirms our main findings. Envelopes containing money are more likely
to get lost. The results in Table 7 confirm that this holds with
recipient-level fixed effects, and the results in Table 8 confirm that
it holds at the same time as does the nonlinear relationship with
neighborhood income (percent classified as poor). This latter result
still remains even, controlling for the manner in which mail is
processed across neighborhoods. The results are intuitive. Mail sent to
neighborhoods with an administrative center is lost more frequently
because, presumably, there are many more employees handling the mail and
this helps to dissipate responsibility. Neighborhoods with more post
office branches lose less mail because it is easier to identify
responsibility.
In Table 8, columns 2 and 3 show that the effect of neighborhood
income is slightly stronger for the envelopes with money, but this is
not significant. The regressions dividing the population receiving
envelopes from family and nonfamily members (columns 4 and 5) confirm
that it is the envelopes with money coming from family members that are
more likely to get lost. Finally, the patterns of lost mail across all
regressions do not seem to respond to the proximity of post offices. The
number of minutes it takes to get to the closest post office is not
significant. (28)
The results reported above also eliminate two alternative
explanations for higher loss of envelopes with money: systematic
misreporting by recipients and misreporting envelopes with money. First,
since the results hold when controlling for recipient fixed effects
(Table 7), individual misreporting is not causing our main result that
envelopes with money are lost more. Second, since the recipients did not
know ahead of time what kind of envelope was sent, there is no reason to
believe that the money effect is due to people not paying attention to
the money envelopes more than no-money envelopes. Also, and this is
crucial, they could keep the money and therefore did not need to say it
was lost.
Finally, the nonlinear effect of neighborhood income on overall
loss rates is not what one would expect if it were only due to
differential shirking by neighborhoods. Table 1 showed that years on the
job for postmen and messengers were lower in low- and middle-income
neighborhoods. If years on the job reflects the consequences of shirking
(i.e., getting fired), we would expect loss rates to be similar in low-
and middle-income neighborhoods, not different.
The results in Table 8 are also robust to learning by the mail
carrier and the inclusion of other socioeconomic information, such as
family size, time in residence, and marital status. (29) Overall,
envelopes with money are more likely to be lost and neighborhood income
is nonlinearly related to loss rates. Also, the fact that money
envelopes are more likely to be lost, even when controlling for
recipient fixed effects, is strong evidence that this result is robust.
VII. ROBUSTNESS CHECKS ON REPORTING BIAS
This section presents additional evidence that the results in the
paper are not due to misreporting by the mail recipients. As discussed
in the data section, recipients in our experiment were recruited among
people involved in field research. They are trained in survey methods
and are aware of the problems associated with misreporting. In addition
we had ten monitors coordinating the collection of data as a way to keep
close vigilance on the process. Despite all this, it is always possible
that mail recipients mistakenly reported lost mail, or worse, purposely
reported losses when they did not exist. While our post-experiment
survey gives us confidence that the envelopes received were indeed
received, it is harder to check if reported losses did indeed occur.
This section presents a series of tests that suggest that the patterns
of lost data are not biased and therefore the main results of the paper
are not due to misreporting.
Table 9 checks the significance of the effect of money on loss
rates. These are OLS regressions of mail loss on the treatment variable
and response time. (30) The estimations exploit the fact that we had a
measure of recipients' responsiveness with the time they took to
respond to and check their e-mail. One hypothesis is that less
responsive recipients or those unable to respond quickly might be more
likely to report missing mail owing to distraction. Since we sent an
email each time the envelopes were sent out, it also allows us to see if
our results are associated in some way to periods in which subjects were
busier or more distracted. We see that, even when controlling for
response time, our results still hold. Columns 1-3 show the rate of mail
loss as a function of whether it contained money and whether the subject
responded to the e-mail or not. As expected, recipients who do not send
an e-mail response are less likely to receive the envelope. However, we
see that envelopes with money are still more likely to be lost and that
this result is explained mainly by the loss of envelopes sent by family
members (coefficients are similar to those reported in Table 8). Columns
4-6 show the rate of mail loss as a function of money and the time it
took subjects to respond to e-mails. Again, we find that money envelopes
are lost more frequently and that this is significant among the
envelopes sent by family members.
Table 10 checks the significance of the effect of money on loss
rates. The estimations exploit the variation in time to confirm receipt
of the mail to see if those confirming relatively later were driving the
results. Confirmation times are measured from the day the envelopes were
mailed from the United States and take advantage of the fact that those
receiving the mail were not equally easy to reach. This is so because
some recipients had to travel for work or because they had no Internet
connection at home or work (and therefore checking email regularly was
less convenient). (31) Table 10 makes the assumption that the longer the
time the mail is not reported the higher the likelihood that reasons
other than crime or bad service explain the results. Columns 1-3
estimate the effect of money after all observations which confirmed
receipt of the mail after 4 weeks are dropped. The regressions show that
envelopes with money are still more likely to be lost in this restricted
sample. The regressions also confirm that the effect is only significant
in the envelopes from family members. Columns 4-6 repeat the analysis
under the strong assumption that later reports were false positives
(i.e., made-up data). The regressions show that the effect of money on
loss rates persists. In all, these regressions show that if misreporting
is correlated with time to confirm, misreporting is not serious enough
to eliminate our main results.
Finally, we test for random misreporting. These estimates assume
that some of the reported losses were mistakes and that these mistakes
were random. To do this, we randomly select some reported losses and
change them to no loss. This amounts to assuming that reported losses
are measured with error. Table 11 presents the estimates of 10,000
repetitions of linear regressions with recipients' fixed effects.
Columns 1-3 report results when 20% of the reported losses are assumed
to be mistakes and columns 4-6 report results when 30% of the reported
losses are assumed to be mistakes. As expected, the estimated parameters
are smaller. However, neither the significance nor direction of the
results are affected. Money envelopes are lost more frequently, and the
losses are significant in the envelopes coming from family members.
While none of our robustness checks is proof that no misreporting
occurred, the checks should dispel concerns that any misreporting is
large enough as to invalidate our results.
VIII. CONCLUSIONS
Using a simple and novel field experiment that opens the door for
opportunistic behavior, we examine strategy and crime in the mail sector
in Lima, Peru. We hypothesize that the very nature of mail delivery
gives an opportunity to those who handle the mail to "lose"
mail if it is beneficial to do so. Our design allows us to differentiate
poor service from targeted crime and to investigate what information is
pertinent to committing crimes and who suffers the most from it.
We have several key findings. First, loss rates are very high. Over
18% of all mail sent never arrived at its destination. These rates are
huge and would imply large barriers for the development of trade that
relies on mail services. Second, these loss rates are partially
explained by poor service but not completely. Envelopes containing money
were 50% more likely to be lost than those without money. So, mail loss
is not random and hints at strategic behavior. Third, when the
sender's last name matched the recipient's last name, the mail
was almost twice as likely to be lost if it contained money. Clearly,
those who handle the mail are looking for clues that might suggest that
an envelope holds something of value. Fourth, middle-income
neighborhoods suffer the highest loss rates and high-income
neighborhoods suffer the lowest. This result (and the previous) lends
support for the crime occurring in Peru rather than the United States
since it would require the U.S. Postal Service to know which
neighborhoods were rich or poor.
Finally, the patterns of behavior we observe are consistent with
expectations that the recipient could receive something of value and the
perceived probability of being caught stealing. This results in a
nonlinear effect of neighborhood income on loss. Looking only at mail
from family members, we see that loss increases as neighborhood income
rises if the mail contains money, suggesting that there is an
expectation that residents in middle-income and wealthy neighborhoods
may have something valuable to steal. The magnitude of loss is higher in
middle-income neighborhoods though, supporting the notion that the
probability of being caught is more likely in wealthy neighborhoods.
That is, while it is more likely to get something of value when stealing
from middle- and high-income residents, stealing in the highest-income
neighborhoods may be tempered by a higher likelihood of being caught.
Middle-income residents then suffer a compounding effect that results in
the highest loss rates.
Put together our results suggest a model of behavior where those
who handle the mail are looking for items of value to steal, but they
take into account the likelihood that valuables are being sent in the
mail and the probability of being caught stealing. Moreover, and
importantly, crime is not independent of the neighborhood's
characteristics.
While our study cannot speak for the presence of large
inefficiencies in all the sectors dealing with the transaction of goods
and services, it highlights the large barriers to market development
that developing economies face. Certainly, our results suggest barriers
are high for e-commerce to emerge since it relies on the mail sector to
transport goods. Other, more secure, transportation services are
available, but they can be two to five times more expensive. Also, while
the cost of losing a piece of mail may be low (depending on what was
lost), relative to other crimes, it is the unreliability of the service
that has a larger cost because it hampers market development.
The sophistication in criminal activity found in our research
suggests that there is a need to design monitoring mechanisms (e.g.,
security cameras) and appropriate incentives to minimize strategic
behavior. Indeed, in a field experiment on auditing, Nagin et al. (2002)
find that an increased perception of monitoring can reduce shirking.
Cameras would work in mail sorting facilities (where they are currently
installed at some locations), but it is more difficult and costly to
monitor mail delivery on foot. Random audits of the nature of our
experimental design could be an effective mechanism for identifying
problem neighborhoods and facilities.
Our study further shows that private firms providing public
services also face incentive problems due to moral hazard in the same
way state-owned enterprises do. The nature of the good seems to be as
important as the nature of ownership. Incentive problems may well
prevent markets from developing. How severe these problems are relative
to other factors, such as inefficiencies in governance or lack of
competition or infrastructure, remains to be studied.
ABBREVIATION
OLS: Ordinary Least Squares
doi: 10.1111/ecin.12046
Online Early publication October 17, 2013
REFERENCES
Abadie, Alberto. and Javier Gardeazabal. "The Economic Costs
of Conflict: A Case Study of the Basque Country." American Economic
Review. 93, 2003, 113-32.
Alesina, Alberto, and Roberto Perotti. "Income Distribution,
Political Instability and Investment." European Economic Review,
40, 1996, 1203-28.
Anwar, S., and H. Fang. "An Alternative Test of Racial
Profiling in Motor Vehicle Searches: Theory and Evidence." American
Economic Review, 96, 2006, 127-51.
Armantier, Olivier, and Amadou Boly. "Can Corruption Be
Studied in the Lab? Comparing a Field and a Lab Experiment."
Working Paper, Federal Reserve Bank of New York, 2008.
Barro, Robert J. "Economic Growth in a Cross Section of
Countries." Quarterly Journal of Economics, 106, 1991, 407-43.
Becker, Gary. "Crime and Punishment: An Economic
Approach." Journal of Political Economy, 76, 1968, 169-217.
Bertrand, Marianne, Simeon Djankov, Rema Hanna, Sendhil
Mullainathan. "Obtaining a Driving License in India: An
Experimental Approach to Studying Corruption." Quarterly Journal of
Economics, 122(4), 2007, 1639-76.
Corman, Hope, and Naci Mocan. "Carrots, Sticks, and Broken
Windows." Journal of Law and Economics, 2005, 235-66.
Di Tella, Rafael, and Ernesto Schargrodsky. "The Role of Wages
and Auditing during a Crackdown on Corruption in the City of Buenos
Aires." Journal of Law and Economics, 46(1), 2003, 269-92.
--. "Do Police Reduce Crime? Estimates Using the Allocation of
Police Forces after a Terrorist Attack." American Economic Review,
94(1), 2004, 115-33.
Duggan, Mark, and Steven Levitt. "Winning Isn't
Everything: Corruption in Sumo Wrestling." American Economic
Review, 92(5), 2002, 1594-605.
Erlich, I. "Participation in Illegitimate Activities: A
Theoretical and Empirical Investigation.'" Journal of
Political Economy, 81, 1973, 521-65.
Fisman, Raymond, and Edward Miguel. "Corruption, Norms, and
Legal Enforcement: Evidence from Diplomatic Parking Tickets."
Journal of Political Economy, 115. 2007, 1020-48.
Fisman, Raymond, and Yongxiang Wang. "Corruption in State
Asset Sales-Evidence from China." Working Paper, Columbia
University, 2010.
Fisman, Raymond, and Shang-jin Wei. "Tax Rates and Tax
Evasion: Evidence from 'Missing Imports' in China."
Journal of Political Economy, 112, 2004, 471-97.
Fried, Brian, Paul Lagunes, and Atheendar Venkataramani.
"Corruption and Inequality at the Crossroads: A Multimethod Study
of Bribery and Discrimination in Latin America." Latin American
Research Review, 45(1), 2010, 76-97.
Gaviria, Alejandro. "Assessing the Effects of Corruption and
Crime on Firm Performance: Evidence from Latin America." Emerging
Markets Review, 3, 2002, 245-68.
Glaeser, Edward L., Bruce Sacerdote, and Jose A. Scheinkman.
"Crime and Social Interactions." Quarterly Journal of
Economics, 111, 1996, 507-48.
Hsieh, Chang-Tai, and Enrico Moretti. "Did Iraq Cheat the
United Nations? Underpricing, Bribes, and the Oil for Food
Program." Quarterly Journal of Economics, 121, 2006, 1283-309.
Hunt, Jennifer, and Sonia Laszlo. "Is Bribery Really
Regressive? Bribery's Costs, Benefits and Mechanisms," Working
Paper, McGill University, 2008.
Jacob, Brian. and Lars Lefgren. "Are Idle Hands the
Devil's Workshop? Incapacitation, Concentration, and Juvenile
Crime." American Economic Review, 93, 2003, 1560-77.
Levitt, Steven. "Using Electoral Cycles in Police Hiring to
Estimate the Effect of Police on Crime." American Economic Review,
87(3), 1997, 270-90.
Nagin, Daniel, James Rebitzer, Seth Sanders, and Lowell Taylor.
"Monitoring, Motivation, and Management: The Determinants of
Opportunistic Behavior in a Field Experiment." American Economic
Review, 92(4), 2002, 850-73.
Olken, Benjamin. "Monitoring Corruption: Evidence from a Field
Experiment in Indonesia." Journal of Political Economy, 115(2),
2007, 200-49.
Olken, Benjamin, and Patrick Barron. "The Simple Economics of
Extortion: Evidence from Trucking in Aceh." Journal of Political
Economy, 117(3), 2009, 417-52.
Pshisva, Rony, and Gustavo Suarez. "Capital Crimes:
Kidnappings and Corporate Investment in Colombia," in The Economics
of Crime: Lessons for and from Latin America, edited by Rafael Di Tella,
Sebastian Edwards, and Ernesto Schargrodsky. Chicago: University of
Chicago Press, 2010, 63-100.
Reinikka, Ritva, and Jakob Svensson. "Local Capture: Evidence
from a Central Government Transfer Program in Uganda." Quarterly
Journal of Economics, 119, 2004, 679-705.
--. "Fighting Corruption to Improve Schooling: Evidence from a
Newspaper Campaign in Uganda." Journal of the European Economic
Association, 3, 2005, 259-67.
Sequeira, Sandra, and Simeon Djankov. "An Empirical Study of
Corruption in Ports," Working Paper, London School of Economics,
2010.
MARCO CASTILLO, RAGAN PETRIE, MAXIMO TORERO and ANGELINO VICEISZA *
* We thank David Solis for conducting the follow-up survey, Cesar
Ciudad for coordinating the mail recipients, and Maribel Elias for
creating the GIS maps. Seminar participants at University of Maryland,
Iowa State University, ICES-George Mason University, Georgia Institute
of Technology, Virginia Commonwealth University, the World Bank, the
Workshop on Economics Experiments in Developing Countries at CIRANO in
Montreal, and the North American Economic Science Association Meetings
gave helpful comments.
Castillo: Associate Professor, Interdisciplinary Center for
Economic Science (ICES) and Department of Economics, George Mason
University, Fairfax, VA 22030. Phone 1-703-993-4238, Fax 1-703-993-4831,
E-mail
[email protected]
Petrie: Associate Professor, Interdisciplinary Center for Economic
Science (ICES) and Department of Economics, George Mason University,
Fairfax, VA 22030. Phone 1-703-993-4842, Fax 1-703-993-4831, E-mail
[email protected]
Torero: Division Director, Markets, Trade and Institutions,
International Food Policy Research Institute, Washington, DC 20006.
Phone 1-202-862-5600, Fax 1-202-467-4439, E-mail
[email protected]
Viceisza: Assistant Professor, Department of Economics, Spelman
College, Atlanta, GA 30314. Phone 1-404-270-6055, E-mail
[email protected]
(1.) Sending money through the Peruvian postal service and the U.S.
Postal Service is not illegal, although it is not advised.
(2.) Fried, Lagunes, and Venkataramani (2010) look at bribery by
traffic police and how it varies by the income class of drivers.
(3.) Compared to the less than 0.5% of mail reported lost in the
United States or the United Kingdom, this is very large. Note that loss
rates in the United States and the United Kingdom are for reported mail
lost. This will underestimate the problem if not all lost mail is
reported. Our experimental measure is for all mail lost that should have
arrived at a destination.
(4.) For examples of empirical and experimental work, see Erlich
(1973), Levitt (1997), Duggan and Levitt (2002), Glaeser, Sacerdote, and
Scheinkman (1996), Jacob and Lefgren (2003), Di Tella and Schargrodsky
(2003, 2004), Olken (2007), Fisman and Miguel (2007), Reinikka and
Svensson (2004), Armantier and Boly (2008), Hsieh and Moretti (2006),
Olken and Barron (2009), Corman and Mocan (2005), among others.
(5.) Becker (1968) and Erlich (1973) present detailed models of
decision making by individuals considering committing a crime. In the
context of their models, the interaction between a mailman and a
customer can be thought of as a zero-sum game. If a person sends
valuables with probability one and there is moral hazard, mail will be
certainly stolen. If the mailman never steals, a customer might feel
safe sending valuables in the mail. In equilibrium, one would expect
that those customers that have a larger marginal benefit of using the
mail to send valuables will face a larger average level of crime.
(6.) The figure graphs the two components of Equation (1): the
probability that a piece of mail contains something of value and the
probability of being caught stealing divided by the expected value of
the piece of mail.
(7.) A model of crime has enough degrees of freedom as to make it
difficult to determine a priori the effect of income on public services
crime. Our experiments provide data that might allow us to determine
regularities, but not necessarily prove or disprove a theory.
(8.) See http://www.serpost.com.pe/transparencia/
DocumentacionTransparencia/Docs2009/Informacion
Adicional/MemoriaAnual/MemoriaAnual2009.pdf
(9.) See http://www.serpost.com.pe/transparencia/
DocumentacionTransparencia/Docs2005/Informacion Adicional/Memoria
Anual/MemoriaAnual2005.pdf
(10.) We define low-income neighborhoods as ones where the percent
of the population considered poor is 30% or higher. Middle-income
neighborhoods are those where the percentage is between 10% and 30%, and
high-income neighborhoods are those where the percentage is less than
10%. The proportion of people living in poverty is from the closest
available poverty map to the time of the experiment (2006) calculated by
the Peruvian Ministry of Economy and Finance using expenditure surveys
from the Peruvian Institute of Statistics.
(11.) Report citation is listed in footnote 9.
(12.) The Peruvian Constitutional Tribunal (the equivalent of the
U.S. Supreme Court) has heard ten cases brought by ex-workers of the
Peruvian mail service asking to be reinstated. Seven out of ten said
they were accused of grave misdemeanors and two out of the seven
explicitly mentioned stealing.
(13.) In South America, including Peru, everyone has two last
names. The first is the last name from the father and the second is the
last name of the mother. We use the first last name. We only test for
family name versus a foreign name, not mismatched Hispanic names. This
is another interesting line of research, but not pursued in this paper.
(14.) There are not an equal number of observations in each cell
for two reasons. First, some households moved before the experiment was
complete, so they did not receive all four treatments. Second, because
of a clerical error, 13 households received four letters, two with money
and two without, but the sender's last name was not equally divided
between family and foreign. All main results hold if these households
are dropped. The estimates are of equal magnitude, but some are less
precise due to fewer observations.
(15.) We could very well have placed folded pieces of paper in the
envelope instead of money, but we want the envelopes and contents to be
realistic, especially in case the envelope is lost or stolen. Also, it
is important to note that the envelope is only thicker and still easily
slips through a mail slot or a sorting machine, as would the envelope
without money.
(16.) We did this to insure that each envelope sent to a household
by a different person was indeed handwritten by a different person.
(17.) Mailings were sent at various dates in November, June, July,
and August.
(18.) We asked the recipient if they received an envelope during a
certain period of time and asked them to report the return address and
color of the envelope. Recipients were able to correctly confirm reports
from 4 to 5 months earlier.
(19.) See footnote 10 for definitions of neighborhood income. We
show later that our results are robust to other definitions of economic
status.
(20.) This is a reflection of our sample procedure; while
recipients resided across many neighborhoods, they all were contacted
through research institutions. Also, recipients regularly received mail,
so the receipt of four envelopes over the course of a year was not
unusual.
(21.) For the mail that arrived, the average arrival time of a
piece of mail was 7.2 days (4.2 s.d.). In high-income neighborhoods, the
arrival time is 7.2 days (4.3 s.d.). In middle-income neighborhoods, it
is 7.3 days (4.6 s.d.) and 6.9 days (3.7 s.d.) in poorer-income
neighborhoods. The differences across neighborhoods in arrival times are
not significantly different.
(22.) t-tests yield p values of .696 comparing low- to
middle-income loss rates, .080 comparing middle- to high-income, and
.189 comparing low- to high-income loss rates.
(23.) The same results hold with a fixed-effects Logit model. We
report OLS results because the regressions do not drop observations and
give us more robust results.
(24.) That is, our results are not due to idiosyncrasies in
recipient reporting, postal carriers, or household location (e.g., at
the end of the postal carrier's route).
(25.) This is a difference-in-difference estimate on the net
benefit (expectation of something of value less the expected cost of
being caught) of an envelope with and without money across
neighborhoods. Or, re-written, this is the net expectation of
encountering something of value less the net cost of being caught taking
something of value (as in Equation (1) in Section II). By splitting the
sample into envelopes from family members and from foreigners, we can
control for the net cost of being caught across neighborhoods. We expect
this to be constant, since otherwise this would say that the rich care
less about receiving mail with money. So, any significant effect on
money or money interacted with neighborhood will be due to expectations.
(26.) As described in the section on the Mail Sector in Peru,
postal workers in richer neighborhoods earn higher wages and have longer
job tenures. This efficiency wage story would be consistent with a
higher probability of being caught stealing. However, with the evidence
we have from naturally occurring data as in Table 1, we only observe
equilibrium outcomes, so we cannot say with certainty that this is the
case. Postal workers in richer neighborhoods might also be more honest.
(27.) The results hold if specified as Logit regressions and are
robust to autocorrelation.
(28.) The number of minutes to the closest post office is based on
an accessibility model which calculates the least cost path surface
(based on time) from any place, using GIS. The accessibility measure
uses three different levels of roads with different speeds of movement.
(29.) Learning is tested by the inclusion of a lagged term for loss
or a dummy variable that equals one if the first envelope did not arrive
(whether it contained cash or not). We also tested for whether the
likelihood of loss varied by whether the envelope was the first, second,
third, or fourth mailing and for the correlation between lost mailings.
There is a slightly higher likelihood a mailing is lost if the first one
was stolen. The most important outcome of these checks is that our main
results still hold. These results are not included in the paper but are
available from the authors.
(30.) Note that all results of the effect of money on loss rates in
Tables 9 and 10 hold with Logit fixed-effects regressions.
(31.) Part of the research was done concurrently with the 2007
National Census which required several of our recipients to be away from
home.
TABLE 1
Labor Force Characteristics across Neighborhoods
Labor Category
Postmen and Postal
Neighborhood Messengers Employees Other
High income Monthly income 840.9 957.3 1032.6
(soles)
Years on the job 6.5 10.0 6.0
Number 128 (51%,) 63 (25%) 59 (24%)
Middle income Monthly income 798.6 959.8 1061.4
(soles)
Years on the job 5.2 10.1 5.9
Number 124 (43%) 90(32%) 72 (25%)
Low income Monthly income 829.1 947.1 1142.5
(soles)
Years on the job 5.2 9.6 6.5
Number 60 (46%) 44 (34%) 26(20%)
TABLE 2
Experimental Design
Contents of Envelope
Sender Last Name Money No Money
Foreign n = 136 n = 131
Family n = 135 n = 139
TABLE 3
Descriptive Statistics of Sample Recipients
Percent SD # Obs
Male recipients 47.5 141
Low income 34.8 49
Middle income 39.7 56
High income 25.5 36
Age (mean, years) 37.2 10.1 124
University education 57.4 136
Married or cohabitating 44.1 136
Family size (mean. number) 4.1 1.5 124
Family in United States 47.1 136
Time in residence (mean, years) 16.5 12.8 124
Minutes to post office (mean) 3.0 8.4 140
Note: Some variables have missing values because of
survey nonresponse.
TABLE 4
Average Loss Rates Overall and by
Experimental Treatment (%)
Number
Overall 18.1 98
(1.7)
Money 21.4 58
(2.5)
No money 14.8 40
(2.2)
Experimental
Treatments
Contents of Difference
Envelope in Mean
Money No Money p value
Foreign 19.8 16.0 4181
(3.4) (3.2)
Family 23.0 13.7 467
(3.6) (2.9)
Difference in mean
p value 5343 5868
Notes: Standard error of the mean in parentheses. Chi-square
distribution test validates the null hypothesis that the
loss rates across treatments are the same.
TABLE 5
Average Loss Rates by Income Groups (%)
Low Middle High
Income Income Income
Overall 18.9 20.4 13.5
Money 19.8 25.7 16.9
No money 18.0 15.3 10.0
Foreign sender name 18.9 18.3 16.4
Family sender name 18.9 22.4 10.3
TABLE 6
Average Loss Rates by Money, Income
Groups, and Sender Name (%)
Envelope with Money
Low Middle High
Income Income Income
Foreign sender name 21.3 20.4 17.1
Family sender name 18.2 30.9 16.7
Envelopes with No Money
Low Middle High
Income Income Income
Foreign sender name 16.3 16.0 15.8
Family sender name 19.6 14.8 3.1
TABLE 7
Evidence of Strategic Behavior Probability of Mail Loss by Sender
Name and Envelope Contents OLS Fixed-Effects Regressions
(1) (2) (3)
Variables All Family Foreign
Money 0.062 ** 0.197 *** 0.035
(0.028) (0.070) (0.063)
Family 0.000
(0.028)
Money x percent poor -0.005 * -0.001
(0.003) (0.002)
Family x percent poor
Constant 0.150 *** (1.139 *** 0.176 ***
(0.024) (0.029) (0.027)
Recipient fixed effects Yes Yes Yes
Observations 541 274 267
[R.sup.2] 0.012 59 0.003
(4) (5)
Variables Money No Money
Money
Family 0.050 -0.081
(0.060) (0.066)
Money x percent poor
Family x percent poor -0.001 0.003
(0.002) (0.003)
Constant 0.199 *** 0.156 ***
(0.025) (0.027)
Recipient fixed effects Yes Yes
Observations 271 270
[R.sup.2] 7 0.013
Notes: Standard errors in parentheses. Dependent variable is Mail
Loss (=1 if mail never arrived and =0 if mail arrived). Independent
variables: Money = I if envelope contained money, Family = I if
sender's last name was the same as recipient's, Percent Poor =
percent of population in neighborhood living in poverty.
* p<.10, ** p<.05, *** p<.01.
TABLE 8
Robustness Check on Main Findings Probability of Mail Loss by
Envelope Contents and Sender Name OLS Regressions
(1) (2) (3)
Variables All Money No Money
Money 0.065 *
(0.034)
Family 0.006 0.031 -0.020
(0.033) (0.049) (0.043)
Percent poor 0.0111, 0.013 0.009
(0.005) (0.008) (0.007)
Percent poor squared -0.000 * 0.000 0.000
(0.000) (0.000) (0.000)
Minutes to post office 0.001 0.004 -0.001
(0.002) (0.003) (0.003)
Administrative center 0.115 *** 0.133 ** 0.096 **
(0.035) (0.053) (0.047)
Number of post office branches -0.029 ** -0.023 -0.034 **
(0.012) (0.018) (0.016)
Mailing number (1-4) 0.002 -0.003 0.005
(0.015) (0.023) (0.021)
Constant 0.041 0.086 0.062
(0.068) (0.086) (0.094)
Observations .537 .269 .268
[R.sup.2] .390 .040 .037
(4) (5)
Variables Family Foreign
Money 0.100 ** 0.029
(0.050) (0.048)
Family
Percent poor 0.013 * 0.010
(0.008) (0.007)
Percent poor squared 0.000 0.000
(0.000) (0.000)
Minutes to post office 0.000 0.003
(0.003) (0.003)
Administrative center 0.075 0.157 ***
(0.050) (0.050)
Number of post office branches -0.033 * -0.023
(0.017) (0.017)
Mailing number (1-4) 0.011 -0.011
(0.022) (0.022)
Constant 0.012 0.078
(0.099) (0.090)
Observations .272 .265
[R.sup.2] .041 .050
Notes: Standard errors in parentheses. Dependent variable is Mail
Loss (= 1 if mail never arrived and = 0 if mail arrived). Independent
variables: Money = 1 if envelope contained money, Family = 1 if
sender's last name was the same as recipient's, Percent Poor =
percent of population in neighborhood living in poverty. Minutes to
post office = number of minutes from residence to closest post
office. Administrative Center = 1 if an administrative center is
located in neighborhood, Number of Post Office Branches = number of
branches in neighborhood, Mailing Number = 1 if first mailing, = 2 if
second mailing, etc.
* p<.10. ** p<.05, *** p<.01.
TABLE 9
Robustness Check on Reporting Probability of Mail Loss Controlling
for Response Time OLS Fixed-Effects Regressions
(1) (2) (3)
Variables All Family Foreign
Money 0.073 *** 0.102 ** 0.017
(0.028) (0.042) (0.038)
Responded by e-mail in a week
Responded by e-mail in 2 weeks
Responded by e-mail in 3 or more
weeks
Responded to e-mail -0.121 *** -0.130 * -0.089
(0.037) (0.069) (0.056)
Constant 0.210 *** 0.203 *** 0.219 ***
(0.026) (0.045) (0.038)
Recipient fixed effects Yes Yes Yes
Observations 541 274 267
[R.sup.2] 39 59 20
Number of recipients 141 139 139
(4) (5) (6)
Variables All Family Foreign
Money 0.062 ** 0.079 * 0.011
(0.028) (0.042) (0.039)
Responded by e-mail in a week -0.165 *** -0.198 ** -0.097
(0.047) (0.082) (0.076
Responded by e-mail in 2 weeks -0.104 ** -0.156 * -0.031
(0.050) (0.093) (0.080)
Responded by e-mail in 3 or more -0.022 0.043 -0.032
weeks (0.050) (0.087) (0.082)
Responded to e-mail
Constant 0.203 *** 0.199 *** 0.203 ***
(0.026 (0.043) (0.037)
Recipient fixed effects Yes Yes Yes
Observations 541 274 267
[R.sup.2] 47 0.106 13
Number of recipients 141 139 139
Notes: Standard errors in parentheses. Dependent variable is Mail
Loss (= 1 if mail never arrived and = 0 if mall arrived). Independent
variables: Money = 1 if envelope contained money, Responded by
e-mail in a week = 1 if recipient responded to initial e-mail
informing the mailing of a card within a week of us sending the
e-mail. Responded by e-mail in 2 weeks = 1 if it took 1-2 weeks,
Responded by e-mail in 3 or more weeks = 1 if it took 3 or more
weeks to respond.
* p<.10, ** p<.05, *** p<.01.
TABLE 10
Robustness Check on Reporting Probability of Mail Loss with
Restricted Sample OLS Fixed-Effects Regressions
Observations Confirmed
After 4 Weeks Dropped
(1) (2) (3)
All Family Foreign
Money 0.070 ** 0.062 * 0.052
(0.029) (0.035) (0.045)
Constant 0.093 *** 0.091 *** 0.108 ***
(0.020) (0.023) (0.023)
Recipient fixed effects Yes Yes Yes
[R.sup.2] within 0.025 0.042 22
N 366 189 177
Losses Confirmed After
Weeks Switched to No Loss
(4) (5) (6)
All Family Foreign
Money 0.048 ** 0.065 ** 0.027
(0.023) (0.033) (0.035)
Constant 0.063 *** 0.052 ** 0.076 ***
(0.016) (0.023) (0.024)
Recipient fixed effects Yes Yes Yes
[R.sup.2] within 11 0.028 5
N 541 274 267
Notes: Standard errors in parentheses. Dependent variable is Mail
Loss (= 1 if mail never arrived and = 0 if mail arrived).
Independent variable: Money = 1 if envelope contained money.
* p <.10, ** p <.05, *** p <.01.
TABLE 11
Robustness Check on Reporting Probability of Mail Loss with Reported
Losses Changed to No Loss Average and 90% Confidence Interval of
Bootstrapped OLS Recipient Fixed-Effects Regressions
20% of Reported Losses
Changed to No Loss at Random
(1) (2) (3)
All Family Foreign
Money 0.042 0.066 0.006
[0.011,0.070] [0.026,0.103] [-0.033,0.044]
Constant 0.107 0.098 0.123
[0.087.0.125] [0.075,0.122] [0.096,0.150]
30% of Reported Losses
Changed to No Loss at Random
(4) (5) (6)
All Family Foreign
Money 0.036 0.056 0.004
[0.004,0.065] [0.017,0.101] [-0.039,0.045]
Constant 0.091 0.084 0.107
[0.070.0.110] [0.056,0.112] [0.072,0.137]
Notes: 9090 CI in brackets. 10.000 bootstraps. Dependent variable is
Mail Loss (= 1 if mail never arrived and = 0 if mail arrived).
Independent variable: Money = 1 if envelope contained money.
