_____  _  __we-51Rqj
PCo1 J(Y   Rl:-SAR, 11   OR KING  PAP ER                                  1999
Heterogeneity among                                                       The small firm sector
comprises several types.
MNexico's MN"ticro-Enterprises                                           Althoughsomearearefuge
for displaced salaried workers,
most were started by workers
a- An  Appu1Cat[ICST o t Q   Factor                                who chose to be
and  Cluster Anialysis                                                    entrepreneurs - and they
show dynamics consistent
WV;eudelv V . (]i,,i )?zlJ1,~lJwl)                                        with patterns in the industrial
WN7IaIiLnz F. Alalhuic'                                                 world.
it I    ,\ 1 I  M t i t     - I  1 h.ani --, gi 0
)\-tv Rcd,tit-ion .uid           M ni n1    N lanageient Sector Ul-lit          U
Octtobcr N'),S                                                                     fif



|POLICY RESEARCH WORKING PAPER 1999
Summ.rary findings
A long tradition sees the smail firm sector as a holding      employrment in order to be independent, collect higher
pattern for workers queuing for jobs in the formal sector     earnings, or follow family tradition. These survey
of a segmented labor market. An alternative                   responses are supported by the finding that income
"entrepreneurial" view suggests that many workers             distribution adjusted for human capital is composed of
prefer self-employment to salaried jobs. These competing    two subdistributions, with the "underperforming"
views can be resolved if the sector is heterogenous.          distribution comprising only 14 percent of the sample.
Using factor and cluster analysis, Cunningham and             The factor analysis also implies that firm owner
Maloney generate a typology of the sector by taking           characteristics and firm size or profitability may not be
advantage of a Mexican data set on micro-firms that           correlated. For example, young workers who we might
offers information on a broad range of small firm             think are forced into the small firm sector due to
characteristics. The methodology permits divisions to         inability to enter the formal job market do not
emerge from the data without the a priori imposition of       necessarily earn less or have less capital than older
a theoretical structure.                                      entrepreneurs. Furthermore, a distribution of the
The data break into several distinct groups, broadly        earnings residual factor shows that very few firms,
characterized as highly profitable and dynamic young          regardless of the firm owner's age, are earning below
firms, older firms that have stabilized at a small size, and  their expected profits.
young firms that act as an employer of last resort. Those       The data suggest that the small size of informal firms
in the last group, comprised of older entrepreneurs with      may not necessarily result from limited access to financial
low levels of education, are the most likely to cite that     institutions or a desire to evade labor or tax laws.
they started their firms because they were unable to find     Instead, the firms simply may be in the beginning stages
a salaried job.                                               of a growth process or owners may prefer to remain
In general most of the firm owners in all groups stated     small.
that they chose self-employment over formal sector
This paper is a product of the Poverty Reduction and Economic Management Sector Unit, Latin America and the Caribbean
Region. Copies of the paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please
contactTania Gomez, room 18-102, telephone 202-473-2127, fax 202-522-2119, Internet address tgomez@worldbank.org.
William Maloney may he contacted at wmaloney@worldbank.org. October 1998. (42 pages)
The Policy Research Working Paper Series dissemiinates the findings of work in progress to necourage the exchange of ideas about
development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than folly polished. The
papers carry the names of the authors and should he cited accordingly. The findings, interpretations, and conclusions expressed in this
paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the
countries they represent.
Produced by the Policy Research Dissemination Center



Heterogeneity among Mexico's Micro-Enterprises:
An Application of Factor and Cluster Analysis'
Wendy V. Cunningham
William F. Maloney
We thank the Mexican National Institute of Statistics, Geography, and Information (INEGI) for the use of the data.
INEGI is in no way responsible for any incorrect manipulation of the data or erroneous conclusions drawn from it.



I. Introduction
Owners of small firms, defined as those of under six employees, constitute 20-25% of the
Mexican work force and those working for these firms roughly 10%. Both their importance as
sources of employment and their tendency to be found in the lower end of the income spectrum
makes understanding the dynamics and reason for being of these firms essential to both issues of
labor market efficiency and poverty and distribution.
A long tradition views the sector as the subsistence holding pattern for workers queuing
for jobs in the formal sector of a labor market segmented by above market clearing labor costs.2
However, an alternate view dating from Keith Hart's (1972) accounts of the dynamism of
Kenya's micro-entrepreneurs takes the basic unit of analysis as thefirm, rather than the
disadvantaged worker, and stresses the large overlaps in the distribution of remuneration between
the self-employed and salaried sectors and that, as in the industrialized world, workers often
prefer self-employment to salaried jobs.3 Within this firm-centered view coexist further
contradictory hypotheses on firm dynamics and the reasons that small firms are small, such as
satisficing behavior vs. structural constraints to growth or a desire to evade formal institutions vs.
lack of access to them.
These conflicting views can be reconciled if, in fact, the sector is very heterogeneous and
contains elements of each story, a possibility acknowledged by the literature in its increasing
reference to "upper" and "lower" tiers. However, to date, the data on the micro-firm sector has
been largely anecdotal and offers little evidence for this division, or other divisions of equal
2 The Harris-Todaro (1970) model is perhaps the traditional statement of this view. Rigidities or excess
remuneration may arise from severance pay, social security, medical benefits, high minimum wages, and costly
regulation in general. See Stiglitz(1974) and Esfahani and Salehi-Isfahani (1989) for an efficiency wage approach.
See Tokman (1989), Fields (1990), Portes and Shauffler (1992).
l



interest. This paper takes advantage of a large and comprehensive data set on Mexican micro-
firms that offers information on a broad cross section of small firm characteristics along which
we may find division or segmentation and that makes generating a richer typology feasible. To
this end we employ two econometric techniques, factor and cluster analysis, that permit the
segmentation of the market to emerge from the data with a minimum of prior structure imposed.
These techniques have been employed in the literature on industrial or labor market dualism in
the industrialized countries by Oster (1979) and Anderson et. al. (1987) for the US, Sloane et. al.
(1993) for the UK, and Flatau and Lewis (1993) for Australia.
II. Possible Sources of Heterogeneity
In the spirit of Hart, Levenson and Maloney (1997) argue that small firm dynamics in
LDCs may be well described by models now standard in the industrialized country literature
rather than representing qualitatively different phenomena. These models imply heterogeneity
that is unrelated to structural features of LDCs or labor market distortions. Lucas (1978) argued
that the vast range of firm size among established firms can be derived from a model where
entrepreneurs have differing cost structures. Building on this, Jovanovic (1982) offers a model of
firm dynamics where entrepreneurs face a multiplier on their costs, x(O+c), that is a function of
entrepreneurial ability or perhaps firm location, 0, and s captures random noise:
C = x(0±+)c(q).                             (1)
It is assumed that c'(q)>O and c"(q)<O where q is the level of output. Entrepreneurs have only a
vague idea of what their true cost structure will be at the time of start up and only with
experience can they generate a more precise estimate. If they find themselves with profits above
2



those expected they will lower their estimates of their costs, invest, and expand to their long run
size. Others may find themselves far less profitable than expected and go out of business. This
framework has two implications.
First, the observed distribution of mature firm size and the co-existence of larger, more
capital intensive, high earnings firms with very small ones may only reflect the underlying
distribution of whatever factors contribute to 0. Assertions of the prevalence of "pre-capitalist,"
"non-optimizing," or "satisficing" modes of production (Tokman 1992 and Elizondo 1992) may
be nested if we argue that entrepreneurs are maximizing if not profits, then utility more generally
and that the high implicit valuation of additional labor after basic needs are satisfied makes 0
very high. Second, among small firms we should find long lived firms at their long run size,
start-up firms that find themselves extremely profitable and desire to expand, and start-up firms
that are about to fail. We should also find differing returns to factors that reflect this normal
heterogeneity. On average, the youngest firms should show the highest mortality rates since they
have the least refined estimates of their viability and are more likely to discover that they are not
than more established firms.4 In sum, small size, lack of growth, or even differential returns to
human and physical capital may be due to variance in individual abilities and preferences and the
noisy process of discovering them instead of due to distorted labor and credit markets or an
inefficient regulatory framework, which themselves may give rise to other divisions in the small
firm sector.
Labor market distortions that lead to a rationing of workers from their preferred position
in salaried work should lead to segmentation in the micro-enterprise sector as well. First, we
4Numerous US and UK surveys support these models showing extremely high rates of firm mortality among young
and small firms that decrease with age and size (See Davis et al 1994, Evans and Leighton 1989, Dunne et al 1989).
3



should find concentrations of young workers, migrants, and perhaps laid off middle age workers
who report having trouble finding salaried work coexisting with those who voluntarily became
self-employed. Further, we should find the standard differences in earnings equations postulated
in the dual labor market literature. Returns to the human capital of those voluntarily in the sector
should be comparable to those in salaried employment while those involuntarily in the sector
who were rationed out of "good" jobs would be lower.5
Credit constraints on entrepreneurs are evident in the industrialized countries (Evans and
Jovanovic 1989) and are very likely in LDCs as well. "Life cycle" behavior where workers enter
into salaried work; accumulate knowledge, capital, and contacts; and then quit to open their own
businesses may represent a natural life progression, but may also reflect credit market
imperfections, or even the poor quality of the education system. If credit market imperfections
constitute an important barrier to growth for some firms, we should expect to find a group with
low O's, earning supernormal profits, who desire to expand, but lack access to resources.
There is no consensus in the literature on how compliance with labor or tax legislation is
related to production decisions. To avoid taxes or regulatory legislation, firms may choose to
stay small or if they grow large, the firms may subcontract to smaller firms (Loayza 1997, Rauch
1991, Portes and Shauffler 1993). But, de Soto (1989) asserts that many aspects of "formality"
are essential for the growth and efficient operation of firms and are something for which firms
are willing to pay, but are often denied by oppressive bureaucracies. Nesting both, Levenson and
Maloney (1996) argue that participation in the institutions of civil society - for example those
I In fact, if the return to factors in the formal sector were imputed, many involuntarily self-employed might have
cost structures that would dictate that they close. In the absence of unemployment insurance, a laid off middle aged
worker unable to find salaried work may invest his severance pay to generate employment, perhaps buy a taxi, even
4



that grant secure property rights, permit formal contracting mechanisms, or access to formal
financing mechanisms - can be considered a factor of production that becomes increasingly
important as firms grow and that has a cost measured in terms of compliance with legal norms.
For many high 0, and therefore permanently small firrns, issues of formality may therefore be
largely irrelevant.
III. Data
The data are from the 1992 National Micro-Enterprise Survey that was collected by the
Mexican National Statistical Institute in 1992:1. This survey was constructed by selecting 11,000
owners of micro-enterprises, defined as firms of under six individuals (16 for manufacturing
firms), from the last quarter of 1991 of the National Urban Employment Survey and examines
reasons for becoming self-employed, incomes, capital structures, costs, enterprise problems,
credit needs, migrant status, and employment patterns, among others. To focus on the firms
commonly thought of as informal, we restrict the sample to males ages 16 - 65 with no more than
12 years of formal education.
The analysis include four broad classes of variables, used both raw and transformed into
indices (see appendix for more detail):
A. Characteristics of the Entrepreneur:
1. Years of Education of the Entrepreneur (Education).
2. Years of Work Experience of the Entrepreneur (Experience).
3. Why initiated enterprise (Voluntary). Captures whether the entrepreneur is voluntarily
employed in the sector.
though if the cost of capital were fully imputed, the firm might not be viable. The worker is borrowing against
future (past) income to subsidize an unviable firm.
5



B. Characteristics of thefirm:
1. Firm Life in Years (Firm Life).
2. Capital-Labor Ratio (K/L). Reported total capital stock of the firm divided by the total
number of workers in the firm.
3. Earnings (Earnings). Monthly total revenues of the firm less expenses.
4. Permanence of Work site (Site). Whether the firm has a permanent site or not.
5. Average Hours Worked by Entrepreneur per Week (Hours).
6. Total number of workers in the firm (Workers).
C. Entrepreneurial Dynamics:
Since the sample does not permit explicitly testing for dynamics, we rely on the
entrepreneur's statements of his plans and needs for financing.
1. Permanence in the Sector (Move/stay). Whether the entrepreneur plans to close the
firm and look for salaried work.
2. Plans to Expand (Expand). Whether the owner plans to go out of business, remain in
business but not expand, or expand.
3. Clients (Clients). captures the nature of clientele ranging from local families to large
businesses.
4. Initialfinancing for firm (Financing-Initial). captures the source of start up capital,
ranges from personal savings, friends, and informal markets up to formal banking
institutions.
5. Demandfor additionalfinancing (Financing-New): Whether the firm has sought
additional financing from any source.
6. Financing problems (No Credit Problems). Whether the owner cites credit availability
as a business problem.
7. No problems (No Problems). Whether the firm reports having no business problems.
6



D. Participation informal market and legal institutions
1. Registration with Tax Authorities (Tax). Ranges from being unregistered with tax
authorities, to being registered locally, to being fully registered with federal tax
authorities.
2. Compliance with Labor Legislation (Benefits). What fraction of workers are covered
by social security and health benefits.
IV. Empirical Methodology
i. Factor Analysis6
Factor Analysis is concerned with finding a small number of common factors that linearly
reconstruct a large number of variables:
h
z = E ajpFp, + eg                                          (2)
p=l
Where zj is the value of the id observation on the j" variable, Fpi is the ii observation on the pi
common factor, a3p is the weighting on that factor, and e1j is known as the j' variable's unique
factor and is analogous to the residual in standard regression analysis. The factors themselves
are linear combinations of the variables such that:
k
Fpi= E qpjzj                                       (3)
j=1
Where Fp; is the value of factor p where there are h selected factors. for individual i for each of
the n individuals with observations on k variables and q is the weighting of the p' factor in
variable j. By isolating a subset of the k possible factors, a large and seemingly chaotic mass of
data can be found to be structured along a few critical dimensions.7
6 See Harman (1976) for a detailed review of the technique and Kim and Mueller (1978) for a more general review.
7 The initial data matrix is rotated to generate a diagonal matrix where the eigenvalues reveal how much of the
variation occurs along each dimension. A subset with the highest explanatory power are chosen and the resulting
7



The technique has four attractive features that aid in teasing out the composition and
dynamics of the micro-firm sector. First, it isolates covarying variables that suggest types of
entrepreneurs or firms, that would not be possible from simple bi-variate cross correlations,
again, without imposing any prior structure on the data. Second, if in fact there are "good" and
"bad" firms within the micro-firm sector, we should be able to identify a "dualistic" factor that
loads heavily on the characteristics commonly associated with firm "quality." Third, since the
value of any particular factor differs for each observation, the factor will show a distribution that
offers a statistical test for segmentation. If, in fact, there are two types of firms with distinctive
modes of operation and dynamics, we might expect the observed distribution to be comprised of
two underlying distributions corresponding to the inferior and superior sub-sectors:
F _ p N([L,,a12) + (l-p) N(1i2, a2)                                 (4)
Using the EM algorithm, we maximize the likelihood across both means, standard deviations and
the weighting on each distribution, the share of the population found in each segment, p, (l-p).
We then compare the likelihoods to test against the null that the data is drawn from a single
normal distribution. Fourth, because the factors are mutually orthogonal by construction, they
cannot be causing the segmentation or patterns observed in each other.8
matrix is rotated again to aid in interpretation by re-expressing the factors so that the loadings on a few variables are
as large as possible. The principal component method of extracting factors was employed rather than the common
maximum likelihood or other methods because it does not require normality in the distribution of the variables.
Though the various alternative rotation methods yielded similar results, the Varimax method offered the clearest
separation of factors and preserved the orthogonality of the initial factors.
8Since the factors identified may be sensitive to the variables included in the analysis, we ran several specifications.
In particular, variables with very high loadings were eliminated to see if that variable alone was driving the factor.
In general, the same factors formed with the same economic interpretations.
8



ii. Cluster Analysis '
The Ward clustering method combines observations on entrepreneurs and their firms into
progressively larger endogenously determined clusters by minimizing the sum of the within
group variance of all clusters:'0
k=g j=ri=nk
W=E        E(XYk -Xjk)                                      (5)
k=l j=1 i=1
where 5jk is the mean value of variable j in cluster k, Njk is the value of an observation assigned
to cluster k, nk is the number of observations in cluster k, m is the number of variables, and g is
the number of clusters. Starting with n clusters each consisting of a single observation, this Ward
value is calculated n times and takes a value of 0, when n=k and the distance of the n
observations from the n cluster centers (themselves) is zero, and then increases as more distant
points are forced to share a common center until n=l. The optimal number g of clusters is
decided and each observation is assigned to the nearest centroid, or mean vector."
9 See Anderberg (1973), Everitt (1974) and Spath (1980) for more detailed treatments.
'° Cluster analysis requires some prior manipulation of the data. First, it is very sensitive to magnitudes, so each
variable is normalized to a scale with a range of [0,1]. Second, extreme values or outliers compress the bulk of the
observations into a very limited range and yield trivial clustering results. In some cases, therefore, these values are
suppressed or recoded to provide a greater variance across the majority of the sample.
" The n values are used in two ways to determine the number of clusters. First, a dendrogram graphically depicts
the hierarchical combination process as clusters are formed and consolidated which suggests the appropriate level of
consolidation. A second, but related method is to stop clustering when the Ward value rises above a certain
threshold or the second derivative of the plot suddenly increases. This reveals when agglomerations are relatively
more "forced" and where the data should be left in distinct sub-clusters. Observations are assigned using the
standard MacQueen's k-Means Method. Since the resulting clusters may be sensitive to the initial centroids, the
analysis was run using different beginning seeds for both steps. Regardless of the initial seed used, the final
centroids were very similar and the characterization of the clusters did not change.
9



V. Results
i Continuous variables
Although in both types of analysis, any kind of variable can be included, in practice,
discrete or non-continuous variables can introduce trivial bi- or multi-modality in the
distributions of factors, and their extreme values tend to dominate the clustering process. We
therefore begin working with just the six continuous variables, education, experience, firm life,
weekly hours worked, capital per worker and earnings. In various combinations, these variables
emerge as elements of segmentation stories: marginal, under-employed workers in low profit,
low capital firms vs. better capitalized prosperous firms; young workers unable to enter formal
employment in short lived firms vs. established voluntary entrepreneurs. If such discrete sub-
groups exits, both techniques should identify them. We then draw on more descriptive evidence
on motivations for being in the sector, business problems, degrees of compliance with social
norms and test directly for segmentation among these groups by comparing estimated earnings
equations. Drawing together these sources, we generate a typology of the subgroups.
ia. Factor Analysis
The loadings on two factors that explain 56% of the variance observed in the data are
presented in table 1 a.
Factor 1: Generational characteristics (32% of the variance). The factor loadings show
firm life and worker experience entering strongly positively and education strongly negatively.
Younger workers are better educated and, clearly, cannot have been in business as long as their
older counterparts.
10



Factor 2: "Dualism? "(23% ofthe variance) What might be considered as a dualism
factor loads heavily and positively on earnings, hours worked by the entrepreneur, capital-labor
ratio, and to a lesser degree, education and firm life.
The orthogonality of these two factors implies that there is no correlation between
generational effects and those determining capital intensity or prosperity. Of greater interest,
however, is the evidence for dual distributions in each factor. Table 1 suggest that the
distributions sketched in figure 1 are mixed. The first factor contains a distribution comprising
roughly 43% of the sample that has higher levels of education, relatively less experience, and,
logically, shorter firm life, and a second with lower levels of education and more experience.
For the tentative dualism factor (figure Ib) 26% of the population corresponds to a
distinct distribution centered at a higher level of capital per worker, earnings, hours worked, firm
life, and education, thus providing some support for an upper tier/lower tier division. However,
dualism in the sense of different returns being earned for the same productive factors can only be
tested by adjusting earnings for these endowments. We employ a standard earnings equation
augmented by returns to capital and interpret the residual, q, as a measure of earnings above or
below what would be expected given levels of human and physical capital: 12
Ln(Earnings) =   . + 13, Educ + 132 Exp +±3 Exp2 + P4 Exp*Edu + f5 K + P6 Ln(Hours) +  '   (6)
2 Taking the log of hours is standard in wage equations because when 16 is constrained to be 1, the dependent
variable becomes the log of hourly wages. Its inclusion here is somewhat problematic since returns to capital
should not be a function of hours worked necessarily. Though not optimal, the present specification at least lets the
data identify the value of the coefficient.
II



Table 2 presents the results of replacing earnings with the residuals. Factor 1 again captures a
generational effect. A "capital intensity" factor emerges that suggests that entrepreneurs with
more capital per worker tend to work longer hours, tend to be better educated and, consistent
with Jovanovic's approach, have been in business longer.
Factor 3 weighs heavily on the residual, but on none of the other variables important in
the previous "dualism" factor. Further, the MLE estimates suggest only 13% of the sample is
found in a distinct, lower distribution that might be attributable to the disadvantaged sector of a
dual labor market. The positive association of earnings in the "dualism" factor arises from the
distribution of human and physical capital rather than differing returns to them.
ib. Cluster Analysis
The results from the cluster analysis tell a broadly similar story. The mean values for the
six clusters are given in table 3, the distance between the final cluster centers in table 4, and the
mean values converted back into their original range in table 5.13 Tables 6-8 provide information
on a variety of firm and enterprise characteristics by cluster.
The generational division identified in factor 1 broadly emerges, with clusters 1-3 being
largely young and capturing 53% of the sample and 4, 5 and 6 capturing older and less well-
educated workers. The other divisions correspond broadly to the second factor which, as the
orthogonality of the factors would predict, cuts across generations. Clusters 2 and 4 are distinctly
higher in capital intensity, although the high earnings, hours worked, and education suggest that
cluster 1 could also be part of an upper tier comprising 25% of the sample-- close to the 26%
suggested by the MLE.
13 In reality, the dendrograms leaned slightly more heavily to there being five factors but the six factor break, while
similar, yielded an additional important division.
12



Though we would again like to compensate for human and physical capital, running the
cluster analysis using the residuals from the earnings equation proved problematic. 4As an
alternative, table 9 presents the results of the earnings equations for each cluster. First, the
returns to education, experience, capital, hours worked and other coefficients were constrained
equal leaving just the intercept to vary. Given the sample-wide return to those factors, this tells
how much each sector earns above or below that of the most numerous case, cluster 3. The
specification accounts for 33% of the variance and the principal coefficients are significant and
of the predicted sign with positive returns to education, experience, capital, and hours worked.
The next 6 columns allow the returns to factors to vary as well. The asterisks on the individual
coefficients, and the joint tests on the coefficients taken as a group indicate statistically
significant differences from cluster 3.
ic. A Tentative Typology Using Continuous Variables
Cluster 1: Very prosperous, low capital-intensity firms (14%). Entrepreneurs in this
sector revealed very low involuntary entrance (17.0%) and very high desires for independence,
higher pay, and continuing family tradition. With a mean age of 39 years, 78% were salaried
employees before. They have very little capital, but are the largest firms (1.02 employees), and
earn roughly 2.5 times the lowest sector. The very high mean wages, even when controlling for
education and experience, suggest that they may have informal or on the job training or skills not
captured by the education proxy. Unsurprisingly, they report the fewest problems with low
earnings or excessive competition."5 They are among the least likely to exit and the most likely to
14 The relatively long tails of the residuals distribution led to cluster means that showed virtually no differences in
adjusted returns. While truncation of the tails yielded enough "spread" in the variable to generate some variance, the
results were somewhat sensitive to how it was done.
" The vagueness of the motivations questionnaire reported in table 7 makes interpretation of the exact nature of
several of the business problems very difficult. In a healthy competitive environment, complaints of "excessive
13



expand. Although they relied primarily on their own resources for start up and have little
demand for additional financing, they do appear relatively more constrained for credit than all
but cluster 2. This sector is clearly the best off of all the sectors, although, importantly, it is not
the sector with the most capital per worker.
Clusters 2 & 4: Highly capitalized firms (7% and 4%0) Paradoxically, both clusters have
much higher capital-labor ratios but lower earnings than cluster 1. This is likely due to the fact
that both groups show 40% of their members working out of vehicles, 10 times that of the other
sectors, a fact that probably explains their disproportionate problems with (traffic) authorities as
well. The fact that they receive the lowest return to capital reflects perhaps over-capitalization
due to the lumpiness of vehicle purchases, or that they may not fully own the vehicle. Neither
shows significant returns to education, and both show extremely low returns to experience as
might be expected in a vehicle operator.
It is in the demographics and the reasons for being in the sector that the two clusters
differ. Cluster 2 resembles the younger cluster 1 although the percentage of entrepreneurs who
left their previous job due to plant closings or being laid off (27%), and having previously
worked for firms of over 50 people (43%), are nearly double that of any other sector. It is
possible that they used their severance pay to invest in a car or truck. Yet the number citing
inability to find another salaried job as the reason to start the business is relatively low (10.7%),
and the share planning to expand the business, 30%, is as high as the prosperous cluster 1.
Whatever their motivation for entering, once in, they choose to stay.
competition" are probably not uncommon. "Lack of economic resources," "low earnings" and "lack of clients" may
be covering similar ground and may not represent evidence of any distortions.
14



Cluster 4 entrepreneurs are clearly of the older cohort with a mean age 15 years higher
than that of cluster 2, and show the longest firm life after cluster 6. As with clusters 5 and 6,
being older appears to make finding alternate salaried employment less easy, but among the older
groups, they are the least pessimistic and exceed all other groups in reporting the desire to be
independent as the reason for leaving their previous job (59%) and starting this business (60%),
or family tradition (14%). Of those who left their previous job, they have the lowest incidence
(15.6%) of having been involuntarily separated. Importantly, this group is overwhelmingly
voluntary in the sector but complaints about a lack of clients or low earnings are not particularly
lower than other groups.
Cluster 3: Younger entrepreneursl startups (32%). This is the largest of the subgroups,
with the youngest workers (31 years), the firms with the shortest life (3.75 years), the second
lowest level of capitalization, and earnings adjusted for sample wide returns to human and
physical capital between 10 to 86% less than those of every other group (table 9). These workers
show the second highest levels of pessimism about finding other salaried work, (15.2%),
downplay independence as a motivation (35.1%), and cite complementing family income as an
important reason for starting the business (14%).
But there are many reasons not to think of the cluster as the safety net lower tier resulting
from a segmented labor market, but rather as the normal start up firm sector with its very high
mortality rates, and a higher proportion of entrepreneurs making insufficient or negative profits.
First, involuntary entries still constitute only 20% of the total, and 32% said they left their
previous job because it did not pay enough, the highest of any cluster. Many may be young
workers, starting out, and shopping among sectors for good jobs. While it is true that 64% report
having insufficient clients as their principal business problem, the highest of any group, this
15



might be expected of younger firms establishing their reputation, and even the clearly voluntary
and prosperous sectors 1 and 4 report rates around 55%. They do not complain
disproportionately about excess competition or low earnings, nor is their growth particularly
constrained by non-availability of credit. Fewer do plan to expand and more plan to leave
business, but again, this may simply reflect the fact that a large fraction of younger firms may go
out of business.
Cluster 5: Life-cycle entrants, mid-life lay-offs, and retirees (25%o). This group
comprises a quarter of the sample and consists of older entrepreneurs who opened their firms
relatively recently. The timing at mid-life offers support for theories of life cycle behavior, but, a
relatively high percentage was laid off from their previous job and 22%, more than any other
sector, report that they started this business because they could not find another salaried job. This
would be the case if employers tend not to hire older workers. Double the percentage of any
other group (3.7%) are retirees, and almost 8%, vastly higher than any other group, left their
previous job because of illness or injury. There is little family tradition, and more than even the
young, many workers (16%) started the business to augment their family's income. Adjusting
for sample wide returns to human and physical capital, the group is perhaps 10% better off than
base cluster 3, making it the second least well off of any sector. While the sector clearly serves as
a safety net for displaced older workers, the sick and the retired, it also shows the highest rate of
return to experience and the second highest rate of return to additional hours worked.
Cluster 6: Older owners of establishedfirms with little capital (18%). These
entrepreneurs are among the oldest with a mean age of 51 and firm lives of roughly 23 years.
They have few plans to expand and have clearly stabilized at a small size. Yet, overall, the group
is more similar to the prosperous clusters 1 and 4 than either 3 or 5. First, adjusted for sample
16



wide returns to human and physical capital, earnings are clearly closer to those of 2, 4, and
perhaps 1, than 3 or 5. For unclear reasons, returns to their 4 years of education are negative and
to experience are insignificant although they show the highest returns to capital. The cluster is
second lowest in entrepreneurs entering because of lay offs or plant closings and, despite the age
of these workers, the pessimism about finding a salaried job is not especially high. It holds an
intermediate position both toward independence and higher pay as motivations for entering the
sector. As with clusters 1 and 4, a sizable group (13.3%) started the firm because of family
tradition. With the exception of reporting a relative lack of clients, its business problems are not
so different from those of the prosperous sector 4.
In sum, the most natural grouping of firms by characteristics frequently associated in
popular theories with segmentation do not generate sectors that are clearly the repository of the
involuntarily self -employed. That roughly 70% of all clusters enter voluntarily as they report, is
broadly confirmed by the estimates of the underlying distributions in the earnings equation
residual that shows a "lower tier" of under 20%. The decisions to remain in the sector are largely
independent of salaried alternatives. The generalized absence of plans to expand across clusters,
again does not appear related to business problems, whose incidence is similar across groups, or
lack of access to credit which is rarely cited as a binding constraint. Finally, there is wide range
of compliance with both social security and tax legislation: between 23% and 65% of the firms
are registered with the national tax authorities, and 11% to 44% offer at least one of their workers
benefits. The micro-firm sector should not be conceptually conflated with informality, although
the nature of the relationship cannot be determined from this data.
17



ii Results using complete variable set.
iia Factor analysis
Table 10 presents the loadings on the six factors that emerge when non-continuous
variables are included. Table 11 shows that running the analysis with the residuals from the
earnings equation yields extremely similar results.
Factor 1: Size/capital intensity (17.1 % of the variance). Factor 1 is essentially an
augmented version of the tentative dualism factor (factor 2) of the previous analysis. In addition
to loading heavily again on capital-labor ratio, earnings, hours worked and less heavily on
education, it now also loads on number of workers, permanence of physical location, likelihood
of paying taxes, and somewhat less reliance on credit for initial borrowing and expansion,
clientele size, and being voluntarily in the sector. All of these things might be expected with
more capital intensive, larger industries, who would need more permanent sites. This size and
permanence, perhaps because of the need to participate more in social institutions, such as
guaranteeing property rights or contracting credit, and the increased difficulty of evading may
imply a rise in compliance with taxation. As in the continuous case, table 10 and figure 2
suggest a mixed distribution with the "upper tier" again, accounting for roughly 60% of the
sample. But again, this does not appear to reflect dualism in the labor market: replacing earnings
with il shows that, adjusted for sample wide returns to human and physical capital, earnings
decline with positive movement along the factor.
Factor 2: Generational characteristics (10. 7% of the variance). This factor is very
similar to factor 1 of the previous case with the addition of a weak weightings on plans to remain
in the sector. It also shares very similar underlying distributions.
18



Factor 3: Future plans and expansion (8.8% of the variance). This factor reveals the
unsurprising fact that those who plan to expand their business are also those who plan to stay in
the sector, and that these tend to have slightly higher raw earnings levels (although not adjusted),
larger clients and are more voluntarily in the sector. Plans to expand are weakly correlated with
problems accessing credit. Figure 2 shows that the distribution is clearly tri-modal, reflecting the
discrete nature of the underlying variables and that the vast majority of the entrepreneurs are not
planning to leave their present position.
Factor 4: Compliance with labor laws, financing/formality (6.9% of the variance). This
factor loads most heavily on the number of workers employed and the share of workers provided
the mandatory social security benefits. This could mean that larger firms less easily evade, but it
could also imply that attracting more workers implies offering them benefits. The factor also
loads heavily on earnings (both raw and adjusted), having used financing to start the firm, plans
for additional financing, registration with tax authorities, permanence of work site, having larger
clients, having no business problems, and being voluntarily in the sector.
Factor 5: Use offormalfinancing (5.9% of the variance). This factor loads heavily on
the use of more formal sources of financing for both initiating and expanding the business.
Importantly, it also loads negatively on the probability of having credit problems. Thus, those
most likely to have credit problems are not those rationed out of credit markets, but precisely
those who have had access before. The surprising absence of any weighting on education or
experience or firm life makes it somewhat difficult to argue that either lack of information or
entrepreneurial capacity prevent desire or access to financing. The factor also loads weakly on
plans to expand, on larger clients, and on a larger capital-labor ratio. The distribution of the
19



factor is highly skewed reflecting, again, that the vast majority of firms are self sufficient in their
financing, relies primarily on local clients, and has relatively low capital-labor ratios.
Factor 6 (5.7 % of the variance): Voluntary and small steady state firm size. This factor
loads heavily on how voluntarily the entrepreneur entered the sector, earnings, and longevity of
the firm. Associated with increased voluntariness is both lower tax payments, and less
permanence in site. Factor 6 is also positively associated with client size, thus implying that
those who voluntarily enter the sector may have started their firms with the express purpose of
subcontracting to a specific client. The more voluntarily self-employed are more profitable--they
were able to choose the project and the time. However, when including the residual in the
factoring, the clientele moves negatively to the voluntary variable. This may reveal that
subcontracting for a specific client does provide employment, but earnings are lower for those
who respond to firm-specific demand rather than to general market demand. As interesting as the
absence of any weight on financing variables, is the absence of any plans to expand or to leave
the sector. Further, the factor loads negatively on experience, and positively on education. The
distribution is arguably mixed although there appears to be a very small percentage of the
population (14%) as involuntary, recently established, low profitability firms, with large clients.
This might be consistent with individuals laid off from previous firms and now subcontracting
for them.
The orthogonality of these factors suggests that many issues that are often assumed
related may not be. As will be more clear in the cluster analysis, issues of credit, voluntariness,
formality, and size or capital intensity are largely unrelated and cut across each other.
20



iib. Cluster Analysis
The extreme values of some of the discrete voluntary, financing, move-stay, and migrant
variables cause clustering that effectively slices the data in a way that reveals the characteristics
associated with them. For this reason, table 12 tests the direct impact of these variables on the
estimated earnings equations. The mean values of, and distances between, the clusters are given
in tables 13 and 14.
Clusters 2 and 4: Large, capital intensive, with and withoutformalfinancing: Clusters 2
(9.1%) and 4 (26%) are located close to one another and differ fundamentally from clusters 1,3
and 5 by having higher mean values in permanence of work site, number of employees, capital-
labor ratios, and tax compliance. Cluster 2 is distinguished from all the other clusters by its
reliance on formal financing for expansion and it shows the characteristics that would be
predicted at the upper tail of factor 5: higher reliance on initial financing, a higher propensity to
expand, more credit problems, more permanent site, greater tax compliance and the highest
earnings. Table 12 suggests that these firms have a lower return to capital, as might be expected
if lack of financing implied inability to build capital to where the marginal product equaled the
market rate, and earn 10% more than those who do not need credit. At the other extreme, the
25% of the sample in cluster 4 represents the lower tail of factor 5: no use of formal sources of
funding, either to begin or to expand, a slightly lower but still the second highest ranking in
capital-labor ratio, reported earnings, level of permanence, number of workers, and size of
clients, and slightly fewer credit problems. As factors 1 and 5 suggest, access to formal financing
is not the critical determinant of size or mode of operation, or plans to expand.
Clusters 1,3 and 5 are similar in representing the lower tail of factor 1 in their few
linkages to formal financing, small client size, fewer workers, small size, lower profitability,
21



lower capital-labor ratios, lack of compliance with legal norms, hours worked, and lack of a fixed
physical location. But this "lower tier" is anything but homogeneous.
Cluster 1: Firm Closures (10.5%) This group differs from the entire remaining
population by declaring no plans to stay self-employed, or logically, to expand and thus captures
the lowest mode of the tri-modality of factor 3. The firms are distinguished by having the
shortest life span, the least experience of all sectors, but a relatively high degree of voluntariness
in their decision to open the business. Although their earnings are similar to those in clusters 3
and 5 who plan to remain in the sector, their high level of education suggests that these firms are
kin in aspirations to the successful cluster 4, rather than to the older, less educated clusters 2 and
3. Those exiting the sector imagined that they had the potential to run a profitable business, but
with experience, discover otherwise. They will leave the sector because, being young, they can.
By contrast, despite similarly low earnings, few workers, very low permanence of site,
lower levels of education, the smallest clients, the lowest capital-labor ratios, and the lowest
degrees of compliance with tax laws or worker benefits, both clusters 3 and 5 are committed to
the sector. They present extremes along factor 6 which largely, but not exclusively, weighed on
degree of voluntariness in entering the micro-firm sector.
Cluster 5: Involuntary (16%) These entrepreneurs were thrown into the sector, the least
education by far, the greatest experience, but not an unusually long firm life. This suggests that
they may have lost their salaried jobs at mid-life and are unable to find a new employer, a similar
profile to cluster 5 in the continuous case. Their earnings adjusted for sample wide returns to
human and physical capital are 24% below the rest of the sample. Returns to education and
experience are lower and borderline significant, although the returns to capital are significantly
higher. Thus, the most disadvantaged sector appears to be older men without the human capital
22



that would outweigh their advancing age, who do not have very good projects in the small firm
sector or, unlike the younger workers, access to salaried jobs.
Cluster 3: Voluntary, long term, successfulfirms (38%). These workers entered the
sector most voluntarily of all sectors. They earn less than the clusters 2 and 4 but more than
workers in clusters 1 and 5 so perhaps they should be considered a third tier altogether. The fact
that they have both less experience and a longer mean firm life than cluster 5 workers suggests
that, perhaps faced with unappealing salaried jobs, they chose to go into business for themselves
fairly early. While it would be tempting to argue that these firms are on their way to becoming
finns in the top tier, their mean age is the same as those in the prosperous sectors suggesting that
they are not quickly catching up. They appear only slightly interested in expanding, but report
the least problems overall in operating. What this suggests is that roughly 40% of the sector is
comprised of voluntary entrepreneurs, who have reached their long run level of output at a small
size and are formal neither in terms of worker protections or tax payments.
Finally, the inclusion of the variable registering whether the entrepreneur migrated to the
city as an adult was omitted because it dominated the cluster analyses. The clusters divided with
one cluster high in future financing and relatively low on non-migrants, reflecting perhaps, that
most who use credit to open businesses are not migrants. The remaining, four clusters
represented extremes along factor 1, size and capital intensity that were, in turn, each split
according to migration. These splits revealed that the characteristics associated with the 27% of
the sample that migrated as adults were virtually identical to those who were native. The
earnings regressions in table 12 reveal a similar story with both migrants and non-migrants
showing statistically very similar earnings functions. We can conclude that migrants are not in a
holding pattern for formal jobs any more than natives are.
23



VI. Conclusions
The continuous and complete variable sets tell a broadly similar story. Overall, the
observed groupings are more suggestive of the normal heterogeneity expected in any small firm
sector than the standard view of a dualistic labor market. The sector does serve as a refuge for
those unable to get salaried jobs, and this group does earn somewhat less given its human capital.
But it is a small minority of the population and concentrated in older workers who find it hard to
be re-hired in salaried jobs if let go, rather than the young. The vast majority of entrepreneurs in
all clusters is there voluntarily with desire for greater independence and higher earnings dwarfing
lay-offs or inability to find a salaried job as motivation. Those leaving the sector are the
youngest, but not necessarily those who entered because they could not find salaried work. They
correspond more closely to the failed, but voluntary entrepreneurs that are found in any small
business sector. There is little evidence that the sector serves primarily as a way of reducing labor
costs for large firms through sub-contracting since only a very small minority report larger
clients. Nor does not appear to be the case that migrants use the sector as a holding pattern for
salaried work any more than locals do, or that they earn less given their human capital.
There is a division along the dimension of capital intensity, size, permanence, and
formality, but this seems to result more from an unequal distributions of human and physical
capital, factors, including perhaps, entrepreneurial talent, rather than any dualism found in the
labor market: there is little relationship between capital intensity or size, and returns to human
capital. Further there is little evidence that the distribution of firm size, and in particular, the
small size of these firms is due to structural problems or market failures: both the optimistic,
prosperous sectors and those less so show relatively similar incidence of business problems and
few cite credit as the primary barrier to expansion. It is possible that the lesson has been
24



internalized that only with collateral and connections is access to formal sector credit remotely
possible and hence most do not try. But it cannot be ruled out that the desired steady state size of
over half the entrepreneurs is two or fewer workers and that they have no desire to expand
further.
25



Table 1: Factor Decomposition
Continuous Variables Only.
Factor I      Factor 2    Communality
Education               -0.78           0.24       0.66
Experience               0.88            ---       0.78
Firm life                0.68           0.16       0.49
Hours                    0.10           0.63       0.41
K/L                       ---           0.67       0.46
Earnings                   ---          0.75       0.56
Eigenvalue               1.93           1.43
% of variation           32.1           23.9
Moments and mixing parameter (p) offactor distributions
111                     -0.89          -0.36
P2                       0.45           1.02
at                       0.56           0.69
aF2                      0.87           1.06
p                        0.43           0.74
LR                     -159.95**      -445.80**
"*Statistically significant at the 5% level.
Kaiser-Meyer-Olkin Measure of Sampling Adequacy = 0.57
Table 2: Factor Decomposition
Continuous Variables Only.
Earnings Residual Replacing Earnings
Factor I    Factor 2    Factor 3    Communality
Experience             0.88           --          --       0.78
Education             -0.79        0.22         0.10       0.68
Firm life              0.68        0.16         0.17       0.51
K/L                   -0.12        0.75         0.15       0.59
Hours                    --        0.74        -0.17       0.59
Residual                 --          --         0.96       0.93
Eigenvalue             1.89         1.17        1.01
% of variation         31.5        19.6         16.9
Moments and mixing parameter (p) offactor distributions
11L                   -0.89       -0.31        -0.97
R2                     0.45        1.28         0.15
a,                     0.56        0.67         1.42
a2                     0.86        1.09         0.82
p                      0.34        0.80         0.13
LR                   -162.02**    -675.88**   -310.35**
**Statistically significant at the 5% level.
Kaiser-Meyer-Olkin Measure of Sampling Adequacy = 0.55
26



Figure 1: Factor Distributions - Continuous Variables Only
C~~~~~~~~~~~~~~~~~C
C~~~~~~~~~~~~~~~~~C
-2  -1    0    1    2    3                    -2       0       2       4
factor I                                        factor 2
27



Table 3: Mean Cluster Centers
Continuous Variables Only.
Cluster 1    Cluster 2  Cluster 3    Cluster 4    Cluster 5    Cluster 6
Education     0.64         0.67        0.66         0.52        0.25        0.33
Experience    0.42         0.37        0.29         0.62        0.66         0.68
Firn life     0.22         0.12        0.10         0.66        0.15         0.45
Hours         0.53         0.53        0.42        0.54         0.45        0.46
KIL           0.12         0.61        0.06         0.56        0.07        0.05
Earnings      0.59         0.42        0.24         0.41        0.25        0.29
%             14.46        6.98        31.83        3.93        24.67        18.12
n             640          309         1409         174         1092         802
Table 4: Distance between Final Cluster Centers
Continuous Variables Only.
Cluster 1    Cluster 2    Cluster 3    Cluster 4  Cluster 5
Cluster 1    0.00
Cluster 2   0.54        0.00
Cluster 3    0.41        0.60        0.00
Cluster 4   0.69        0.61         0.86        0.00
Cluster 5    0.58       0.77         0.55        0.78         0.00
Cluster 6    0.75       0.98         0.84        0.58         0.62
Table 5: Mean values of final cluster centers
Cluster
1        2          3         4         5       6
Firm owner characteristics
Age(years)                 38.76      36.05   31.43       49.40    48.26   50.76
Experience (years)         25.07      22.03    17.54      37.16     39.32   40.82
Education (years)           7.70       8.02    7.89        6.24      2.94    3.93
Hours(workedweekly)        52.66      52.51   41.97       53.56    44.45   45.38
Firm characteristics
Workers                      2.02       1.41    1.32        1.61     1.46     1.50
Firm life (years)           7.12       4.38    3.75        7.77      5.23   22.86
Capitalperworker'        2360.06   12,514.27  1160.00   11,441.47   1366.10  997.31
Earnings'                 1774.86    1260.92  722.72    1253.43    773.33  868.28
'in 1992 pesos where 3.1 pesos US$ I
28



Table 6: Motivations for Leaving Previous Job and Starting Business
Cluster
Motivation                        1       2       3      4       5       6
Why Did You Leave Your Last Job?
Laid off or business closed    13.7   26.7    19.5    9.9    19.8   13.3
Contract ended                  3.3    4.6      6.3    5.7      6.4    6.0
Pay too low                    25.6   18.2    31.8   17.4    25.6    25.4
To be independent              49.3   44.1    35.1   58.5    31.2   42.9
Retired                          1.9    1.5     0.9    1.7      3.7      1.7
Illness or injury                1.9    1.2     2.7    0.9      7.8    4.5
Other'                          4.1    3.8      3.7    5.9      5.4     6.2
No Response                     17.7   14.9    22.0   30.3    14.5   24.7
Total                            100    100      100    100      100     100
Why Did You Start This Business? 2
Fired or laid off               2.8    7.4      3.5    0.6      5.1      1.5
No other work available          5.2   10.7    15.2   13.2    22.0    15.7
To be independent              52.5   41.0    41.8   59.2    42.9    44.6
Family tradition                12.7    8.4     9.9   14.4      3.6    13.3
Complement family Y             10.8   13.3    13.8   10.9    16.1    12.9
Higher pay than salaried       33.3   23.3    27.8   24.7    22.4   24.1
Flexible hours                   1.7    2.3     3.7    2.9      3.5     3.5
Total Observations               640    309    1409    174    1092       802
%of sample                       14.5    7.0    31.8    3.9    24.7    18.1
' Includes family responsibilities or moving. 2 Multiple'responses possible.
Table7: Main Business Problems (in percent)'
Cluster
Problem                            1     2      3      4      5       6      Total
Lack of clients                  53.6   48.2   63.7   56.9   62.3   61.7   60.2
Lack of credit                   12.5   19.4    9.2    9.2    6.4    5.4    9.0
Lack of economic resources       18.4  22.7   21.2   13.2   23.9   17.1   20.5
Low earnings                    40.9   42.7   43.6   47.7   49.0   47.8   45.4
Problems with authorities         7.2   19.7    6.7   16.1    5.6    4.4    7.4
Excessive competition           43.0   52.4   44.9   53.4   40.8   45.9   18.4
Problems with workers            2.0    2.3    0.6    2.3    0.7    1.1    1.2
Late payments                    13.4   13.9   12.2    8.6   10.6   9.1    7.5
Raw materials problems            3.1    3.6    3.6    2.3    3.3    2.9    3.3
'Multiple responses possible.
29



Table8: Firm and Owner Characteristics by Cluster
Cluster
(as a percent of the cluster)  1     2       3       4       5       6
Site
automobile                 3.1   41.8     2.8    37.9      4.0     1.5
imnpermanent              52.3   50.8    75.7    27.0    76.6    75.8
permanent                 44.5   27.8    21.5    35.1    19.5   22.7
Clients
big frms                   6.2   5.1       2.5     7.8     3.0    3.1
small firms                11.9    9.6     8.3    13.2    6.8      7.0
public                     81.8   85.3    89.2    78.9   90.2    89.9
Financing to start firm
bank                       2.2   2.1       1.0    2.2      0.8     0.6
private loans              16.9  23.1    17.9   22.5    12.4    11.6
personal savings          60.9   61.5    52.5    51.1    60.4    51.4
credit (clients/suppliers)  4.7    3.8    2.3      4.8     3.4     3.7
donotneed financing        15.4   9.5   26.3    19.4   23.0    32.7
Additionalfinancing
do not need               67.1   63.4    73.9    56.3    74.3    75.4
too costly                 15.0   13.7    13.1    17.2    13.6    11.0
don't know how to apply     1.6   2.9    4.5       0.6    4.5      3.6
applied, did not receive    1.4   3.9      1.2    2.3      1.2     1.4
received financing from:
bank                      5.94   5.83    1.49   10.34    1.56    2.87
private loan             4.22   6.80    3.34    5.75    2.56    2.74
credit(suppliers/clients)  4.53   2.91    2.34    7.47    2.20    2.49
Maquila                      1.7    1.0     1.5    0.6      1.6    0.9
Plans
no change in firm         63.4   57.8    62.6   75.58    79.7    85.4
expandthefirm             29.9   29.9    19.9    19.2    11.1    10.2
shutdown the firm          6.2   11.2    16.5      5.2     9.0    4.3
Registrationltax
state treasury            31.4   30.1    18.0   36.8    19.1    18.8
Secretary of Treasury     53.8   56.0    23.2    64.9    24.7    27.6
Size of lastfirm:
self-employed               7.7    7.8     8.3     8.6    13.7    10.2
2-5                       29.1   22.0    28.7    25.9    30.9    29.6
6-10                       11.7   6.5      8.2    6.9      9.2     9.4
11-15                      5.3    3.9    4.2      4.6     4.4     4.1
16-50                      9.2    8.4    10.7    6.3      9.2     8.0
50+                       27.3    4.3    25.8    25.9   27.1    19.3
not previously employed      9.5    8.4    14.0   21.3      5.2    19.2
PayIMSS(asa%offirms       27.31   40.8   12.33   43.75   10.94   16.92
with salaried workers)
30



Table 9: Regression of log(earnings) on human and physical capital
Cluster'
All         1         2          3        4         5         6
Education               0.03      0.07        0.05      0.03      -0.08    0.05      -0.08**
(2.80)    (4.37)     (1.49)    (1.67)    (-1.06)  (0.96)   (-1.94)
Experience              0.02      0.01        0.01      0.03      0.01     0.06*    -0.002
(3.51)    (1.62)     (0.23)    (3.34)    (0.23)   (4.25)   (-0.074)
Experience2 (104)       -3.07     1.71**      -0.22**   -7.78     -2.05    -8.72     -1.64**
(-5.11)   (1.66)     (-0.079)  (-5.03)   (-0.47)  (-4.86)   (-0.67)
Ed*exp (10')            -6.95     -30.00*    -5.30      -6.45      18.52    -17.11    22.84**
(-2.62)   (-5.43)    (-0.36)   (-0.75)   (0.31)   (-1.49)   (2.21)
Capital (10-5)          1.48      1.15        0.77**    2.03       1.38     3.1      4.05**
(11.46)   (7.36)     (4.02)    (3.68)    (5.08)   (6.85)   (5.55)
Log(hours)              0.29      -0.11**    0.19       0.27      0.06*    0.38**   0.39**
(15.84)   (-5.43)    (3.03)    (9.48)    (0.51)   (8.95)   (8.49)
Intercept               5.08      7.42**      5.68      5.06      6.53     4.08**   5.41
(44.11)   (34.28)    (12.46)   (27.48)   (6.92)   (12.92)  (10.95)
dummy for cluster 1    0.86**
(29.70)
dummy for cluster 2    0.27**
(6.45)
dummy for cluster 4    0.29**
(5.47)
dummy for cluster 5    0.094**
(2.84)
dummy for cluster 6    0.24**
(7.07)
R2 adjusted             0.33      0.14       0.10       0.09      0.13     0.15      0.14
F value                 201.2**   18.57**    6.79**    24.45**   5.27**   33.3**   21.9**
joint test of coefficients  10.24**   127.12**   11.74**           6.39**   3.93**   11.48**
number of observations  4426      640         309       1409       174      1092     802
Numbers in parentheses are t-statistics. 'The asterisks on coefficients show significant differences from the coefficients of the
base case, cluster 3. *=Statistically significant at the 10-percent level, *  5-percent level.
31



Table 10: Factor Decomposition
All Variables
Factor                                Communality
1        2         3         4        5        6
Tax                      0.73        --        --     0.26       --   -0.11          0.62
Site                     0.71        --        --     0.27       --   -0.12          0.60
Hours                    0.62        --        --        --      --      0.10        0.40
K/L                      0.58     -0.10        --   -0.14      0.24        --        0.43
Earnings                 0.41        --      0.11     0.37       --    0.34          0.45
Experience                 --     0.88         --       --       --   -0.16          0.80
Education                0.20     -0.77        --       --       --    0.15          0.67
Firm life                  --     0.68         --       --       --      0.36        0.60
Move/stay                  --     0.13       0.92       --       --        --        0.87
Expand                     --       --       0.91       --    0.12         --        0.86
Workers                  0.24        --        --    0.72         --       --        0.58
Benefits                   --       --         --     0.72        --       --        0.53
No problems                --        --        --    0.15    -0.65    -0.12          0.47
Credit problems            --       --    -0.10         --   -0.61       0.12        0.41
Financing (new)          0.15        --        --     0.30     0.55        --        0.41
Financing (initial)      0.17        --        --     0.13     0.45        --        0.26
Voluntary                0.13     -0.13      0.14     0.18       --    0.70          0.57
Clients                  0.22        --      0.11     0.27     0.11      0.48        0.38
Eigenvalue               3.07      1.93      1.58     1.23      1.05     1.02
% of variation          17.15    10.73       8.76     6.88     5.85      5.71
Moments and mixing parameter (p) offactor distributions'
,1                      -0.71    -0.85         --    -0.33        --   -0.64
PL2                      0.48     0.48         --     1.16        --     0.62
a,                       0.36     0.58         --     0.47        --     0.94
CF2                      1.01     0.86         --      1.41       --     0.59
p                        0.40     0.36         --     0.78        --     0.49
LR                   -848.6**  -157.8**       --  -2225.0**       --  -308.5**
**Statistically significant at the 5% level. Kaiser-Meyer-Olkin Measure of Sampling Adequacy = 0.67
I Factors 3 and 6 are clearly tri-modal so moments could not be estimated.
32



Table 11: Factor Decomposition
All Variables
Earnings Residual Replacing Earnings
Factor                                Communality
1        2          3        4         5        6
Tax                       0.71        --         --.    0.36        --       --        0.64
Site                      0.71        --         --     0.35       --        --        0.64
Hours                     0.66        --         --       --        --       --        0.44
K/L                       0.53     -0.10         --       --     0.26        --        0.37
Experience                   --     0.87         --     0.10        --   -0.14         .080
Education                 0.22     -0.77         --       --        --     0.15        0.67
Finn life                 0.11      0.69         --    -0.12        --     0.28        0.58
Move/stay                    --     0.13       0.91        --       --       --        0.87
Expand                       --       --      0.90      0.12     0.11        --        0.86
Workers                   0.21        --         --     0.69        --     0.13        0.54
Benefits                    --        --         --     0.60        --     0.10        0.38
Clients                   0.12        --       0.10     0.47     0.10    -0.27         0.34
No problems                  --       --         --     0.20    -0.61        --        0.43
Credit problems              --     0.10      -0.12    -0.11    -0.59      0.12        0.40
Financing (new)           0.11        --         --     0.30     0.57      0.11        0.44
Financing (initial)       0.13        --         --     0.16     0.48        --        0.28
Voluntary                 0.19     -0.11         --       --        --     0.71        0.56
Residual                 -0.13        --         --     0.24        --     0.68        0.54
Eigenvalue                2.87      1.94       1.60     1.23      1.13     1.03
% of variation           15.95     10.79       8.87     6.81     6.27      5.75
Moments and mixing parameter (p) offactor distributions'
Pi                       -0.70     -0.85         --    -0.36        --    -0.54
Pu2                       0.46      0.49         --     1.09        --     0.54
a,                        0.37      0.58         --     0.58        --     1.05
Cr2                       1.02      0.86         --     1.19               0.56
p                         0.40      0.37                0.75               0.50
LR                     -767.9**   160.3**           -1124.3**           -415.9**
**Statistically significant at the 5% level. Kaiser-Meyer-Olkin Measure of Sampling Adequacy = 0.65
Factors 3 and 6 are clearly tri-modal so moments could not be estimated.
33



Figure 2: Factor Distributions - All Variables
CD''-                                           ,o.                                            oCa,
o                                               o
C')
-2      0       2       4                     -2   -1    0    1    2    3                    -3   -2   -1    0  
factor I                                        factor 2                                        factor 3
CI                                                          I )
C..J~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~C
o S , I ° 1S AX ° 0
-2      0      2       4                        -2       0       2      4                       -6   -4   -2    0        2
factor 4                                        factor 5                                        factor 6



Table 12: Regression of log(earnings) on human/physical capital and possible segmentation variables
need more credit           stay/leave          migrant status        voluntariness
i       ii           iii        iv           v        vi          vii       viii
Education                0.04       0.04       0.04        0.04      0.04       0.04        0.05       0.05
(4.29)    (3.76)    (4.27)    (3.70)    (4.39)    (3.20)    (4.74)    (5.05)
Experience               0.03       0.03       0.03        0.03      0.03       0.06        0.03       0.04
(6.1)      (5.87)    (5.92)    (5.90)    (6.17)    (4.20)    (6.55)    (6.61)
Exp2 (104)               -5.14      -5.35      -5.06       -5.52     -5.19      -4.29       -5.23      -5.45
(-7.82)   (-7.69)   (-7.73)   (-7.94)   (-7.89)   (-5.49)   (-8.02)   (-7.72)
Ed*Exp (104)             -6.20      -6.25      -5.77       -4.69     -0.06       -3.16      -7.58      -0.001
(-2.13)   (-2.02)   (-1.99)   (-1.54)   (-2.16)   (-0.89)   (-2.62)   (-3.23)
Capital (10-5)           1.80       1.98       1.86        1.83      1.88        1.86       1.83       1.73
(15.23)   (13.84)   (16.13)   (15.53)  (16.26)   (13.28)   (15.91)   (14.55)
Log(hours)               0.39       0.41       0.38        0.36      0.39       0.40        0.38       0.38
(19.47)   (18.97)   (19.02)   (16.84)  (19.49)   (16.88)   (19.29)   (17.35)
Intercept                4.58       4.53       4.66        4.72      4.57       4.62        4.59       4.55
(36.49)   (34.16)   (37.14)   (34.82)  (36.38)   (31.97)   (36.91)   (33.91)
Dummy'                   0.10       0.29       -0.21       -0.58     0.001      -0.34       -0.24      0.22
(3.42)    (0.68)    (-7.11)   (-1.63)   (0.07)    (-1.09)    (-9.16)   (0.61)
Dummy interactive
Education                          0.04                  0.03                  0.03                   -0.05
(1.33)                (0.89)                (1.13)                 (-1.73)
Experience                         0.004                 -0.02                 0.03                   -0.02
(0.21)                (-1.02)               (2.16)                (-1.61)
Exp2 (104)                         1.27                  4.50                  -3.63                  2.35
(0.58)                (2.09)                (-2.27)               (1.27)
Ed*Exp (104)                       -5.85                 -9.58                 -10.05                 16.56
(-0.61)               (-0.96)               (-1.50)               (2.00)
Capital (10-5)                     -0.58                 0.03                  0.06                   1.61
(-2.29)               (0.62)                (0.26)                (3.46)
Log(hours)                         -0.15                 0.11                  -0.04                  0.01
(-0.25)               (1.84)                (-0.87)               (0.14)
R2 adjusted              0.19       0.19       0.19        0.20      0.19       0.19        0.20       0.20
Fvalue"2                 11.70      4.50       50.59       10.19    0.016       0.93        83.88      14.60
Numbers in parentheses are t-statistics. Number of observations = 4426 for each regression. l The dummy corresponds to the
segmentation variable being tested in each column. It takes a value of I if credit is a constraint (regressions , ii), if the firm
owner wants to shutdown his business (regressions iii, iv), if the firm owner migrated as an adult (regressions v, vi), and if the
firm owner involuntarily entered the small firm sector (regressions vii, viii), 0 otherwise.2 where the degrees of freedom for the
regressions with only a dummy term (i, i, v, vi) are (1, 4418) and the degrees of freedom for the regressions with all the dummy
interactive terms (ii, iv, vi, viii) are (7, 4412).
35



Table 13: Cluster Centers
All Variables
Cluster I    Cluster 2    Cluster 3    Cluster 4    Cluster 5
Education           0.54        0.55        0.47        0.56         0.36
Experience          0.41        0.50        0.47        0.49         0.59
Firm life           0.16        0.30        0.30        0.27         0.27
Voluntary           0.64        0.87        0.98        0.89         0.18
Hours               0.42        0.52        0.43        0.54         0.41
K/L                 0.11        0.24        0.08        0.19         0.08
Earnings            0.25        0.44        0.31        0.38         0.23
Site                0.25        0.68        0.10        0.85        0.16
Workers             0.10        0.37        0.09        0.26         0.07
Benefits            0.03        0.15        0.02        0.08         0.02
Tax                 0.16        0.60        0.05        0.63         0.09
Move/stay           0.00        0.95        1.00        0.99         1.00
Expand              0.00        0.68        0.58        0.64         0.56
Clients             0.21        0.33        0.19        0.29         0.20
No problems         0.11        0.09        0.20        0.19         0.18
Credit problems     0.94        0.81        0.96        0.92         0.96
Financing (initial)  0.08       0.24        0.05        0.11         0.07
Financing (new)     0.06        1.00        0.01        0.00         0.06
%                   10.48       9.08        38.05       25.62        16.76
n                   464         402          1684        1134        742
Table 14: Distance between Final Cluster Centers
All Variables
Cluster 1    Cluster 2    Cluster 3    Cluster 4
Cluster 1    0.00
Cluster 2   1.71        0.00
Cluster 3    1.24       1.37        0.00
Cluster 4   1.46        1.05        1.00        0.00
Cluster 5   1.27        1.49        0.82        1.20
36



Appendix: Construction of Variables
A. Personal Characteristics:
1. Education. The responses were categorized as count variables where 0 is no education,
1 is one year of education, and so forth up to 12 is completed high school. Those with more than
a high school diploma were omitted since we are primarily concerned with the relatively less
educated entrepreneurs.
2. Work Experience of the Entrepreneur. (Experience) This is total work experience in
any sector and is calculated as age - 6 - years of education.
3. Why initiated enterprise. (Voluntary) This variable is intended to capture the how
voluntarily the entrepreneur entered the sector by using the motivational responses. The variable
is coded into three categories: involuntary (to supplement family income, couldn't find other
job, fired from last job) coded at 0, semi-voluntary (family tradition), coded at 0.5, and voluntary
(to be independent, to earn a higher salary than possible if employed, flexible working hours)
coded at 1.
B. Characteristics of the firm:
1. Firm life. Measured in years. The data showed a vast number of very young
enterprises and then an extremely long tail encompassing successful long-lived firms. Simply
normalizing the data compressed 90% of the observations into the range [0,0.39]. The apparent
lack of variation diminishes the power of this variable to drive clusters. Thus, the values were
recoded into deciles in increments of 2.5 years with an upper bound of 1 representing all
individuals with firm lives greater than 25 years.
2. Capital-Labor Ratio. (K/L) This variable divides the reported total capital stock of the
firm by the total number of workers in the firm. This variable also showed some extreme
outliers so the top 5% were omitted.
3. Earnings Profits net of all expenses. The existence of extreme outliers meant that on
normalization, 95% of the sample is in the range [0,0.04] leaving little variance for the bulk of
the sample around which to factor or cluster. By omitting the top 5%, approximately 95% of the
new sample is in the range [0,0.71].
4. Permanence of Worksite (Site) takes a value of 1 if the firm has a permanent locale,
0.5 if the firm can be moved, but it has a specific work site as recognized by the other firm
owners and its customers and 0 if the firm does not have a permanent work site.
5. Hours Worked (Hours) is average number of hours worked per week.
6. Total number of workers (Workers) counts the number of workers (including the
owner) working in that firm. Since 68.1% of the firms are self-employment (1 worker) and
86.7% of the firms have only one employee while 0.1% of the firms have eight employees, the
37



top end of the distribution is collapsed into the value for the upper bound, thus stretching left side
of the distribution.
C. Entrepreneurial Dynamics:
1. Permanence in the Sector (Move/stay) takes a value of I if the entrepreneur reports
that he plans to remain in the sector, and 0 if he plans to look for salaried work.
2. Plans to Expand (Expand) takes a value of 0 if the Move/stay variable takes a value of
0, 1 if the entrepreneur plans to expand the line of products, undertake new investments or hire
more workers, and 0.5 if he plans to stay in business, but not expand.
3. Clients captures the degree of personalism clientele. The responses were categorized
into three groups: large formal clients (large businesses, large factories) coded as 1, smaller less
formal clients (small businesses, small factories or workshops) coded as 0.66, informal clients
(the public) coded as 0.33 and family are coded as 0.
4. Initialfinancingforfirm (Financing-initial) captures where the funds to start the
business were obtained. There were nine possible responses in the survey that were reclassified
into three general categories: formal credit markets (banks, or savings banks) coded as 1, semi-
formal credit markets (credit from clients, credit from suppliers) coded as 0.5, and informal credit
markets (loans from friends or relatives, private loans, savings, severance payments) coded as 0.
5. Demandfor additionalfinancing (Financing-new) captures whether the firm has
sought additional financing from any source.
6. Financing problems. (Credit Problems) codes how problematic credit is. It takes a
value of 1 if credit is not cited as a business problem, 0.66 if credit is the third most pressing
problem, 0.33 if it is the second most important problem, and 0 if credit is the primary problem
in the firm.
7. No Problems takes a value of 1 if the entrepreneur states that his firm is free of
problems and a 0 if the owner does not face market, institutional, or resource problems.
D. Participation informal market and legal institutions
1. Registration with Tax Authorities (tax) This variable could take a value of 0 if not
registered at all, 0.5 if registered with either the state or municipal authorities or with the federal
treasury, and 1 if both. Though often used as the critical definition of formality, we treat is as
only one of several firm characteristics.
2. Compliance with Labor Legislation. (Benefits) By law, all salaried workers are to be
registered in the national health benefits program. The fraction of covered workers is a measure
of one dimension of "formality." However, because the vast majority of the sample are self-
employed and have no employees (65%) and entrepreneurs are not required to pay IMSS for
themselves or their unpaid workers, this variable is relevant to only a fraction of the
observations.
38



VII. References
Anderberg, M.R. (1973), ClusterAnalysisfor Applications, Academic Press, NY.
Anderson, K. H., J. S. Butler and F. A. Sloan (1987), "Labor Market Segmentation: A Cluster
Analysis of Job Grouping and Barriers to Entry. " Southern Economic Journal, 53:3 January.
Aronson, R.L.(1 99 1), Self-Employment, A Labor Market Perspective. ILR Press, Ithaca.
Balan, J. and H.L. Browning, E. Jelin (1973), Men in a Developing Society, Institute of Latin
American Studies, U.T. Austin.
Beneria, L.(1989) Subcontracting and Employment Dynamics in Mexico City, in Portes et al.,
The Informal Economy, Johns Hopkins.
Castells, M. and A. Portes (1989), "World Underneath: The Origins, Dynamics, and Effects of
the Informal Economy," in A., M. Castells, and L. A. Benton, eds., The Informal Economy:
Studies in Advanced and Less Developed Countries, Baltimore, Maryland: Johns Hopkins
University Press.
Davis, S. J., J. Haltiwanger, and S. Schuh (1994), "Small Business and Job Creation: Dissecting
the Myth and Reassessing the Facts," in L. C. Solmon and A. R. Levenson, eds., Labor Markets,
Employment Policy, and Job Creation, Boulder, Colorado: Westview Press.
de Soto, H. (1989), The Other Path: The Invisible Revolution in the Third World, (New York:
Harper and Row).
Dunne, T., M. J. Roberts, and L. Samuelson (1989), "The Growth. and Failure of U.S.
Manufacturing Plants," Quarterly Journal of Economics, 104:4, 671-698.
Elizondo, N. (1992), "Illegality in the Urban Informal Sector of Mexico City," in V.E. Tokman,
ed., Beyond Regulation, The Informal Economy in Latin America, Lynne Rienner, Boulder,
Colorado.
Esfahani, H. and D. Salehi-Isfahani (1989), "Effort Observability and Worker Productivity:
Towards and Explanation of Economic Dualism," Economic Journal, 99, 818-836.
Evans, D. S. and L. S. Leighton (1989), "Some Empirical Aspects of Entrepreneurship,"
American Economic Review, 79:3, 519-535.
Everitt, B. (1974) Cluster Analysis, John Wiley and Sons, NY.
Fairlie. R. W. and B.D. Meyer, "Trends in Self-Employment Among White and Black Men:
1910-1990," Mimeo, U.C. Santa Cruz.
39



Fields G. S. (1990), "Labor Market Modeling and the Urban Infonnal Sector: Theory and
Evidence," in OECD, The Informal Sector Revisited, Paris.
Flatau, P. R. and Philip E. T. Lewis, Segmented Labour Markets in Australia, Applied
Economics, 25, 285-294.
Gregory, P. (1986), The Myth of Market Failure: Employment and the Labor Market in Mexico,
Baltimore: World Bank and Johns Hopkins.
Harman, H.H. (1976), Modern Factor Analysis, 3rd ed, University of Chicago Press, Chicago.
Harris, J.R. and M. P. Todaro (1970), "Migration, Unemployment, and Development: A Two
Sector Analysis," American Economic Review, 60:1, 126-142.
Hart, K. (1972), Employment, Income and Inequality: A Strategy for Increasing Productive
Employment in Kenya, Geneva, ILO.
Jovanovic, B. (1982), "Selection and Evolution of Industry," Econometrica, 649-670.
Kim, J. and C. W. Mueller (1978), Statistical Methods and Practical Issues, Sage Publication,
Beverely Hills.
Levenson, A.R. and W. F. Maloney (1997), "The Informal Sector, Firm Dynamics and
Institutional Participation" Mimeo, University of Illinois and the Milken Institute.
Loayza, N.V. (1994) "Labor Regulations and the Informal Economy," Policy Research Working
Paper 1335, World Bank, August.
Lopez-Castano, H. (1990) "Inestabilidad Laboral y Ciclo de Vida en Colombia," (Labor
Instability and the Life-Cycle in Colombia) Coyuntura Economica, 20:1, p 173-191.
Lucas, R. E., Jr. (1978), "On the Size Distribution of Business Firms," Bell Journal of
Economics, 9:2, 508-23.
Maloney, W. (1996) "Are LDC Labor Markets Dualistic?" University of Illinois Working Paper.
Maloney W. (1997) "Labor Market Structure in Mexico: Time Series Evidence on Competing
Views" University of Illinois Working Paper
Marcouiller, D. V., Ruiz de Castilla and C. Woodruff (forthcoming), " Formal Measures of the
Informal Sector Wage Gap in Mexico, El Salvador, and Peru", Economic Development and
Cultural Change.
40



Marquez, C. and J. Ros (1990), "Segmentacion del Mercado de Trabajo y Desarrollo Economico
en Mexico," (Labor Market Segmentation and Economic Development in Mexico) El Trimestre
Economico, Abril-Junio.
Mazumdar, D. (1981), The Urban Labor Market and Income Disitribution, Oxford University
Press, New York.
Mazumdar, D. (1983). "Segmented Labor Markets in LDC's." American Economic Review, 73, 2
254-259.
Oster, G. (1979), A Factor Analysis Test of the Theory of the Dual Economy, Review of
Economics and Statistics, 33-41.
Ozorio de Almeida, A.L., and S.E.M Graham and L. F. Alves (1994) "Poverty, Deregulation and
Informal Employment in Mexico," ESP Working Paper, World Bank, July.
Peattie, L. (1982), "What is to be Done with the 'Informal Sector'?: A Case Study of Shoe
Manufacturers in Colombia" in H. Safa ed, Towards a Political Economy of Urbanization in
Third World Countries, Delhi:Oxford University.
Peattie, L. (1987) "An Idea in Good Currency and How It Grew: The Informal Sector," World
Development, 15:7 851-860.
Portes, A., M. Castells, and L. A. Benton, eds. (1989), The Informal Economy: Studies in
Advanced and Less Developed Countries, (Baltimore: Johns Hopkins University Press).
Portes, R. and R. Schauffler (1993), "Competing Perspective on the Latin American Informal
Sector," Population and Development Review, 19:1, 33-59.
Rauch, J. E. (1991), "Modeling the Informal Sector Fornally," Journal of Development
Economics, 35:1, 33-47.
Roberts, B. (1989) "Employment Structure, Life Cycle and Life Chances: Formal and Informal
Sectors in Guadalajara," in Portes et al., The Informal Economy, Johns Hopkins.
Secretaria del Trabajo y Prevision Social (1993), El Sector Informal en Mexico (The Informal
Sector in Mexico), no 2.
Sloane, P. J., P. D. Murphy, I. Theodossiou and M. White (1993), "Labour Market
Segmentation: a Local Labour Market Analysis using Alternative Approaches." Applied
Economics, 25, 569-581.
Spath, H. (1980), Cluster Analysis Algorithms for Data Reduction and Classification of Objects,
John Wiley and Sons, NY.
4'



Stiglitz, J. E., "Alternative Theories of Wage Determination and Unemployment in LDC's: The
Labor Turnover Model," Quarterly Journal of Economics, 88:194-227.
Tokman, V.E. (1989), "Policies for a Heterogeneous Informal Sector in Latin America," World
Development, July.
Tokman, V.E. (1992) Beyond Regulation, The Informal Economy in Latin America, Lynne
Rienner, Boulder.
Turnham D. and D. Erocal (1990), "Unemployment in Developing Countries, New Light on an
Old Problem," Technical Paper, OECD Development Centre.
U.S. Dept of Labor (1992) The Informal Sector in Mexico.
U.S. Dept of Labor (1993) Work Without Protections: Case Studies of the Informal Sector in
Developing Countries.
World Bank (1995), "Labor and Economic Reforms in Latin America and the Caribbean,"
Regional Perspectives on World Development Report.
42



Policy Research Working Paper Series
Contact
Title                            Author                   Date              for paper
WPS1 979 Banking on Crises: Expensive      Gerard Caprio, Jr.      September 1998      P. Sintim-Aboagye
Lessons from Recent Financial                                                38526
Crises
WPS1980 The Effect of Household Wealth     Deon Filmer             September 1998      S. Fallon
on Educational Attainment:       Lant Pritchett                              38009
Demographic and Health Survey
Evidence
WPS1981 Evaluating Public Expenditures     James E. Anderson       September 1998      L. Tabada
When Governments Must Rely       Will Martin                                 36896
on Distortionary Taxation
WPS1982 Analyzing Financial Sectors in     Alan Roe                September 1998      D. Cortijo
Transition: With Special Reference   Paul Siegelbaum                         84005
to the Former Soviet Union       Tim King
WPS1983 Pension Reform in Small Developing Thomas Charles Glaessner September 1998     M. Navarro
Countries                        Salvador Valdes-Prieto                      84722
WPS1984 NAFTA, Capital Mobility, and       Thomas Charles Glaessner September 1998     M. Navarro
Mexico's Financial System        Daniel Oks                                  84722
WPS1985 The Optimality of Being Efficient:    Lawrence M. Ausubel  September 1998      S. Vivas
Designing Auctions               Peter Cramton                               82809
WPS1986 Putting Auction Theory to Work:    Paul Milgrom            September 1998      S. Vivas
The Simultaneous Ascending Auction                                           82809
WPS1987 Political Economy and Political    Ariel Dinar             September 1998      F. Toppin
Risks of Institutional Reform in  Trichur K. Balakrishnan                    30450
the Water Sector                 Joseph 'Wambia
WPS1 988 The Informal Sector, Firm Dynamics, Alec R. Levenson      September 1998      T. Gomez
and Institutional Participation  William F. Maloney                          32127
WPS1989 Contingent Government Liabilities:   Hana Polackova        October 1998        A. Panton
A Hidden Risk for Fiscal Stability                                           85433
WPS1990 The East Asia Crisis and Corporate  Michael Pomerieano     October 1998        N. Dacanay
Finances: The Untold Micro Story                                             34068
WPS1991 Reducing Air Pollution from Urban    Mark Heil             October 1998        R. Yazig,
Passenger Transport: A Framework  Sheoli Pargal                              37176
for Policy Analysis
WPS1992 The Present Outlook for Trade      John Croome             October 1998        L. Tabada
Negotiations in the World Trade                                              36896
Organization



Policy Research Working Paper Series
Contact
Title                            Author                  Date              for paper
WPS1993 Financial Safety Nets and Incentive  Philip L. Brock      October 1998        K. Labrie
Structures in Latin America                                                 38256
WPS1994 Estimating Wealth Effects without   Deon Filmer            October 1998       S. Fallon
Expenditure Data - or Tears:     Lant Pritchett                             38009
with an Application to Educational
Enrollments in States of India
WPS1995 What Macroeconomic Policies       Mansoor Dailami         October 1998        B. Nedrow
Are'Sound?'                      Nadeem ul Haque                            31585
WPS1996 Namibia's Social Safety Net:       Kalinidhi Subbarao      October 1998       P. Lizarondo
Issues and Options for Reform                                               87199
WPS1 997 On Measuring Literacy             Kaushik Basu           October 1998        M. Mason
James E. Foster                            30809
WPS1998 The Structure and Determinants of   Sudharshan Canagarajah  October 1998      A. Garscadden
Inequality and Poverty Reduction  Dipak Mazumdar                            38400
in Ghana, 1988-92                Xiao Ye