Female workers undercounted

FEMALE WORKERS UNDERCOUNTED
THE CASE OF LATIN AMERICAN AND CARIBBEAN CENSUSES

Zulma Recchini de lattes*

Catalina H. Wainerman**

Working Paper No. 12 June 1982

FEMALE WORKERS UNDERCOUNTED
THE CASE OF LATIN AMERICAN AND CARIBBEAN CENSUSES

Zulma Recchini de lattes*

Catalina H. Wainerman**

Working Paper No. 12 June 1982

United Nations, New York

** Centro de Estudios de Poblacion (CENEP), Buenos Aires, Argentina

TABLE OF CONTENTS

Page

PRESENTATION

FROM CHAPTER 1: Introduction

FROM CHAPTER 3:

The Labor Force Approach. Its application
in population censuses and household surveys 8

Quantitative evidence of censal omission in the
measurement of the participation of women in
economic activity

FROM CHAPTER 4: Evaluation and Suggestions

Validity of the collection of information on
"activity status"

Elaboration of measures that partially overcome the
validity problem of censal information on
"activity status"

REFERENCES

Suggestions and Initiation of the New Stage

ANNEX:

-2-

PRESENTATION

The English speaking public is offered here selected excerpts from the

book by Catalina H. Wainerman and Zulma Recchini de Lattes El trabajo femenino

en el banquillo de los acusados. La medicion censal en America Latina (Women's

work on trial. Censal measurement in Latin America), published in Mexico in

1981 by Terra Nova*. This publication was possible thanks to the initiative of

the program of the Social Development Division of the Economic Commission for

Latin America (ECLA) on "Integration of Women in Development," to the support

of the Regional Office for Latin America of the Population Council (Mexico),

and to the institutional support of the Centro de Estudios de Poblacion CENEP)

in Buenos Aires (Argentina). This English version, a translation of excerpts

from the above-mentioned publication, was made possible with a grant from the

Women in Development Office of USAID.

The present selection mainly draws on certain sections of Chapters 3

and 4 of the original book. More specifically, we are presenting here (i) the

results of an attempt to quantify the flaws discovered in the measurement of

female labor force by Latin American and Caribbean censuses taken around

1970;(ii) suggestions for modifications to be introduced in future censuses

with a view to improve the quality of information on female economic activity;

and (iii) suggestions for the elaboration of new measures to apply to past cen-

suses to overcome, though partially, the low degree of validity and reliability

verified in order that trends and international comparisons might be estab-

lished on a firmer basis.

*The present document was translated by Anna M. Mundigo.

- 3 -

We trust our findings will alert and trouble both the users, who are

often fully confident working with these figures, as well as those who produce

them. We hope that the dissemination of these results will call attention and

lead the groups to undertake the necessary task of improving the measurement of

female participation in the economy.

Faithfulness to the original Spanish text included keeping within this

document references to tables, chapters, annexes, which are not included here.

The Annex presented at the end of this version was especially written for it.

It contains a brief report on a new stage of the project already completed

which involved the establishment of personal contacts with people from the

Statistical Offices of five Latin American and Caribbean countries. It also

contains preliminary lines suggested for carrying on the next stage of this

research enterprise.

-4-

FROM CHAPTER 1: INTRODUCTION

The reader might wonder why this book is devoted to the measurement of

the economic activity of women rather than to that of the population, without

sex differentiation. The question is a valid one, among other reasons because

the data gathering instruments used by population censuses as well as by house-

hold surveys have up to now been designed, following international recommenda-

tions, without reference to either sex. In other words, there is no section in

the questionnaires used either in censuses or household surveys specifically

designed to probe the employment situation of women as there is, indeed, for

probing their fertility. In both, the section on employment situation refers

to males and females, and seldom are different instructions given for the one

or the other sex.

The reason justifying this apparently "sexist" work is that female

labor behavior is different from that of males and that this difference makes

for different qualities of the measurement of both sexes, more concretely for a

generalized underenumeration of the female labor force. Before elaborating on

these assertions we wish to make clear that the women we are referring to are

actually only those in the middle age-groups of the life cycle. They do not

constitute the only sector of the population undercounted as members of the

economically active population. The younger and older sectors of the popula-

tion (irrespective of sex), also tend to be ignored when carrying out economic

activities (students or retirees who in addition engage in activities classi-

fied as economic). If our work does not deal with these other groups it is,

among other reasons, because they occupy, in strictly numerical terms, a less

important place among the human resources of a society. Nevertheless, and al-

though it is not our main intention, some of the conclusions we have reached

- 5 -

and some of the recommendations we have formulated do in fact have implications

for these other sectors of the population.

To assert that female labor behavior is different from that of males is

as valid at the individual as well as the societal level. It is also valid

within as well as between countries. In the majority of known societies the

males who are in the economically active stage of the life cycle remain in the

labor market from entry to retirement, save exceptional situations such as ill-

ness, death, war or economic recession. During the most active years of the

life cycle their behavior is relatively independent of the educational level

attained, of the family situation (both in terms of the marital status and of

the number of children), of the place of residence, be it urban or rural, and

even of the economic structure of the society. In fact, in most of the coun-

tries for which records exist, the proportion of males in the labor market

among those between 25 and 54 years of age exceeds 90 percent, a figure scat-

tered throughout a wide gamut of occupations and employment status.

The main characteristic of the labor behavior of women is diversity,

instead. The homogeneity of the high level of economic participation of males

contrasts with the enormous diversity found among females within and between

countries. In effect, in the 1970's such disparate figures for female activity

could be found as an 86 percent in the Soviet Union, 46 percent in the United

States, 24 percent in Argentina, 18 percent in Brazil, and 8 percent in Egypt.

The high female economic participation in the Soviet Union is, in truth, an un-

usual phenomenon since it is more often the case that the proportion of females

not taking part in the labor market is very high. Among those who do so it is

a common experience to enter and to leave several times from the market

throughout their active lives. Theirs is a discontinued economic activity with

- 6 -

interruptions usually associated with points of a change in the life cycle:

marriage, birth of first child, school entrance of last child, etc.

It is also common among women to have part-time, sporadic, and seasonal

employment, all forms of economic participation which in general obey to the

need to make their reproductive role compatible with their productive one. On

the other hand, unlike males, the level of education attained whether there is

a husband or companion present, whether childless or with one or more children,

and whether living in an urban or rural setting does indeed make a difference

to women and a great one in terms of their probability of entering the labor

market. In general, the market selectively recruits working women among those

with the higher educational levels, without husband or companion single,

separated or divorced, and widowed and without children. Those that are

recruited are permitted access to a short gamut of occupations and to a narrow

range of occupational categories within a few sectors of the economy. In the

majority of known societies, women form the majority in occupations such as

domestic servants, nurses, and teachers which bear a similarity to their

nurturing and reproductive role. Very often their activity is carried out in

the home (as in the case of self-employed seamstresses), or not too far from

home and from their small children (as is the case with street vendors who

carry them on their backs, while offering their merchandise to passersby).

These characteristics discontinuous, seasonal, part-time work, which

is often difficult to distinguish from household activities, carried out in the

traditional sectors of the economy, in family enterprises or as own account

workers are intimately related to the sexual division of labor prevalent in

societies. Together with the effects of prejudices favoring women's reproduc-

tive role in detriment of the productive one (also present among people

-7-

responsible for the design and collection of statistics), these features make

this behavior to be poorly measured and usually underestimated by the statis-

tical registers. It must be borne in mind that these registers have been

designed to detect one type of activity, that which is carried out by the

majority of males in developed economies and which is characterized by being

carried out continuously, eight hours per day, five to six days per week, and

48 to 52 weeks per year.

The underestimation of women's labor force participation varies with

the degree of "invisibility' of certain occupations and employment statuses.

It is greater in the occupations perceived as a part or extension of household

duties, such as women who prepare lunch for the workers of an agricultural en-

terprise or that of a woman who carries out in her own home a part of a se-

quence of tasks required in the garment industry. "Invisibility" is also high

for activities that are not registered in the accounting system of an enter-

prise, for example, the "unpaid family worker" or the "selfemployed" woman in a

traditional sector of the economy. Visibility is greater, instead, in the

salaried activities of the modern sectors of the urban economy.

- 8 -

FROM CHAPTER 3: THE LABOR FORCE APPROACH. ITS APPLICATION IN
POPULATION CENSUSES AND HOUSEHOLD SURVEYS

Quantitative Evidence of Censal Omission in the Measurement of the Participa-

tion of Women in Economic Activity

Before evaluating the degree of validity of the information produced by

population censuses in relation to that produced by other kind of data collec-

tion such as household and demographic surveys, it is pertinent to recall cer-

tain basic characteristics of these operations, whether multi-purpose or not.

These operations differ in their objectives, coverage, frequency with which

they are conducted and characteristics of information gathering. All these

features cannot but affect the quality of the data they furnish.

The population census is a complicated and costly undertaking aimed at

obtaining information on relatively new characteristics of the total population

of a country at a given moment (age, sex, marital status, migration status,

educational and economic characteristics of individuals, size and distribution

of the population throughout the territory, etc.). The census, which tends to

be repeated every ten years, is normally carried out in one single day or in a

very short period of time, by means of a questionnaire which necessarily must

be brief, simple and of clear interpretation for interviewees and for the

numerous interviewers who, in Latin America, normally receive little training

and no remuneration for a task that is usually imposed upon them.

The major feature which distinguishes a survey from a population census

is that the former gathers information on a sample of the population, something

which makes the undertaking less complicated and costly. This is done by means

of a questionnaire applied by a small group of especially trained interviews

which are remunerated for a task which normally extends over a period of time

that exceeds one day. A survey is usually oriented towards the more or less

-9-

thorough investigation of some topics considered basic, hence the questionnaire

might be long and complex or short and simple. An example of the first case

would be the continuous household survey (taken once per year or more often)

designed to attain a thorough investigation of topics like employment, unem-

ployment, underemployment and the like. An example of a short and simple

questionnaire is provided by the demographic survey which tends to deal with

the same subject matters as the population census and others such as the in-

vestigation of mortality and fertility at a given moment.

Differences in objectives as well as in methodology undoubtedly grants

greater validity to the information gathered by surveys vis a vis that gathered

by censuses on the "activity status" of the population. In the case of house-

hold surveys, the utilization of several questions instead of only one (as is

almost the rule in censuses) not only increases the quantity but also the

quality of the information. In effect, when the questions are several, if the

answer to the first one of a sequence should result in an erroneous answer, the

possibility would exist to introduce corrections through the following ques-

tions, something impossible when there is only one question.

In what follows comparisons are made of measures obtained from census

enumerations and that obtained by other types of data collection. For the

reasons detailed above, information provided by household surveys and similar

data collecting operations is granted greater validity than those of the censal

type.

The specific purpose of the following pages is to provide empirical

evidence which reveals how different measurement procedures of the same pheno-

menon obtain different magnitudes of it. Two examples are given in which con-

trasts are made between censuses and household surveys. One of them is Panama

- 10 -

in 1970, a case where it was possible to differentiate between the metropolitan

area and the rest of the country. The other, also in 1970, is the State of Sao

Paulo in Brazil, a State which includes one of the largest urban agglomerations

of Latin America and which has a not insignificant rural population (around 20

percent).

It would have been desirable to analyze more than two instances of

measurements of the same universe through a census and a household survey but,

unfortunately, cases in which both operations use, if not identical, at least

comparable conceptual definitions are not abundant. (Two cases supplied by

Argentina and Chile were discarded because the minimum working period required

for a person to be classified as economically active by the census and survey

of each country differ notably.) Neither were abundant cases where the results

are published and/or are accessible. Two other pairs of measurement have been

found that involve population censuses on the one hand, and special purpose

surveys, on the other. Both cases have been included because they also contri-

bute to demonstrate the low validity of census information in relation to the

female labor force. The two cases are: Bolivia, which has information from

the population census of 1976 and from the National Demographic Survey con-

ducted a year earlier; and Costa Rica, which, in addition to the census of

1973, has a continuous register of the working population covered by the Costa

Rican Social Security System (Caja Costarricense de Seguro Social).

In none of the analyzed cases do the data rigorously correspond to the

same populations, defined in time and space, as would have been the ideal.

But, as can be seen in Table 1, they are reasonably equivalent. The most

important differences are to be found in the period in which they were con-

ducted. While in the four cases the censuses were taken on one single day, or

Table 1
Some Characteristics of the Sources of Data Contrasted for Measuring Female Economic Activity

Place Source Date and time Spatial Refbrence Minimum time Size Sample
used for enu- coverage period (percent) total N
meration

Panama Census 5/10/1970 (1 day) Entire Week prior Not specified except Universe Universe
Total, Metro- country to census for non remunerated
politan Area, day family workers (1/3
Rest of of working time)
country
Survey of -Metropolitan Area in Entire Week prior Not specified for 5-6 13 000
labor force a continuous way during country a/ to interview all occupational households
entire year. categories
-Rest of country:
First week of each quarter

Sao Paulo Census 9/1/1970 (1 day) State of 1 year (9/1/69 Not specified 25 (particular- 3 333
Sio Paulo to 8/31/70) households or thousand per-
families making sons
up a collective
household)

Household
Survey Ist. quarter 1970 State of Employed: week Not specified 0.1 17 201
(1 quarter) Sao Paulo prior to sur-
vey. Unemployed:
last two months

Bolivia Census 9/29/1976b/ Entire Week prior to Not specified 3.3 139 434
country census (asks for the
greater part)
Demographic June-October Week prior to Not specified X 1.2 52 293
survey 1975 survey
(4 months)

Costa Rica Census 5/14/1973b/ Entire Week of May 1 hour a week for Universe Universe
country 7 to 12 all occupational
categories
Caja Costa- June 1973/ Entire d/ Not existing Not existing Universe Universe
rricense de country-
Seguro Social

a. Particular households; native Indian population is excluded and also those residing in the Canal Zone.

b. There is no information regarding length of time for the enumeration (one day or more),

c. Exact details of the enumeration are unknown. From the introduction preceding the statistical
information reproduced in Appendix A it is assumed that it is a monthly declaration and
therefore it refers to a month.

d. Refers only to managers, wage-earners and so-ne self-employed and members of cooperatives. It
is estimated that in 1973 Social Security covered 49 percent of the economically active population.

- 11 -

in a period presumed to be very short (Costa Rica and Bolivia), the surveys

were invariably conducted during a much longer period: three months in Sao

Paulo; four months in Bolivia with almost one year's difference between the

time when one and the other were conducted); throughout the entire year or in

each quarter, as in Panama, depending on whether reference is made to the

Metropolitan Area or the rest of the country. Although very drastic changes in

behavior cannot be expected within such narrow time-spans, the fact that the

data collection was undertaken in different periods within the year can indeed

be of importance due to the possible presence of seasonal work activities both

in Bolivia a predominantly rural country as well as in Sao Paulo and

Panama, above all outside their metropolitan areas.

Appendix Al reproduces the measurement instruments just as they

appear in the census schedules, in the instructions for census enumerators, in

the questionnaires, and in the definitions used by the household surveys and by

the Costa Rican Social Security System. The degree of divergence between the

instruments used for each case of comparison varies both with respect to con-

cepts as well as data collection techniques. The smallest divergence is found

between the instruments used by the two Bolivian sources; the greatest, between

the Costa Rican ones; and in between, although tending to be very divergent,

those between the sources of Panama and, above all, those of the State of Sao

Paulo. There is no information regarding the type of training given to the

enumerators in the different cases but it is undoubtful that it is more intense

and of better quality in the surveys than in the censuses.

1 These documents are not attached here but are part of the original
Spanish version, available on request from Judith Bruce, The Population
Council, One Dag Hammarskjold Plaza, New York, New York 10017.

- 12 -

a) Comparison of population census versus household survey: Panama

The conceptual definitions of the economically active population em-

ployed by both Panamanian sources contrasted are very similar. The population

census as well as the labor force survey include among the active both the

employed and the unemployed, with practically identical descriptions about who

should be included in each of those categories. Both used a one week time-

reference period, and both leave undefined the portion of that week that a

person is required to have worked in order to be counted as economically

active, with the sole exception of unpaid family workers in the census. For

this category the census requires that the person must have worked at least

one-third of a normal work week to be considered active. No requirement is

established in the survey; that is, the time is left undefined, just as it is

for the other categories. This would be the only source of conceptual discre-

pancy between the census and the labor force survey. Since this category has

little weight on the total female labor, this difference would account for only

a minor part of the differences that exist between these two measurements of

the economically active female population. According to the census, only 4.3

percent of the total active females are unpaid family workers, the correspond-

ing figure according to the labor force survey is 5.6 percent. That is, assum-

ing that all the other categories would have been measured with the same degree

of accuracy by both sources which of course is not the case, as will be seen

further on the change of conceptual definition would only give rise to a 1.3

point difference.

Whereas the conceptual aspects are similar, the technical ones concern-

ing the information gathering tools are markedly different. As can be seen in

Figures 1 and 3 of Appendix A, only once in the census questionnaire is an in-

- 13 -

quiry made about the different economic activity alternatives for the reference

week ("worked", "looked for work having worked before", "looked for work not

having worked before"). In the labor force survey queries were twice made

about work in addition to the formulation of a third question regarding

whether a person has an employment or business whether working or not during

the previous week and twice about having searched for work. The use of a

reiterative instrument greatly increases the probability that the survey will

record more of the active population than the census. Added to this is the

fact that the enumerators of a survey operation are better instructed than

those of a census and that their performance is likely to be superior because

interviewing is a stable and renumerated job.

There are other reasons for possible discrepancies between both

measurements. While the census collected information in one single day, the

survey did it (alternatingly or not according to the area) over the course of

the entire year. Little can be said about the possibility of a survey discern-

ing seasonal labor since there are insufficient test elements for evaluating

whether in the "rest of the country" (that which is predominantly rural) the

first week of each quarter coincides or not with seasonal tasks. From the sur-

vey report it is also not clear whether four data collections were made and

then averaged or whether in each quarter a different region was surveyed. How-

ever, it seems sensible to assume that the survey has a greater probability of

detecting seasonal employment than the single day census operation.

Other possible sources of divergence between the two measurements

reside in the inclusion or exclusion of the native Indian populations and in

the lowest age limit employed which determines the starting point for the

- 14 -

applicability of questionnaires. But these have been compensated through the

selection of adequate tabulations that include similar populations.

The comparison between activity rates by sex and age calculated from

census and survey data reveals very clear results: the survey registers a

greater number of active persons than the census, as can be seen in Table 2.

(The only exceptions occur among the youngest and oldest women of the Metro-

politan Area. These cases might be the result of sample errors given that one

is dealing, undoubtedly, with groups of lesser frequency among the active popu-

lation.) The results coincide with what is expected according to the quality

of the instruments utilized.

But the differences are not the same for both sexes. Though signifi-

cant, they are not too great in the case of males; in turn, they are enormous

for females. That is, the quality of the collecting tool seems to have had

little effect on the measurement of the male labor force but it has indeed

affected that of women even in an urban setting as is the case of the Metro-

politan Area of Panama, where the recording of the active population is more

valid generally.2 The differences between the rates of femal participation

of the two sources (with the exceptions already pointed out) are not below 5

percent. Indeed, they usually go beyond 10 percent and as high as 43 percent.

The conclusion is clear and immediate: the population census did not result in

a valid tood for measuring the economic participation of women, both in an

urban population (the Metropolitan Area), and in a predominantly rural one

(rest of the country).

2 The bibliography on the economically active female population is insis-
tent in pointing out that underenumeration is greater for agricultural and
unpaid family workers, less frequent in urban than in rural areas.

Participation

TABLE 2
Ratesa/by Sex and Age Derived from the Census and the Labor Force Survey of Panama
(Total country, Metropolitan Area and the rest), 1970

WOMEN MEN
Age Survey Census S Survey Census S
-. 100 100
C C

Total country
15-19 32.9 31.3 105.0 63.5 60.0 105.7
20-29 44C3 39.6 111.9 95.6 94.8 100.8
30-39 41.1 33.6 122.8 99.2 97.4 101.9
40-49 35.8 31.2 114.9 98.4 96.4 102.1
50-59 30.2 23.7 127.4 95.0 91.6 103.7
60 and + 11.5 10.3 111.5 65.6 61.1 107.3
Metropolitan Area
15-19 35.8 36.0 99.5 48.5 46.8 103.5
20-29 57.2 50.8 112.6 94.3 93.2 101.5
30-39 56.2 44.9 125.1 99.2 96.6 102.7
40-49 48.2 42.2 114.2 98.4 95.3 103.2
50-59 39.1 32.1 121.8 94.2 89.0 105.9
60 and + 13.0 13.3 98.0 52.6 47.7 110.2

Rest of the country
15-19 29.5 25.6 115.4 77.0 72.3 106.5
20-29 28.8 25.8 111.7 97.0 96.5 i00.5
30-39 24.9 21.7 115.0 99.2 98.1 101.1
40-49 23.2 18.6 124.9 98.4 97.5 100.9
50-59 19.2 13.7 140.1 95.8 94.3 101.6
60 and + 10.0 7.0 143.3 75.3 72.6 103.8

a. The participation is the quotient between the economically active population and
the corresponding total population of each sex and age-group, times one-hundred.

SOURCES: Panama (1972), Table 5. Panama (1976), Table 30 of Vol. III and Table 6 of
Vol. IV.

- 16 -

b) Comparison of population census versus household survey: Sao Paulo

The case of Sao Paulo (Brazil) is entirely different from that of

Panama as to the degree of conceptual and technical divergence. The census

includes within the economically active population persons having worked during

the twelve months previous to the date of the census, even if at the census

date they were unemployed, on leave or on vacation, or imprisoned "awaiting

trial", and also persons aged ten years and over who at the date of the census

were for the first time looking for a job (Brazil, IBGE 1973, pages xxxi and

xxxii).

The household survey considers that the economically active population

is constituted by the totality of employed persons (those who in the reference

week were working and those who, although not working, had some employment or

business from which they were temporarily absent for reasons such as illness,

vacation, bereavement, strike, etc.), plus the unemployed (those not employed

in the reference week but who in the last two months had taken some steps to

seek work).

These definitions differ in the reference period and, therefore, in the

concept of economically active population. While the census adopts one year,

the survey adopts one week for the employed and two months for the unemployed.

Given the characteristics of the participation of women, this discrepancy would

lead one to expect that the recording of that population would be larger in the

census than in the survey.

As to the techniques of collection, those of the census and the survey

are absolutely different. Criticism has already been made in a previous sec-

tion of the collection instrument used by the Brazilian census: the question

apparently directed at measuring the "activity status" contains a double nega-

- 17 -

tive, an indeterminate reference period even though in the publication of the

censal results it is said that the reference period is of one year, this is not

mentioned either in the censal schedule or in the manual for enumerators and

a set of response categories which begins with the alternatives of economic in-

activity. Further on another two questions are formulated to distinguish be-

tween employed and unemployed, whose reference period in one case is the last

week, and in the other, one year. The instructions in the manual for the cen-

sus takers do not clarify either the sense of the questions nor to whom these

should be asked, as can be seen in Figure 6a of Appendix A. It is not known

what kind of training the enumerators received but faced with such a compli-

cated and confused collection instrument it is doubtful, even if much energy

was invested in the training process, that the replies would be superior to the

formulation of the questions themselves.

The survey, which follows the outlines of the Atlantida model (see Fig.

7 of Appendix A) devotes several questions towards the identification of econo-

mically active persons, so that whoever was erroneously classified as inactive

in the first question, can be correctly classified as economically active at

the second or third opportunity when questioned on the same subject.

Based entirely on the instrument (i.e., disregarding conceptual as-

pects), it seems clear that the survey's questionnaire is a more valid measur-

ing instrument than the census. Therefore, a greater enumeration of active

persons would be expected by the survey, above all among women, since in addi-

tion to the reason pointed out earlier, in this census in particular, the first

alternative offered each person was "household chores". However, based on the

conceptual aspects--e.g. length of adopted reference period--one would expect a

greater registration of active persons by the census than by the survey. If

- 18 -

the comparison of both measurements were to show that the survey registered

more active persons than the census, the lack of validity of the Brazilizn cen-

sal instrument for measuring female economic activity would acquire a much

stronger significance.

Another source of incomparability between both measurements, just as in

the case of Panama, is the time of the year when they were carried out. While

the census was taken on 1 September 1970, the survey was conducted during the

first quarter of that same year. Since the existence of seasonal activities is

highly probable, especially in the rural part of the State of Sao Paulo, the

change of time period to which the economic activity refers could have a bear-

ing on the counting of the female labor force. But, again, the census should

enumerate more completely than the survey, because the census made reference to

activity during the last year, while it is not known whether or not the quarter

to which the survey refers corresponds to a time of heightened agricultural

activity.

The rates of participation calculated with data from the two sources

appear in Table 3. The conclusions are even more convincing than for the

Panamanian case: the survey enumerated more active persons than the census.

(the only exception is constituted by persons 65 and over, attributed to the

size of the sample, since the relative frequency of cases is very low in these

ages.) Once more, the censal underenumeration is much greater for women than

for men. The censal underenumeration in relation to the survey oscilates be-

tween 14 and 33 percent among women and between 2 and 6 percent among men.

Fortunately, the available information for Sao Paulo makes it possible

to delve deeper into the comparison by permitting separate calculations of the

activity rates for agricultural and non-agricultural workers, and for unpaid

TABLE 3
Participation Rates-/ by Sex and Age Derived from
of the State of So Paulo

the Census and
(Brazil), 1970

Household Surveys

WOMEN MEN
Age Survey Census S 100 Survey Census S 100
C C

20-24 47.4 39.3 120.6 92.8 88.5 104.9
25-34 31.7 26.5 119.6 97.9 96.0 102.0
35-44 29.0 22.6 128.3 97.1 95.3 101.9
45-54 19.4 17.1 113.5 92.8 88.2 105.2
55-64 12.6 9.5 132.6 72.7 68.6 106.0
65 and + 2.4 3.7 64.9 31.6 34.2 92.4

a. The participation rate is the quotient between the economically active population
and the corresponding total population of each sex and age-group, times one-hundred.

SOURCES: Brazil (1973), Table 21. Brazil, (n.d.), Table 3.1.2.

- 20 -

family workers and wage earners. The results shown in Table 4 clearly demon-

strate that, at least in this case, the population census is not a good enumer-

ator of the female labor force in general. Furthermore, the census is especi-

ally inadequate for women working in agricultural activities and for unpaid

family workers of both sexes. In these categories too, the divergence is

greater with regard to women than to men. In fact, the figures are very im-

pressive, although it might be argued that the survey's sampling error might be

great in these relatively small groups. But the differences, in addition to

being large, are consistent with each other and also with the results obtained

by Pecht (1974) in his analysis of data from Brazilian population and agricul-

tural censuses, both taken in the same period. Therefore, the census of Sao

Paulo supplies clear evidence that the Latin American population censuses

significantly underenumerates active women and especially so those employed in

agricultural activities and as unpaid family workers (within or outside the

agricultural sector). It gives further evidence, though to a somewhat lesser

degree, of a high underenumeration of males working as unpaid family workers.

c) Comparison of population census versus demographic survey: Bolivia

The general methodology of a demographic survey such as the one con-

ducted in Bolivia in 1975 is in many aspects similar to that of a population

census. It coincides with the latter in the use of a short and simple ques-

tionnaire as well as in the variables investigated. It differs in some aspects

like that among the "economic characteristics" it only investigates the "activ-

ity status" and not other topics common in censuses (such as "occupation", "in-

dustry" and "employment status"), and it delves deeper into the investigation

of fertility and mortality. The most notable difference, however, is that the

- 21 -

TABLE 4
Participation Rates by Sex and Age, Agricultural and Non-Agricultural
Sector, Wage earners and Unpaid Family Workers, Derived from the
Population Census and the Household Survey of Sao Paulo (Brazil), 1970

Age Survey Census 100 Survey Census 100
C C

WOMEN

Agricultural

1.9 236.8
1.1 427.3
1.2 350.0
1.2 233.3
0.8 475.0
Unpaid family workers

1.1 318.2
0.5 1,120.0
0.4 975.0
0.4 700.0
0.3 700.0

Agricultural
19.0
18.1
19.7
22.7
23.6

89.5
102.2
100.0
101.3
97.5

Unpaid family workers

4.6
S 1.2
0.4
0.3
0.5

210.9
233.3
300.0
100.0
320.0

Non-agricultural

42.8
26.9
24.8-
16.6
8.8

42.2
23.1
21.0
13.1
6.9

75.4
79.4
77.4
69.9
49.8

77.5
72.4
67.6
60.4
38.7

37.4
25.3'
21.4
15.9
8.7
Wage-earners

36.4
23.3
18.5
13.1
6.6

Non-agricultural
69.6
77.9
75.7
65.4
45.1

Wage-earners

73.4
75.0
66.6
56.3
39.4

a. The participation rate by activity sector or occupational category is defined
as the quotient between the economically active population of each sector and
category and the corresponding total population of each sex and age-group,
times one-hundred.

SOURCES: Brazil (1973), Table 21. Brazil (n.d.), Table 3.2.2.

20-24
25-34
35-44
45-54
55-64

20-24
25-34
35-44
45-54
55-64

17.0
18.5
19.7
23.0
23.0

20-24
25-34
35-44
45-54
55-64

20-24
25-34
35-44
45-54
55-64

114.4
106.3
115.9
104.4
101.1

115.9
99.1
113.5
100.0
104.5

108.3
101.9
102.2
106.9
110.4

105.6
96.5
101.5
107.3
98.2

- 22 -

survey utilizes a sample of the population which is studied by a relatively

small number of well trained interviewers. These move about the territory

during a period of three to four months, corresponding to the data collection

phase.

The concepts of an economically active population implicit in the 1976

population census of Bolivia and in the demographic survey of 1975 are very

similar. The census definition says that the economically active population

"encompasses all persons who in the adopted reference period (the week prior to

the date of the census) were employed and unemployed" (Bolivia INE: n.d. page

21). From the available survey materials one infers a similar definition. It

includes the employed and the unemployed, and the period of reference is also

one week, although with a less precise definition than in the census as to the

amount of working time required during that period for a person to be consi-

dered active. Another small difference is that the census specifically in-

cludes the unpaid family workers whereas it is not clear whether the survey

does it or not because since the occupational category is not investigated,

this category is not mentioned.

There are some differences between the measuring instruments used by

the census and the survey both in the phrasing of questions as well as in the

instructions given to interviewers. Whereas the censal question inquired about

the economic activity during the "greater part of your time", no reference to

time exists in the corresponding item of the survey, as can be seen in Figures

9 and 11 of Appendix A. The instructions in the census, short and clear, indi-

cate only that the response alternatives must be read in the indicated order

until an affirmative answer is obtained (see Figure 12 of Appendix A). The one

included in the survey's questionnaire says that the alternatives must be read

- 23 -

in the indicated order and that the first to register a positive answer should

be marked. However, the one in the survey's manual, repeats this instruction

but also says that "all and each of the alternatives" should be read to the in-

terviewee, which is contradictory to the former. Since the instructions in-

cluded in the questionnaire are more frequently read than those included in the

manual, given that the interviewer is confronted with them each time the ques-

tion is posed, one assumes that it is the questionnaire instructions that were

actually followed by the interviewers. If this be so, it can be concluded that

both census and survey tools are similar in this sense and that both clearly

give preference to the alternative corresponding to economic activity. Al-

though undoubtedly there is room for doubt, we will continue the analysis

assuming that the two tools for collecting information (census and demographic

survey) were similar in terms of the aspects analyzed, so far.

The censal instructions in the manual for interviewers are accompanied,

as can be seen in Figure 10 of Appendix A, by a series of apparently descrip-

tive drawings of each option presented by the census: all characters repre-

senting categories of economic activity (1, 2, 3 and 4) are masculine. The

only feminine characters are those corresponding to the categories of "home-

maker" and "student". The choice of sex does not appear to be casual. Could

those responsible for the census have had in their minds that work is a pro-

vince of men, even in a country such as Bolivia with a long tradition of high

female participation in economic activities? Though not intended, it is sen-

sible to assume that the drawings may have biased the census takers.

All indicates that the demographic survey fieldwork (as well as the

household surveys) must have been done with greater care than that of the cen-

- 24 -

sus. All aspects considered, it is expected that the demographic survey would

offer more valid results than the population census.

Lastly, mention must be made of the time period in which each was con-

ducted as a possible source of incomparability. In fact, the survey was taken

between June and October 1975 and the census at the end of September of the

following year. There is a certain overlapping in the period of the year in

which both were conducted, but is is not total and thus, if the first months of

the survey would correspond to a certain intensification of agricultural activ-

ities, this would result in a relatively greater number of persons economically

active measured by the survey.

Data from the survey and the census are compared in Table 5. The male

data are shown next to the female data as a point of reference. The relation-

ship between the participation rates of the survey and those of the census for

the totality of women in Bolivia, controlling by age-groups, shows that the

survey counted between 33 and 48 percent more active women than the census.

These differences are huge if compared to those observed for the capital and

the rest of the urban areas areas for which the survey and census give par-

ticipation rates that tend to differ very little, and even give difference of a

contrary sign for some age-groups-,3 and are especially dramatic when com-

pared to those for the rural areas. In the latter areas, the rates for women

3 This result, so different from that found in the metropolitan area of
Panama is surprising. Faced with it, several questions need to be raised,
although no satisfactory answer can be found for any of them. Was the
survey in the urban areas of poorer quality than in the rural areas? Was
the census fieldwork better implemented in the urban areas? Were the urban
and rural limits defined differentially by one or other operation and, are
these the ones responsible for the results that were found?

- 25 -

TABLE 5

Participation Rates / by Sex and Age for Bolivia, total country and regions, derived
from the 1976 Population Census and from the 1975 National Demographic Survey

WOMEN HEN
Age
Survey Census 100 Survey Census .
C C

Total country
20-29 37.0 25.0 148.0 83.8 89.5 93.(
30-39 34.1 23.5 145.1 98.3 98.6 99.;
40-49 30.2 22.8 132.5 98.1 98.4 99.:
50-59 27.1 19.5 139.0 96.0 96.2 99.1

Capital City
20-29 35-4 35.3 100.3 69.0 84.2 81.
30-39 36.8 35.9 102.5 97.3 98.3 99.
40-49 28.6 33.1 86.4 96.6 98.2 98.
50-59 27.8 27.2 102.2 90.3 93.2 96.

Other urban areas-
20-29 35.1 32.3 108.7 73.5 82.0 89.
30-39 30.6 32.9 93.0 .97.7 98.0 99.
40-49 26.0 31.5 82.5 97.0 97.5 99.
50-59 23.5 23.3 100.9 94.0 92.3 101.
Rural areasN
20-29 38.2 '17.4 219.5 91.6 95.6 95.
30-39 35.0 15.7 222.9 98.8 99.0 99.
40-49 32.1 16.5 194.5 98.8 98.8 100.
50-59 28.4 15.7 180.9 97.8 98.2 99.

a. The participation rate is the quotient between the economically active population
and the corresponding total population of each sex and age-group, times one-hundred.

b. The rates corresponding to the population census are based on simple averages using
those calculated by Polo Najera by quinquennial groups, since the original infor-
mation was not available.

SOURCES: Torrez (1977), Tables 11 and 12. Polo Najera (1978), Table 4. Bolivia
(1977), Table 9 and unpublished tabulations from the National Demographic
Survey of 1975, kindly supplied by the Instituto Nacional de Estadistica
(La Paz).

- 26 -

between 20 and 39 years of ages obtained on the basis of survey data, are twice

as large as those obtained on the basis of census data and, although for women

between 40 and 59 years of age these differences are of a smaller magnitude,

they are still quite large. The rates for men calculated on the basis of in-

formation from one and the other source differ the least, even less than those

for women from urban areas. In addition, those derived from the census present

a slight tendency to exceed those from the survey.

It is possible that the large differences in economic activity found

for the female population residing in rural areas, is accounted for by the

seasonality of agricultural tasks, hence, by the different dates of the data

collection. (However, one rather tends to believe that, as with the Brazilian

and Panamanian censuses, the Bolivian also turns out to be an operation of

little validity for collecting information on the female labor force engaged in

agriculture, and therefore, in the rural areas. The results of the survey, on

the other hand, are also more consistent with the previous census of Bolivia

(1950) which showed higher rates of participation for the agricultural sector

than for the non-agricultural. The urban-rural differential shows the same

pattern according to the 1975 survey data (Recchini de Lattes: 1979), a fact

that is consistent with the conditions of life of the residents of the Bolivian

altiplano, that is, the majority of the population. The differential revealed

by 1976 census data, on the other hand, shows the opposite case, as can be seen

in Table 5, although similar to those presented (possibly in an erroneous way)

by the other countries of the region.

- 27 -

d) Comparison of population census versus continuous register: Costa Rica

These two information sources for the working population present strong

divergences, both with reference to concepts as well as to instruments.

The 1973 population census of Costa Rica defines the economically ac-

tive population in a way that is similar to other censuses of the region, in-

cluding both the employed and the unemployed. Among the former it includes

those who have worked in the reference week, for remuneration or for income, or

in a family enterprise without receiving income, or had not worked but had an

assured employment; and among the latter, those who sought work in the same

reference period, whether having or not having worked before.

The statistics from the Costa Rican Social Security System, to the con-

trary, refer not to the economically active population but to those directly

insured by the Security System, which comprise: 1) workers or wage earners

registered in payrolls by their employers; 2) persons who were wage earners but

who are self-employed at present; and 3) members of cooperatives. Some tables

using these statistics also include employers, defined as "the natural or

juridical person, private or by public right, who utilizes the services of one

or more persons covered by social security benefits" (Costa Rica: 1974a, page

v.).

It can be seen that the definitions of these different categories have

little to do with the definitions of the censal categories, even though they

respond to the same or to similar generic names (wage earners, employers) and

that these categories do not encompass the entire labor force. To point out

the most obvious differences one might say that the Costa Rican Social Security

System does not cover, and therefore, does not register: 1) self-employed

workers who never were wage earners; 2) unpaid family workers; 3) wage earners

- 28 -

not registered in the payroll by their employer;4 4) the unemployed, unless

they have registered on their own volition. Furthermore, among employers, it

does not differentiate between natural or juridical persons. In light of such

differences with the census categories for the economically active population,

it seems irrelevant to speak of a reference period, of the date of the data

collection, and even of the instruments themselves. Census data is collected

from a single item plus its instructions in the censal schedule; Social

Security data are collected from a monthly payroll on which different informa-

tion is entered for the employer and for the insured, shown in Figures 13, 14

and 15 of Appendix A.

Why are these two sources of information from different population

sectors being contrasted? Because the analysis of persons registered by these

two sources leads one to believe that the Social Security greatly underesti-

mates the economically active population enumerated by the population census,

since the latter by definition includes more worker categories. And these ex-

pectations are amply borne out in general but not in the case of active women

engaged in agricultural tasks. Therefore, this comparison demonstrates once

again that the census is not a good tool for measuring women's participation in

agriculture. In fact, as can be observed in Table 6, the Social Security

System encompasses approximately 46 percent of the total economically active

population, or 62 percent of it, if one only takes into account the salaried

workers. It appears that Social Security coverage is greater for active women

than for men, which may be due to the fact that the Social Security System --

4 Since enrollment means payment of insurance, that is, apportionment of
money on the part of the employer, it can be surmised that there might be
evasions and, therefore, important omissions from these lists of workers.
In addition, there might be employers who do not register as such.

- 29 -

TABLE 6

Proportion of Workers Covered by the Social Security System in
Relation to the Economically Active Population Registered
by the Population Census of Costa Rica, 1973 (in percent)

Sector used to calculate T l W
proportion Total Women Men
proportion

LF,* total 46 61 42
LF, wage-earners 62 67 60
LF, in agriculture 24 117 22
LF, wage earners in
agriculture 41 133 38

SOURCES- Costa Rica (1974a), Table E-9.

Costa Rica (1974b), Table 53.

* Labor Force

- 30 -

which includes maternity benefits -- reaches more satisfactorily the female

than the male population, or that active women were more poorly enumerated than

men by the population census. Despite this higher coverage, the Social

Security figure for working women is far from that of the census: it repre-

sents only a 61 percent of the latter.

Coverage by the Social Security System of the economically active popu-

lation in agriculture is lower than coverage of the total of actives, as can

well be expected in a rural setting where the difficulties for making monthly

declarations are greater due, among other reasons, to the relatively greater

lack of information and services. But what is surprising is that the direc-

tion of the difference is inverted in the case of women in agricultural activi-

ties: coverage by the Social Security System represents 117 or 133 percent of

the censal register of this sector, depending on whether all women in agricul-

ture or only salaried women in agriculture are included. These figures speak

for themselves about the censal underenumeration of women in agricultural

activities. In fact, as already mentioned, although the Social Security System

does not cover the total active population and even less so wage earning males

of the agricultural sector (as seen in Table 6), it still registers more women

in agriculture than the population census. That is, there is evidence that the

agricultural sector is one of those least covered by the Social Security System

- 38 percent of the wage earning males in agriculture compared to 60 percent of

the total salaried workers -, and that the Social Security System covers a

greater number of women in agricultural activities than were enumerated by the

population census. It can therefore be deduced that the actual number of women

working in agriculture in Costa Rica largely surpasses both the census and the

Social Security System's figures. Once more this case serves to show, for our

- 31 -

specific purpose, that the census is not a good tool for measuring female par-

ticipation in agricultural activities.

e) General conclusion

The four comparisons between the results of population censuses and

those from other independent sources of information suggests that censuses are

tools little suited for measuring female participation in economic activity,

especially in the agricultural sector or in the category of unpaid family

workers.

In the case of Sao Paulo (Brazil), where different conceptualizations

employed by each of the information sources led one to expect that the census

would show a greater relative number of persons economically active than the

household survey, the opposite was true. The same applies to Costa Rica where

the census was compared with a register containing fewer categories of workers

than are implied in the censal definition of the economically active popula-

tion. In Bolivia, where the demographic survey's instructions suffer from

serious defects, it seems that the quality of the interviewers still brought

about a better measurement of female economic activity. In the case of

Bolivia, as well as in the Panamanian part outside the Metropolitan Area ("rest

of the country"), doubts remain as to the possible influence of different data

gathering periods with possible seasonality of certain agricultural tasks.

But, if this were so, why did it not become a reason for a difference in the

measurement of males? There are enough reasons for attributing the discrepan-

cies between both sources contrasted to the low validity of the population cen-

suses for measuring the female labor force.

- 32 -

It bears pointing out that the results found in these four Latin

American cases are similar to those reported by Bancroft (1958) for the United

States. Comparisons made with conceptually equivalent data from the Current

Population Survey show that economically active women were underestimated in

the 1940 and 1950 U.S. censuses by 17 percent (based on 41 countries) in the

former census, and by 5 percent (based on the entire country) in the latter

census. The underenumeration was consistently greater among young persons of

both sexes, women in general, and especially among those employed in agricul-

ture. These results were attributed to the different quality of census and

survey interviewers: "a trained interviewer will make fewer assumptions and

will tend to classify more persons as members of the labor force than a begin-

ner" (Bancroft: 1958, page 162). The U.S. example shows that the problem of

censal underenumeration of the economic activity of women is not restricted to

Latin American countries nor to developing countries, although it might perhaps

attain a much greater magnitude in the latter due to the prevalence of tradi-

tional economic activities frequently carried out in the home environment and

alternated with and therefore often undifferentiated from household

duties.

- 33 -

FROM CHAPTER 4: EVALUATION AND SUGGESTIONS

Validity of the Collection of Information on "Activity status"

As seen in the previous chapter, the degree of validity of the measure-

ment of female labor force in household surveys is clearly superior to that of

censuses. Hence, the analysis and suggestions that are here formulated will

refer to the latter. Let us be clear that the suggestions derive from an

analysis of a very defined universe: the censuses that were taken in the

decade of the 1970's in Latin America and the Caribbean. In other words, in

this publication the conceptual definitions of "work", "activity status", "eco-

nomically active population", "unemployment", etc., which have been used in

these censuses are not discussed but rather taken as given. On the other hand,

the operational interpretations made by each country in light of those concep-

tual definitions were evaluated. The suggestions being made are, therefore,

the product of the assessment of a particular approach which does not represent

that of all countries that have taken censuses nor the only possibility. To

undertake a broader evaluation would be the object of another work, which would

transcend what has been done on this topic in the region and which would in-

clude, furthermore, a discussion of the conceptual basis on which the measure-

ments of female labor rest. The analysis was done here of the quality of the

conceptualization of the phenomenon of female labor, of the way to achieve this

measurement, and of the adequacy of the one to the other. The suggestions are

addressed at increasing the adequacy.

Three aspects were judged as requiring the greatest consideration

because of their effects on the measurement of the economic participation of

women: a) items (and instructions investigating the "activity status");

- 34 -

b) reference period with regard to which that status is investigated; and c)

minimum time required to be considered "economically active".

a) Items (and instructions) on "activity status"

The following suggestions arose from the analysis of the wording of

items and of instructions accompanying these items addressed to determine the

"activity status" of the population of active age. First of all, and following

the most basic methodological recommendations applicable to all social re-

search, it is necessary to insist that when opting for one single question (the

stimulus designed to produce the information), its wording must be clear, pre-

cise and uniform for the entire population. As to clarity, the use of the

negative form, especially if reiterated, as in the case of the census of

Brazil,5 goes against the most elemental principles for the wording of items

designed to record valid information. As to uniformity (constancy of stimuli),

it is necessary that the information on the "activity status" be obtained

through one question (or several) that is included in the censal schedule and

not by means of a heading for a column or row which forces each enumerator to

give it (at their discretion) an interrogative form.

Secondly, it is important that the presentation of the preceded re-

sponse alternatives of that item starts by those on economic activity and fol-

lows by those on economic inactivity. It is also important to instruct the

enumerator to stop the reading of the alternatives at the first answer chosen

by the interviewee. If the economic activities are not conveniently ordered or

5 The question referred to reads as follows: "If you do not work nor were
looking for a job, what situation and occupation did you have?" talics
added.)

- 35 -

if the interviewee is given the possibility of replying after knowing all the

alternatives, it is probable that, as already said, many of the active women

would be classified erroneously as inactive. This may occur because a given

culture "says" that for women the proper and suitable tasks are the domestic

ones, or because under given circumstances household tasks are considered as

more prestigious than certain economic activities and, furthermore, because in

general, economically active women tend to carry out a double role. Precisely

owing to this latter reason, it is convenient to accept the double answer which

may possibly increase the probability of detecting a considerable number of

housemakers who are, at the same time, economically active.

Thirdly, and although apparently obvious, it seems necessary to insist

that the instructions for the enumerators must not detract from and contradict

the wording of the items as was seen to occur in many cases. It is especially

recommended that it should not be suggested in the instructions manual, for

example, that household tasks are proper to women and that only as an exception

are these performed by men, or that for persons of specific ages it is most

possible that they be inactive or unemployed. Quite the contrary, the instruc-

tions should tend to do away with the stereotype that women are only devoted to

household tasks by insisting that if they carry out an economic activity, even

though at home or part-time, they should likewise be classified as active.

Up to now we have assumed the use of a single item, which was common in

the majority of the cases analyzed. But given that the item on "activity

status" in its usual form of a single question actually condenses several, one

would have consider whether it is convenient that a series of alternative an-

swers be unfolded in a sequence, thus contributing to clarify what is meant by

"work" in addition to ensuring the internal control of the answers. But this

- 36 -

is a topic that merits an investigation all of its own and which does not form

part of the present work.

b) Reference period

As several authors have already pointed out (D'Souza: 1978; United

Nations: 1978, Torrado: 1979), the choice of a short period such as a week

affects above all the female agricultural workers who only participate in

periods when those tasks are intensified. In fact, it is reasonable to assume

that the very high participation rates observed in Haiti --the highest in Latin

America and the Caribbean according to the latest available census information,

as will be seen further on-- is due to the long reference period chosen by that

country (six months). But one more step needs to be taken as part of the set

of considerations for this recommendation. Research on participation in non-

agricultural activities of some populations outside the region indicates that

female participation is characterized by several entries and exits from the

labor market, even in relatively short periods such as one year. Ostry's

(1968) study on female labor force in Canada gives quantitative evidence re-

garding its greater occupational "elasticity" relative to that of males: the

comparison between the participation rates obtained in 1961 for the week and

for the year prior to the census highlights a difference that reaches 14 per-

cent among women, but does not reach 5 percent among men. Although the infor-

mation based on one week underestimates the one based on one year for both

sexes, the effect is significantly greater for females. More adequate measure-

ment of the female labor force in the region would require extending the refer-

ence period bearing in mind the peculiarities of female participation not only

by those in which agricultural activities predominate.

- 37 -

c) Minimum activity time

The analysis of cases including instructions regarding the amount of

time required to be considered economically active revealed that only one coun-

try (Mexico) specified a minimum time in addition to including a reference

period in the item on "activity status" of the censal schedule. Of the rest of

the countries indicating a precise time -- one day, 15 hours, etc. -- the great

majority gave different instructions for "unpaid family workers". As was shown

in Chapter 3, this divergence may have confused the census taker determining

him not to take these instructions into consideration. It is here recommended

not to include a minimum time either in the question or in the instructions for

the item on "activity status". As will be seen in the next section, it is con-

sidered more profitable to investigate the time actually worked. But when it

is decided to put this in, there should be no differences between the time

established for "female unpaid family workers" and for the rest of the "econo-

mically active women" at the risk of having to formulate very complicated in-

structions not proper for an operation such as the census or to make inco-

herent or absurd recommendations, as has happened in the past.

Elaboration of Measures that partially Overcome the Validity Problem of
Censal Information on Activity Status

The conclusion that Latin American population censuses do not appear to

have been up to now valid tools for the measurement of female economic partici-

pation might not surprise those who repeatedly have written or read about the

censal underestimation of female activity. It is nevertheless new, and in more

than one sense. Up to now emphasis had been placed on the low validity of the

census as a data source on women's economic activity and, in particular, on the

underestimation of those employed in agricultural activities and as "unpaid

- 38 -

family workers". The novelty consists in having based this conclusion in an

attempt to quantify, even though for only a few cases, the magnitude of that

underestimation and in having detected that the problem is more general.

The underestimation seems to affect not only the female agricultural

workers and the unpaid female family workers but also women economically active

in other occupations. Thus, for the users of censal data doing historical re-

search on female labor force these conclusions are discouraging. Not only dis-

couraging but difficult to accept. In fact, how can one renounce the use of

the data source for the study of past trends of female participation? Is there

no part of the information whose validiy might be accepted?

As was seen when contrasting independent sources of measurement the

operational aspects of the labor force approach that in such a remarkable way

affected the female population did not seem to have similar problems when re-

ferring to males. This seems to be due to perceptual distortions of a cultural

nature that make both the interviewers as well as the respondents perceive

women, as do women themselves, as economically inactive when they are actually

participating in an economic activity. The same does not hold true for men.

But it is much more probable that this erroneous perception which therefore

leads to a poor declaration occurs with a greater frequency in the cases

where the economic activity is performed part time or within the home environ-

ment. In other words, when it takes place outside the organized capitalist

sector of the economy or of a governmental activity. Following a similar line

of thought and with the object of overcoming the low validity of censal

measurements, Boserup (1975) proposed to use a measure of female participation

in activities that she calls "modern". Her premise, accepted here, is that

modern activities performed full-time for renumeration, will always be better

- 39 -

registered than other economic activities. What she had in mind when proposing

this new measure of female activity was the comparability between countries at

a given moment, and in time for the same country. These suggestions are also

considered essential here: to attain a more valid measurement of female labor

force participation which would permit the study of trends and, on the other

hand, render more reliable comparisons of economic participation between Latin

American countries, and of the region, with respect to other countries and/or

regions of the world.

Several measures of female economic activity for Latin American coun-

tries that conducted population censuses in the decade of the 1970's are pre-

sented in this section. But the information for each of the countries of the

region is not always available with the necessary degree of disaggregation to

allow to calculate the rate considered as giving the most valid measurement and

at the same time the most comparable between countries that is, the one

taking into account modern activities, as mentioned above. The one that is

possible to calculate for practically all the countries of the region is a rate

that underenumerates active women. This is the refined rate of total partici-

pation, defined as the quotient between the active women 15 years and above and

the total number of women of corresponding ages.6 This rate leads to a

picture of Latin American and Caribbean female activity in which Haiti stands

out from the rest of the countries of the region because of its very high level

of participation (Table 7). In fact, Haiti's refined rate of total participa-

tion is practically twice that of the country following it in order of impor-

6 Use of a lower age limit relatively high is preferred not only because
in this way comparability is gained among countries, but also because, with
regard to the denominator of the rate by excluding women younger than 15 -
differences in age-structure among countries are largely eliminated.

- 40 -

TABLE 7

Some Measures of Female Participation for Latin American and Caribbean
Countries from Available Census Data of the 1970's

Refined participation rate (15 years and over)
per hundred population

Country Total Modern Wage Domestic Non- Gross years of
occupa- earners servants agricul- active life,
tions turald/ urban zones
(15-59 years)

Argentinaa 27.3 13.0 21.0 6.3 26.2 15.1
Barbados 34.1 *
Boliviaa 23.9 5.5 9.4 4.2 19.1 15.3
Brazil 21.1 14.9 17.1 *
Chile b 21.9 9.0 16.2 6.3 21.1 12.9
Colombia 23.9 7.1 15.1 0.2 19.6 11.5
Costa Rica 20.7 10.7 19.0 6.1 19.9 15.1
Cuba 18.3 18.0 11.8
Dominican Republic 26.7 3.8 10.4 3.2 19.2 13.0
Ecuador 17.1 5.6 10.2 4.1 15.1 13.1
El Salvador 25.7 c 14.4 c 22.5 c
Guatemala 14.2 4.2 9.5 4.9 13.8 13.8
Haiti 70.3 *
Hondurasa 16.8 5.8 10.6 4.3 15.6 15.3
Jamaica 36.1 *
Mexico 17.9 6.5 12.0 0.5 18.6 *
Nicaraguaa 21.6 5.7 14.2 7.7 19.9 16.2
Panama b 30.3 13.7 22.4 9.0 29.2 21.6
Paraguay 22.4 6.0 11.6 6.0 20.1 17.5
Peru 19.9 6.1 10.8 3.7 16.4 12.7
Uruguay 29.4 20.7 29.1 17.8
Venezuela 22.6 10.7 17.0 4.9 19.9 12.6

Information not available.

a. Economically active female population 10 years and over.

b. Economically active female population 12 years and over.

c. OMUECE available data absolutely not comparable to those from national publication.

d. Includes unknown activities.

e. Economically active female population 7 years and over.
\
SOURCES: National population censuses (see details at end of present chapter); Data Bank
of CELADE (OMUECE 70). The total rates calculated by one and another source
coincide reasonably well, save for El Salvador where the difference was so
great that it was decided not to incorporate OMUECE data and, in smaller measure,
for Colombia.

- 41 -

tance Jamaica whose rate is, in turn, two and a half times that of

Guatemala, a country that according to this measurement would have the lowest

female participation in the region.

What has been shown up to here does not differ significantly with works

such as PREALC (1978) for the same period or those of Elizaga (1974) and Durand

(1975) using data from the previous decade. But these data do not give a reli-

able picture of female economic participation in the region. Each of the

figures shown must have very different degrees of validity because they derive

from data collection approaches whose degree of validity also differs. Using

the measure proposed by Boserup (1975) and up to the point where the available

data permits, an attempt will be made in what follows to improve the existing

measurement.

This new measure will be called "refined rate of participation in

modern occupations".71t is defined as the quotient between women 15 years

and over who worked in modern occupations and the total number of women of cor-

responding ages. That is, the denominator for this measure is the same as that

used in the prior measure. The numerator includes the following occupations:

professionals, technical and related female workers in all occupational cate-

gories; managers, directors and top level administrators in all occupational

categories; salaried office workers and related; wage earning saleswomen; non-

7 It should be noted that this is a rate and not the proportion of women
active in modern occupations. In this way the underenumeration of the rest
of the occupations does not affect this measure. In those cases where
female labor force information by occupations was not available for active
women 15 years and over, the available figure was used in the numerator (for
example, women 10 years and over in modern occupations), but the age limit
of the denominator was retained at 15 years. In this way, the comparability
is much better assured since, although there are few women under 15 in
modern occupations, a change in the denominator would significantly affect
the magnitude of the rate.

- 42 -

agricultural wage earning laborers and workers and related female wage earners.

In other words, for the computation it was necessary to have tabulations that

classify female labor according to occupation and employment status. (The

occupational classification used corresponds to COTA 1970 and the data derives

from the CELADE OMUECE 1970 data bank.) It must be pointed out that the

problem of correctly measuring the totality of the female labor force is not

solved in this way. On the contrary, information is lost especially with

regard to all occupations that are not of the modern type, but gains are made

in validity and comparability with respect to a sector of a female labor force,

as mentioned before.

The refined rates of participation in modern occupations for each coun-

try are also included in Table 7. The picture these new rates present is com-

pletely different to the earlier one not only because they are all lower as

was expected, since fewer occupations are included but, basically, because if

one ranks the countries according to the magnitude of female participation

measured by one or other rate, these will be situated, save a few exceptions,

in completely different locations. What has been said becomes evident in the

number of crossings that the left-hand side of Graph 1 presents. Here coun-

tries are ranked from larger to smaller female participation in economic activ-

ity according to different rates. Graph 2 completes the differential picture

offered by one and the other measure if only the bars corresponding to the

total rate and to that of modern occupations are observed since the differ-

ences in magnitude between both are more readily discernable in this graph.

According to the rate of participation in modern activities, four Latin

American countries stand out: Panama, Argentina, Costa Rica and Venezuela,

with rates ranging between 10.7 and 13.7 percent. Next comes Chile with 9.0

- 43 -

GRAPH 1

Latin America: Rank Order According to Some Measures of
Female Participation- Countries with Available Necessary
Censal Data for the 1970's

Total Modern Occupations Wage earners

1 Panama 1 Panama 1- Panama
2 Argentina 2 Argentina 2 Argentina
3 Dominican Rep. 3,5 Costa Rica 3 Costa Rica
4,5 Bolivi 3,5 Venezuela 4 Venezuela
4,5 Colombia 5 Chile 5 Chile
6 Venezuela- 6 Colombia 6 Colombia
7 Paraguay Mexico 7 Nicaragua
8 Chile Per Mexico
9 Nicaragua/ 9 Paraguay 9 Paraguay
10 Costa Rica 1' 0 Honduras 0 Peru
11 Per l 11 Nicaragu 11 Honduras
12 Mexico 12 Ecuador 2 Dominican Rep.
13 Ecuador 13 Bolivia. 13 Ecuador
14 Honduras/ 14 Guatemala 14 Guatemala
15 Guatemal 15 Dominican Rep'. 15 Bolivia

SOURCE: Table 7.

- 44 -

GRAPH 2

Latin America: Refined Participation Rates for Women in Modern Occupations,
Wage-earners and Totals- Countries with Available Necessary Data
for the 1970's (per hundred women 15 years and over)

Scale COUNTRIES Scale
3 20o 10 o 0 10 to 30

Panoma

Argentinoa

Uruguay
Costa Rica

Cuba

Venezuela

Chile / ,

CV ombia

Brazil
Ozjl

El Solvador

Md xica

SPeru

Honduras

Nooroguo

Eacuador

SBolivia

SGuatemalo

Dominican Rep.

MModern occupations -] Wage-earners EM Total

SOURCE: Table 7.

- 45 -

percent of its women 15 years and over in modern occupations, and finally the

rest of the countries with 7.0 percent or less.

How do these countries rank with regard to those from other regions of

the world? Some available data are found in the already mentioned article by

Boserup, based on information for 34 countries around 1960. The group of those

countries with greatest female participation in modern ococcupations (the

United States, New Zealand, and Canada) presented an average rate of 21.7 per-

cent, followed by those (Japan, Denmark, France, Belgium, and Norway) with 18.0

percent, and the group of countries (Puerto Rico, Cuba, Chile, Jamaica, and

Mauritius) in the next rank order, 8.6 percent. Thus, the Latin American Coun-

tries with greatest female participation in modern occupation around 1970 would

fall within the countries that around 1960 were grouped in Boserup's second and

third rank orders, closer to the latter. At the other end, none of the Latin

American countries presents in 1970 a rate as low as Boserup's lowest average

rate for 1960 (1.0 percent), not even that of the next group of countries (2.4

percent)8. The lowest of the 1970 Latin American rates corresponds to the

Dominican Republic where it reaches a value of 3.8 percent.

In this presentation of the refined rates of participation in modern

occupations, 7 of the 22 countries for which the refined rate of total partici-

pation had been calculated remained outside the comparison due to the lack of

tabulations that could permit their computation. As the reader might suspect,

if the comparison were to be attempted for earlier dates, the "loss" of cases

would probably be greater. The computation of other rates was then attempted,

8 These two groups are integrated by Sierra Leone, Pakistan, Korea,
Morocco, and Liberia, in the first; and Nicaragua, Egypt, Syria, the
Dominican Republic, and Ghana, in the second.

- 46 -

testing whether the picture that these others offered would be good substitutes

of the supposedly better rate: that of modern occupations. The following

measures were calculated: refined rate of participation of female wage

earners, refined rate of participation in nonagricultural activities, and gross

years of active life in urban areas.9 With each series of rates the ranking

of the countries was again established only for the 15 which the refined rate

of participation in modern occupations had been calculated always from larger

to smaller, building up Graph 3, similar to Graph 1.

When observing the two graphs one can conclude that the only rates that

produce a similar ranking to that of modern occupations are those of wage

earners. In fact, very few countries significantly change their ranking:

Nicaragua, and in second place, the Dominican Republic. This can also be ob-

served in Graph 2: here can be seen that, generally speaking, and with the

noted exceptions, the rates for wage earners reproduce fairly well at a

higher level the ranking for the rates of those employed in modern occupa-

tions. But in his graph the group of countries with relatively high rates is

not so clearly differentiated from those with relatively moderate rates. The

dividing line is not so easily drawn. One must take into account that in the

category of wage earners, in addition to the modern occupations already men-

tioned, are included some labeled as very traditional, such as domestic ser-

vices. And, as can be seen in Table 7 itself, some of the countries with the

9 The latter computation was done instead of the refined rate of participa-
tion in urban areas because the computations were begun with the urban par-
ticipation rates by age. The ranking of countries would not have changed
much if the refined rate had been used.

- 47 -

GRAPH 3

Latin America: Rank Order According to Some Measures of
Female Participation Countries with Available Necessary
Censal Data for the 1970's

Non-agricultural Modern Occupations Urban zone
occupations activities

1 Panama 1 Panama 1 Panama
2 Argentina 2 Argentina Paraguay
3 Chile 3,5 Costa Ric Nicaragua
4 Paraguay, > 3,5 Venezuel 4,5 Bolivia
6 Costa Rica 5 Chile- ,5 Honduras
6 Nicaragua 6 Colombi ,5 Argentina
6 Venezuela 7 Mexico
8 Colombia 8 Peru ,5 Costa Rica
9 Dominican Re 9 Paraguay / Guatemala
10 Bolivia 10 Honduras 9 Ecuador
11 Mexico 1 Nicaragua 0 Dominican Rep.
12 Peru 12 Ecuado 11 Chile
13 Honduras 13 Bolivi 2 Peru
14 Ecuador 14 Guatemal 13 Venezuela
15 Guatemala 15 Dominican Rep. 14 Colombia

SOURCE: Table 7.

-48-

highest rates of participation in modern occupations as well as some with the

lowest have, coincidentally, very high rates of participation in domestic ser-

vice.10

The rates for wage earners permit to compare a larger number of countries

since the information needed for this calculation exists for 19 of the 22

cases. The rank that each of these "added" countries would occupy can be ob-

served in Graph 2: Uruguay, occupying a place very close to that of Argentina

and Panama; Cuba, although not as high it must be taken into account that

this country does not include remunerated domestic service among the economic-

ally active ranking among the Latin American countries with high female par-

ticipation; Brazil, in an intermediate place, although in all probablility it

is badly placed due to the very poor formulation of the question on activity

status; and finally El Salvador, close to Nicaragua.

The other measures shown in Table 7 refined participation rates in

non-agricultural occupations and gross years of active life in urban areas-

give, on the contrary, a totally different picture than the rates of participa-

tion in modern occupations. It is therefore, not possible to take these

measures as substitutes, not even as an approximation to the first. The idea

implicit upon calculating these two last measures was that non-agricultural, or

urban, could be considered almost synonymous to modern. But it has been amply

proven that this is not so. In fact, many of the occupations that are carried

out in the urban area in similarity to the group formed by all non-agricul-

tural occupations include traditional activities such as domestic service,

10 One is struck by the very low rates of participation in domestic service
in Mexico and in Colombia. They seem exceedingly low and unacceptable as
valid measures. But it is beyond the scope of this work to investigate
their real magnitude. One suspects problems of codification.

- 49 -

plus a wide gamut of tasks that may even be characterized as occasional, as are

probably many of those carried out by women who participate in the labor force

as unpaid family workers or as self-employed. As to domestic service it is in-

teresting to point out a collateral finding revealed by examining Table 7. If

to the column of refined participation rates of female wage earners one sub-

stracts that of domestic service, with few exceptions a column of rates very

similar to that of modern occupations is obtained. Aside from a few excep-

tions, the discrepancy between both rates does not go beyond 21 percent and, in

the majority of cases, is between 5 and 10 percent. This not only gives an

idea of the importance of this activity among women, but also suggests that it

constitutes perhaps one of the traditional female activities most often de-

clared with relative accuracy in the population censuses. The latter is a con-

jecture that needs to be contrasted with other independent measurements.

To conclude, and on the basis of the evidence already examined, we assert

that censal data may be used although with great caution since, as was seen,

they tend in general to underestimate active women but especially the agricul-

tural female workers, the unpaid family workers and in general, all those not

engaged in modern activities. Quite the contrary, those best detected by the

population censuses would be those carrying out tasks which, following Boserup,

are called "modern". Since not all countries have the necessary information to

calculate these rates which also could be computed with a much greater degree

of specificity: according to age-groups, civil status, etc. a fairly accept-

able substitute could be on the computation of the participation rates of wage

earners.

Finally, it is believed that a realistic image has been offered of the

relative place that each Latin American country occupies as far as female par-

- 50 -

ticipation in modern economic activity is concerned. What remains to be estab-

lished is the ranking of female participation in all economic activities. This

question of disputable formulation cannot be answered starting from an in-

vestigation that uses censal data.

Conclusions and Suggestions

The two objectives of the present work have been: 1) to evaluate the

quality and identify the deficiencies that statistics for the analysis of

women's economic participation suffer from, especially those of the 1970 census

round taken in Latin America and the Caribbean; and 2) to record the gaps in

the gathering of information on those variables relevant to the analysis and

explanation of women's labor performance. Both objectives have been met and

the conclusions derived emerged clearly: female labor measurements have little

validity and relevant variables on which information is gathered are very

scarce. But, as will be evident to the reader who has followed this exposi-

tion, the conceptual and operational deficiencies of population censuses for

measuring female labor are so great, as the quantified analysis of a few cases

indicated, and underestimate this labor so much that it would be entirely out

of place to plead for more information. Before that, the objective should be

better information.

Therefore, some suggestions to overcome the flaws found in the censal

statistics have been elaborated. This was done without departing from the uni-

verse of existing approaches and definitions institutionalized by the usual

practices recommended internationally. That is, criticisms and suggestions

were made from the perspective of approaches currently in use in censal statis-

tics, and not from other conceptual frameworks. The tasks implied evaluating

- 51 -

the quality and quantity of statistical information were gathered around 1970.

This was done starting with a review of the conceptual frameworks proposed for

the study of female labor force; followed by the analysis of the operational

forms used, the biases detected in censal measurements, and the tabulations

that were elaborated and published. Now the moment has come to put the con-

clusions arrived at together so they can stand out and be evaluated. Those

judged of greatest importance are presented below.

Firstly, among the persons responsible for the 1970 Latin American cen-

suses especially those charged with the wording of questionnaires and in-

structions to interviewers prejudices and stereotypes were found concerning

the role of women in society and, therefore, of the activities that they carry

out. These prejudices and stereotypes, when present in the operationalization

of the concept of an economically active population diminish the validity of

the censal measurement of female labor force.

Secondly, in many Latin American censuses serious inconsistencies were

identified between different instructions or between the wording of items and

instructions, if not a flagrant flaunting of the rules that apply to all social

research. Among these inconsistencies, one frequently found and whose origin

might be probably due to a mechanical adherence to international recommenda-

tions (inconsistent in themselves) relates to the minimum time of work required

from a person to be considered "economically active" and the one required for

the status of "unpaid family worker". These inconsistencies, although affect-

ing both men and women, affect the latter more because it is they who work most

frequently part-time and as unpaid family workers.

Lastly, the attempt to quantify the underestimation of female labor

force, though based on a few cases, brought forth very conclusive results: the

- 52 -

population censuses taken in Latin America in the decade of the 1970's do not

provide in comparison to alternative sources of information a valid measure

of female labor force. This, which is especially true of the sector that in-

cludes women dedicated to agricultural activities and occupied as "unpaid

family workers," is also true of that made up by women employed, in general, in

non-modern activities. Among these, remunerated domestic service was found to

be one of the occupations that should be investigated with special care due to

its rather high relative frequency in some countries of the region and its sur-

prisingly low one in a few others.

As already mentioned, the problem of low validity of censal measure-

ments is so serious that, in spite of having identified important gaps in the

spectrum of variables on which information is gathered, it was not recommended

to broaden information gathering with the exception of one item which would

permit better identification of the employment situation: the period of time

worked. In the case of household surveys, where the problem of validity does

not seem to be as serious, it was indeed suggested to collect additional infor-

mation.

It is necessary to point out, however, that not all the defects derive

from the way in which Latin American censuses rendered operational the concept

of an economically active population. Some of the problems noted seem to go

way back and have been dragged along. This seems to be due not only to those

in charge of the censuses, but also to social science researchers. (As an ex-

ample, just as it is frequent to find evaluations of the quality of Latin

American censuses with regard to the coverage of the total population, by age

and by sex, there are hardly any examples of attempts to evaluate the quality

of the information on the economically active population.) The concept of

- 53 -

"work" itself continues to be insufficiently clear and in spite of recent

developments originating within the framework of marxist economics and the time

budget approach, obsolete patterns persist, such as continuing to consider only

some activities and not others as economic: for example, the non-remunerated

domestic activities.

The result of the evaluations carried out here should not lead one to

conclude that the census is necessarily a bad tool for measuring the female

labor force. On the contrary, we believe that it can and must be improved. A

first step towards an awareness of the issue was already taken when social

science researchers and some international agencies (see United Nations: 1978,

and ECLA: 1978) recognized that the task of detecting female economic activity

gives rise to specific problems. Earlier in this chapter, some suggestions

were already made to improve the validity of the collection of information on

the employment situation. Obviously, these are not definitive, as additional

research is needed to formulate more precise recommendations, as for example,

concerning the reference period on the time worked, and on the wording of the

sequence of questions for measuring the "activity status."

Up to here, the recommendations formulated have not paid attention to

the fact that older available information contains certain basic flaws that

might invalidate further analyses if used indiscriminately. To overcome this

obstacle Boserup proposed in 1975 the computation of refined activity rates for

modern occupations in preference to the generally used measures since these

activities tend to be registered more completely by censuses. We have followed

this suggestion here and have calculated such rates for all the countries of

the region that had the necessary data. It must be pointed out that this mea-

sure, which only takes into account one sector of the economically active

- 54 -

female population, is not proposed here as a replacement for a total rate since

it is not intended to give information for the entire female labor force of

each country. If its use is suggested, it is because the censal data referring

to the totality of economic occupations are not valid. The proposed measure

does indeed offer information that is valid even though for only a part of the

economic occupations, and above all, useful for the purposes of comparative

studies both in space (among units) as in time (covering the history of each

unit). Given the greater validity of this rate relative to the total rate, the

comparisons based on it must also be more valid.

Because the computation of this rate requires the existence of special

tabulations and because not all countries have them, we have computed a substi-

tute measure that requires a type of information that is available among a

greater number of countries and which is sufficiently acceptable. This is the

refined participation rate for wage earning women. But here again, this matter

has raised some questions requiring further investigation. Are Boserup's re-

commended occupations or activities the most adequate, or would it be advisable

to redefine what is meant by "modern occupations"? Should some of the tradi-

tional occupations perhaps be included, such as "domestic service," that ap-

parently is more accurately registered as an economic activity than other

traditional occupations? Or, is it that, perhaps, the dividing line between

valid and non-valid registration has to do with the social recognition of the

economic nature of the task rather than with its modernity?

As already stated, the present study has dealt with only a selection

among the many topics that merit consideration. Many others were also identi-

fied which undoubtedly will give rise to future research. Among these, the

problem of how much censal underestimation of the female labor force is due to

- 55 -

the deficient conceptualization and operationalization of the variable and how

much to characteristics proper to the data collection operation. It would be

advisable to try to evaluate how the information obtained through censuses and

surveys of various population groups among which interviewers are recruited

(students, teachers, etc.) is influenced by whether these are or are not

remunerated, the type of training they receive, etc. It is necessary also to

investigate how the seasonality of certain agricultural tasks affects the

measurement of female labor by censuses and surveys, as well as how the defini-

tion of head of household and the selection of the respondent influences the

declaration of female economic activities. These and many other topics require

further investigation before we are able to formulate concrete recommendations

directed at improving the data on the status of women in the economic sphere.

- 56 -

REFERENCES

BANCROFT, Gertrude (1958), The American labor force: Its growth and changing

composition, New York: John Wiley & Sons.

BOSERUP, Ester (1975), "Employment of women in developing countries," in TABAH,

Leon (ed.), Population growth and economic development in the Third

World, Dolhain, Belgium: Ordina Editions, Vol. 1, chapter III.

D'SOUZA, Stanislas (1978), Sex-based stereotypes; sex biases and national data

systems, paper presented at the Seminar: "A Mulher na Forca de

Trabalho na America Latina", Rio de Janeiro, Brazil, 23-26 November.

DURAND, John D. (1975), The labor force in economic development, Princeton,

N.J.: Princeton University Press.

ELIZAGA, Juan C. (1974), "Participacion de la mujer en la mano de obra en

America Latina: la fecundidad y otros determinantes," Revista

International del Trabajo, Vol. 89, Nos. 5-6, May-June, pp. 569-588.

OSTRY, Sylvia (1968), The female worker in Canada, Ottawa, Canada: Dominion

Bureau of Statistics.

PREALC (1978), Participacidn femenina en la actividad economic en America

Latina analysiss estadfstico), Working Document PREALC/161.

PECHT, Waldomiro (1974), La poblaci6n economicamente active en las actividades

agrfcolas en Brasil y Mexico: un analysis a partir de los censos

economics y de poblaci6n, Santiago, Chile: CELADE, IPI/6, Rev. 1.

POLO NAJERA, Jose (1978) Caracterfsticas de la poblaci6n economicamente active

de Bolivia, INE-CELADE, Seminario sobre Situacion Demografica de

Bolivia, La Paz (17-20 April).

- 57 -

RECCHINI DE LATTES, Zulma L. (1979), "Las mujeres en la actividad economic en

Argentina, Bolivia y Paraguay", Demografia y Economfa, Vol. XIII, No. 1

(37), pp. 19-48.

TORRADO, Susana (1979), "Conclusiones y recomendaciones relatives a las fuentes

de informaci6n necesarias para estudiar la participation de la mujer en

los mercados de trabajo en America Latina", Relatorio Geral do

Seminario "A mulher na Forca do Trabalho na America Latina", Rio de

Janeiro, Brazil: IUPERJ.

TORREZ P., Hugo (1977), La mujer boliviana y sus caracterfsticas demograficas

en la fuerza de trabajo 1975, La Paz: Ediciones CIS, Serie Estudios de

Poblacion y Desarrollo, No. 12.

UNITED NATIONS (1978), Draft Principles and Recommendations for Population and

Housing Censuses, Part Two, Topics and Tabulations for Population

Censuses (E/CN.3/515/Add. 2).

- 58 -

STATISTICAL S URCES

ARGENTINA, Instituto Nacional de Estadistica y Censos (INDEC) (n.d.), Censo

Nacional de Poblacion, Familias y Viviendas 1970. Resultados

obtenidos por muestra. Total del pals, Buenos Aires:: INDEC.

BARBADOS, Barbados Statistical Service (1972), Commonwealth Caribbean

Population Census 1970. Barbados. Preliminary Bulletin. Working

Population, Part I. Garrison, St. Michael: Barbados Statistical

Service.

BOLIVIA, Instituto Nacional de Estadistica (1977), Resultados Anticipados por

Muestreo. Censo Nacional de Poblacion y Vivienda 1976. La Paz:

Institute Nacional de Estadfstica.

INSTITUTE NATIONAL DE ESTADISTICA (INE) (n.d.), Resultados del Censo

Nacional de Poblacion y Vivienda 1976, La Paz: Instituto Nacional de

Estadistica (Vol. 2).

BRAZIL, Fundacao Instituto Brasileiro de Geografia e Estadistica (IBGE) (1973),

Censo Demografico, Brasil VIII Recenseamento Geral 1970. Serie

Nacional, Vol. 1, Rio de Janeiro: Fundacao IBGE.

FUNDACAO IBGE (n.d.), Sao Paulo, pesquisa national por amostra de

domicilios. Populacao; habitacao; mao de obral salario; instrucao.

Doc. GEPD No. 48. 1st Quarter of 1970, IBGE, Lucas, GB.

CHILE, Instituto Nacional de Estadfsticas (n.d.), Poblacion, Resultados

definitivos del XIV Censo de Poblacion, 1970, Total pals, Santiago.

COLOMBIA, Departamento Administrativo Nacional de Estadistica (DANE) (1975),

XIV Censo Nacional de Poblaci6n y III de Vivenda, Muestra de Avance.

Poblacion, Bogota, D.E.: DANE.

59 -

COSTA RICA, CAJA COSTARRICENSE DE SEGURO SOCIAL (1974) a), Estad(stica

patrons, trabajadores, salaries y cuotas 1973, Departamento Actuarial

y Estadistico, Seccion de Estadfstica.

DIRECTION GENERAL DE ESTADISTICA Y CENSOS (1974) b), Censos Nacionales

de 1973. Poblacion (2 volumes); San Jos: Direccion General de

Estadistica y Censos.

CUBA, Junta Central de Planificacion (1975), Censo de Poblacion y Viviendas

1970, Havana: Editorial Orbe.

DOMINICAN REPUBLIC, Oficina Nacional de Estadistica (1972), Comentarios sobre

los resultados definitivos del V Censo Nacional de Poblacion Segunda

parte, Santo Domingo, mimeo.

ECUADOR, Instituto Nacional de Estadfstica y Censos (1977), III Censo de

Poblacion 1974. Resultados definitivos. Resumen Nacional, Quito:

INE.

GUATEMALA, Direccion General de Estadfstica (1975), VIII Censo de Poblacion:

26 de marzo de 1973. Poblacion total. Poblacion indigena. Cifras

definitivas. Serie III, Vol. I. Republica, Guatemala.

HAITI, Institut Haitien de Statistique (1978), Recensement General de la

Population et du Logement. Aout 1971. Vol. IV. Resultats pour le

Department Traditionnel de L'Artibonite, Port-au-Prince.

HONDURAS, Direcci6n General de Estadistica y Censos (1976), Censo Nacional de

Poblacion. Marzo 74, Resumen por Departamento y Municipio. Vol. 1,

Tegucigalpa, D.C.: Direccion General de Estadistica y Censos.

JAMAICA, Division of Censuses and Surveys. Department of Statistics (1974)

Commonwealth Caribbean Population Census 1970, Jamaica. Population

- 60 -

Census 1970. Bulletin 5. Economic Activity, Kingston: Department of

Statistics.

MEXICO, Direccion General de Estadistica (1972), IX Censo General de Poblacion

1970. 28 de enero de 1970. Resumen general abreviado, Mexico, D.F.:

Talleres Graficos de la Nacion.

NICARAGUA, Ministerio de Economia, Industria y Comercio (1974), Censos

Nacionales 1971, 20 de abril de 1971; Poblacion, Vol. III.

Caracterlsticas Econ6micas, Managua.

PANAMA, DIRECTION DE ESTADISTICA Y CENSO (1972), Estadisticas del trabajo

(Encuestas de Mano de Obra) Aro 1970, Panama: Direccidn de Esta-

dfstica y Censo.

DIRECCION DE ESTADISTICA Y CENSO (1976), Censo Nacionales de 1970, Vol.

III and IV, Panama: Direccion de Estadfstica y Censo.

PARAGUAY, Direccion General de Estadfstica y Censos (1975), Censo Nacional de

Poblacion y Vivienda. 4 de junio de 1972. Resultados Definitivos.

Nivel Nacional. Vol. II, Lima: Oficina Nacional de Estad'stica y

Censos.

URAGUAY, Direccion General de Estadfstica y Censos (1977), V. Censo de

Poblacion. III de Viviendas ano 1975. Muestra de anticipation.

Cifras y Comentarios, Montevideo: Direccion General de Estadistica y

Censos.

VENEZUELA, Direcci6n General de Estadistica y Censos Nacionales (1975). Censo

71, X Censo de Poblacion y Vivienda. Venezuela. Resumen Nacional:

Fuerza de Trabajo. vol. VI, Caracas: Direcci6n General de Estadistica

y Censos Nacionales.

CELADE, "Banco de datos: Operacion Muestra de censos 1970 (OMUECE 70)".

- 61 -

ANNEX: SUGGESTIONS AND INITIATION OF THE NEW STAGE

For many of the Latin American and Caribbean countries the census is

the only source of information of national coverage. Because of this and of

its high cost, it cannot be allowed to offer, even if in one or a few items,

results that are of little validity, reliability, and comparability. It is

therefore imperative to improve the statistics of the next round of population

censuses. The suggestions summarized in this annex and the ideas outlined for

future research are precisely guided by this aim.

a) Recommendations

The first stage of the research project -- completed between April 1979

and October 1980, -- consisting of the assessment of the quality of the censal

measurement of female labor force, allowed one to identify a number of actual

and potential sources of invalidity. In fact, the conceptual analysis of all

items and instructions to enumerators on "activity status" contained either in

the censal schedule or in the handbook for enumerators, lead to formulate a set

of recommendations and a set of suggestions. The latter -- suggestions -- dif-

fer from the former -- recommendations -- in that they require to be tested

empirically; the adequacy of the former requires no proof.

The recommendations apply to the items, their response categories, the

time required from a person to be considered economically active, and to the

training of the census takers. They point out the need to: i) eliminate sex

biases from instructions to enumerators; ii) sensitize enumerators to the case

of double "work", so widespread among women (as well as youths and elders of

either sex); iii) formulate the item on "activity status" as a question rather

- 62 -

than as a title looking for the constancy of the stimulus; iv) present the re-

sponse alternatives starting with those of activity and following with those of

inactivity; v) instruct the census takers to stop reading the response alterna-

tives when the first answer chosen by the interviewee is obtained; vi) see that

the minimum amount of time required from a person to be considered economically

active is not inconsistent with that required from a person to be classified

within any specific category of economically active.

The suggestions, i.e., those related to the potential sources of in-

validity of the measurement of female labor force stemming from different

sources include: i) using different wording of item(s) on activity status; ii)

employing interviewers with different levels of training; iii) varying the re-

spondent's and the interviewer's sex; iv) collecting data at times which do

and do not take into account the seasonality of certain agricultural and agri-

cultural activities.

b) Report on the trip1

Between September 1981 and January 1982 contacts were made and official

invitations were received by Catalina H. Wainerman from the Statistical Offices

of various Latin American and Caribbean countries. Barbados, Ecuador, Panama,

Paraguay, and Peru were chosen on the basis of the reliability of their human

resources and of their manifest interest in the project. The visits (between

early March and mid April 1982) had a two-fold aim: 1) to acquaint the people

in charge of producing the official statistics with the limitations of the

1 This trip, which was supported by USAID, was made possible by the initia-
tive and enthusiasm of Judith Bruce from the Population Council.

63 -

censal measurement of female labor force revealed by the analysis completed

during the first stage of the research project,and whose outcome is contained

in the book El trabajo femenino en al banquillo de los acusados. La medicion

censal en America Latina; and 2) to assess the interest of the producers of

statistics to participate in the design and testing of alternaitve items

addressed to improve the quality of current census measurement.

The presentation of the results of the first stage of the research was

done through seminars involving up to 25 persons involved in the production of

data on the economic characteristics of the population through censuses, con-

tinuous household surveys, and labor statistics. Demographers, statisticians,

sociologists, and economists from the Statistical Offices and, in some cases,

from the Ministry of Labor and the Ministry of Planning, composed the audience.

The seminars had a didactic aim. They were started with a presentation of the

results of the research on the censuses collected in the region around 1970,

and followed by a practical exercise which consisted on working together in the

search for local examples of problems involved in different measurements of the

female labor force. The interest demonstrated by the people and the frequency

with which personal experiences on this topic appeared are a proof of the posi-

tive effect the seminars had as a tool for sensitizing people with respect to

this subject matter.

It may be asserted that the visits achieved a great success in terms of

both its aims. In most cases the people responsible for the collection of

statistics were much in favor of participating in a project of this kind. In

all cases there was'clear awareness of the difficulties inherent in the mea-

surement of this phenomenon. In all cases, especially in Barbados, people were

equally clear as regards the impossibility of their institutions to contribute

- 64 -

with funds to such a project. Most were willing, however, to contribute with

work, mainly through labor force within one of the annual data collections of a

household survey.

c) Some ideas for the next stage

The next stage will proceed along two lines involving: i) a conceptual

or arm-chair type of research, and ii) and empirical or field type of research,

in that order.

The former, i.e., the arm-chair activity, aims at enlarging and deepen-

ing the conceptual analysis conducted in the first stage. Its purpose is to

identify additional sources of invalidity of the censal measurement of female

labor force that the review of the Latin American and Caribbean censuses of the

1970's might have not revealed. Such conceptual analysis will imply: 1) to

review the critical literature on measurement by censuses and by household sur-

veys conducted beyond the limits of the Latin American and Caribbean region,

not only in developed countries like the USA, Canada, and the like, but also in

other developing countries which differ widely in cultural terms, like India;

2) to review attempts conducted in Latin American and the Caribbean countries

to measure the female labor force through data collection operations and tech-

niques other than those used by censuses and household surveys (e.g., those

which have used the time-budge approach, the "realistic" list of response al-

ternatives to activity, the investigation of activities which are perceived as

being "work" or "non-work", etc.); 3) to identify additional pairs of measure-

ment of female labor force obtained by censuses and by other kinds of data col-

lection operations to be used as non-intentional test experiments of the effect

of some of the variables identified as possible sources of invalidity of female

- 65 -

labor force measurement during the first stage of the project; 4) to design

field experiments to test the suggestions mentioned in a) above.

The need to review the critical literature produced beyond the limits

of the region obeys to the need to draw on the knowledge produced or that is

being produced in other societies, which differ either in terms of their econo-

mic structure, level of development, or cultural orientations towards women's

economic participation. For instance, in a recent seminar on "Women's Work

and Employment" held at the beginning of 1982 in Delhi (India), a number of

papers bearing directly on the issue of the censal measurement of the female

labor force were discussed.2 Two of the four substantive topics of the

agenda were especially relevant: "Secondary analysis of Indian data on female

work participation", and "Methodological issues in measuring women's work".

Among the achievements of the seminar it is worth mentioning the recognition of

the urgent need to improve the statistics on women's work; the need to bridge

the gap between researchers and national level data collection agencies; and

the need to provide input for the "methodological leap" between small-scale

anthropological and time use studies of women's work and the requirement of

large-scale national data collection activities.

Measurement of female labor force (especially in rural areas and among

women working in highly "invisible" activities in the informal sector, as

unpaid family workers or on own account basis) using the time budget approach

or other kinds of approaches and techniques, though not applicable to censal

operations, might be a source of insights. There already exist a number of

such attempts in the region. For instance, the research conducted by Judith

2 "Technical Seminar on Women's Work and Employment", organized by the
Institute of Social Studies Trust, Delhi (India), 9-11 April 1982.

- 66 -

Fircher Laird on the rural women in Paraguay where lists of "reality-based"

preceded activities (like washing, cleaning, knitting, plowing, harvesting,

candle-making, etc.) were used to measure "work."3 The study directed by

Magdalena Leon de Leal on the rural women in Colombia using a long question-

naire to investigate the division of labor by gender within the household is

another case in point. There techniques of self assessment of the frequency

with which each activity is developed by each member of the unit were used.4

Time-budget techniques were applied by E.A. Cebotarev in her study of Latin

American rural women.5 Elizabeth Jelin's research on female labor from urban

sectors using life histories is another quite insightful source.6

Additional cases of measurement of female labor force by census and by

other type of operations may help to identify new sources of invalidity and/or

supply evidence in favor or against the sources already identified. Certain

censuses might provide also non-intentional experiment. For instance, the 1980

Peruvian population census applied two questionnaires -- a long and a short one

-- to different samples of the population. Both were applied by equally "badly

3 The study mentioned is La mujer rural en el Paraguay. Dimension socio-
econ6mica, Direcci6n General de Estadfstica y Censos, Ministerio de Hacienda
de la Repdblica del Paraguay, 1979. It contains the computation of five
activity rates (including or not unpaid family workers, work outside the
domestic unit, etc.) obtaining a wide range which varies between 15.7 and
65.4 per cent.

4 Magdalena Leon de Leal Compp.) Mujer y capitalism agrario, Bogota:
ACEP, 1980.

5 E.A. Cebotarev, "La organization del tiempo de actividades domesticas y
no-domesticas de mujeres campesinas en Lationoamerica", mimeo, 1978.

6 Elizabeth Jelin and Maria del Carmen Feijoo, Trabajo y familiar en el
ciclo de vida femenino: el caso de los sectors populares de Buenos Aires,
Buenos Aires: CEDES, Estudios CEDES Vol. 3, no 8/9, 1980.

- 67 -

trained" census takers. Both share one item addressed to measure the "activity

status", but whereas this is the sole item in the short questionnaire, it is

one of a set of items on the same variable in the long questionnaire. Once the

tabulations are available, the effect of the activity rates obtained for

females with one or the other questionnaire (and the same training of inter-

viewers) will be assessed.

Field experiments will be designed in an attempt to empirically test

the effect of the sources of invalidity identified. These experiments will be

designed to carried out in pairs, keeping constant certain features while vary-

ing others. For instance, two operations of data collection using the same

conceptual and operational definitions of activity status and differing in the

level and quality of training of the interviewers will be designed. Other

pairs of data collection will be planned so that the wording of the items(s)

and the order of presentation of the preceded alternatives of activity is

varied keeping other features constant.

The empirical stage of this project will consist on the field experi-

ments, which will be conducted in five countries within the region differing in

their economic structure, degree of urbanization, cultural orientations towards

women's participation in the labor market, etc. To keep costs relatively low

it is suggested that the field experiments be conducted as part of a continuous

household surveys which are carried out by the Statistical Offices of the dif-

ferent countries. The positive reactions obtained from the majority of the

Latin American and Caribbean countries visited as part of the second stage of

this project assures the feasibility of this research enterprise.

MILLISECOND	CLASS.METHOD	MESSAGE
0	sobekcm_page_globals.constructor
0	sobekcm_page_globals.constructor	Application State validated or built
0	sobekcm_database.verify_item_lookup_object
0	sobekcm_page_globals.constructor	Navigation Object created from URI query string
0	sobekcm_database.verify_item_lookup_object
0	sobekcm_page_globals.display_item	Retrieving item or group information
0	sobekcm_page_globals.get_entire_collection_hierarchy	Retrieving hierarchy information
0	sobekcm_assistant.get_entire_collection_hierarchy
0	cached_data_manager.retrieve_item_aggregation
0	cached_data_manager.retrieve_item_aggregation	Found item aggregation on local cache
0	item_aggregation_builder.get_item_aggregation	Found 'all' item aggregation in cache
0	system.web.ui.page.page_load (ufdc.page_load)
0	sobekcm_page_globals.constructor.on_page_load
0	html_echo_mainwriter.add_style_references	Adding style references to HTML
0	html_echo_mainwriter.add_text_to_page	Reading the text from the file and echoing back to the output stream
38	html_echo_mainwriter.add_text_to_page	Finished reading and writing the file