Newsletter

August 1999 - Number 4

User Questions

This occasional section of the FLS Newsletter incorporates questions of general interest from data users. It deals mainly with questions that clarify existing variables and relationships between variables, respondents, and subfiles in the public-release databases. Topics covered relate to questions not addressed in the public-release documentation.

IFLS-1

I get more than one household with the same value of CASE, the household identifier. What's going on? I get income values that look like 998998976 and 996996992. Are these missing value codes? I expected to see 998998998 and 996996996 according to the documentation.

The problems above arise for Stata users who converted the SAS transport files into Stata files and forgot to make double-precision format the default for data conversion. Variables such as CASE and income are numeric integer variables with a large number of digits. The values exceed what single-precision (or float) format can accurately store; thus, the double-precision value 998998998 is stored as 998998976 in float format. See the READMEs for the FLS databases for more on this issue.

Stata versions of the FLS databases have been created so that Stata users will not have to worry about conversion problems for large integer values.

Is there an ethnicity variable in the IFLS database?

Ethnicity is not an approved question for inclusion in research questionnaires in Indonesia, even in government surveys, and so is not collected in the IFLS. However, a proxy measure can be constructed using language of interview, residence, and, sometimes, religion to provide clues to the respondent's ethnicity. For the language of interview to be other than Indonesian (Bahasa) provides a good measure of ethnicity. If the language used is Indonesian, it can be combined with residence and religion to guess at ethnicity. For example, if the person using Indonesian is Hindu, he or she is almost certainly Balinese, particularly if living on Bali; if Indonesian is used and the person is Christian and lives in North Sumatra, he or she is almost certainly Batak. Additional reading on the diversity of Indonesia will help. One recommendation is the introductory chapter in The Demographic Dimension in Indonesian Development, by G. J. Jugo, T. H. Hull, V. J. Hull, and G. W. Jones, New York: Oxford University Press, 1987.

What is the direction of dowries reported in K12 in Book 3 of the IFLS survey? There is no indication if this is a dowry received or given by the individual, as dowry information is asked of both males and females. Also, are the premarital assets in K14 those of the respondent's family or of the respondents themselves?

With respect to KW12 (dowries), the bride generally receives the dowry and the groom gives the dowry. This can vary among ethnic groups, but the feeling is that, in the vast majority of cases, the direction is to the bride from the groom. The objective of the question was to identify what money, goods, etc., come into the household at marriage, not necessarily who they came from. Below is some information from an Indonesian colleague that might be of use.

The giving of a peningset, which is not a dowry but a gift, is a common practice for a Javanese family. The gift can be very expensive or simple, depending on the family's socioeconomic status. A dowry, or mas-kawin, is common among the Moslems. (Javanese are also Muslim, so they are more familiar with mas-kawin.) The amount and type of dowry vary by ethnic group. In Southeast Sulawesi, dowries have very high values. Sometimes the dowry is paid as a certain number of cows or buffaloes. Sometimes it is paid in money, with the amount declared in the ritual ceremony in public. More-modern people tend to pay the dowry in recognition of Al-Quran, furnishing a set of praying gear--a sajadah, the mat for praying, and a mukena, the veil.

In which direction is the dowry given? In most of the Indonesian ethnic groups, the dowry goes to the bride or the bride's family, following the patriarchal system. Only one ethnic group, the Minang of West Sumatra, follows the matriarchal system, whereby a girl has to buy her future husband with a dowry. If the dowry is paid in goods, jewelry, and money, it may not go to the new family but may become a part of defraying the cost for the wedding party.

As for premarital assets (KW14), the intent of the question was to get at the respondent's own assets, i.e., those he or she owned personally. The question's wording and the interviewer's instructions specifically refer to what the respondent owns. Thus, a woman should respond about her own assets and not those of her family, and likewise for a male respondent. There is no way of knowing whether a respondent did not understand the meaning of the question and gave a response based on family assets. Let us assume that if this happened, it did not happen often.

In comparing the household nonlabor income in PH09 from BUK2P2 in the IFLS-1 data to the sum of the Book 2 respondent's nonlabor income (PH10) and the nonlabor income of the respondent's spouse (PH11), I find cases where PH09, PH10, and PH11 all have the same value. Aside from those cases, there are others where the sum of PH10 and PH11 is greater than PH09. What does this mean?

Out of 43,000 records, there are 133 instances in which a BUK2Ph2 record has PH09R1=PH10R1=PH11R1. There are cases in the IFLS-1 data where, for jointly owned or received assets or income, respondents reported the value of the asset/income as the full amount for each owner. Thus, if dividend income is viewed as being received jointly, then PH10 may well equal PH11, because no specific interviewer instructions existed to tell interviewers how to deal with such cases.

In just 40 cases, PH10+PH11 is greater than PH09 and PH10 is not equal to PH11. Some of these are most likely miscoded values, and others may be cases where the head gets all the income and then gives some portion to his spouse (or vice versa). The user must decide how to handle these cases, because of the lack of clarifying information.

MFLS-1

I'm trying to replicate the creation of MFLS-1 individual and household income variables described in Kusnic and DaVanzo, Income Inequality and the Definition of Income: The Case of Malaysia (Santa Monica, CA: RAND, R-2416-AID, 1980) but am having trouble getting numbers similar to those in Kusnic and DaVanzo. Can you help?

The MFLS-1 Individual Level Dataset, which is available from the FLS Web site, was the basic data set used for the Kusnic and DaVanzo analysis. The data set has one record per person in the MFLS-1 Household Roster. That record includes detailed characteristics from MF1, MF2, and MF3, wage earnings from MF4 and MF5, and income and wealth from MF6. Various wage and income measures were created and put into the data set. This may be the best place to start in trying to construct earnings and income measures for MFLS-1 households and individuals.

EGSF

In PEMITL02, employment, I find women who have more than one record with the same L06 (type of activity). There are also a handful of cases where the total hours worked per day are greater than when summing across all the respondent's activities. What's going on? Are these errors in the data?

A woman could have more than one employer for whom she does the same type of work. If she chose to treat the work for each employer as a separate activity, there would be more than one record for a given activity. For farming activities, a woman who cultivates more than one crop might report the work for each crop as a separate activity. The variable L08, number of days doing the activity in the last 2 weeks, can provide some assistance here, because it may vary by activity. Likewise, L08 can solve the handful of "works more than 20 hours a day" cases. A woman with two activities may work 12 hours a day at each activity, but may work 3 days a week at one activity and 2 days a week at the other. Thus, she is actually working 12 hours a day, 5 days a week. Don't forget to also look at L08 when looking at L07 (time spent doing activity).