29 July 2010

English

MALAWI NATIONAL SEMINAR ON CENSUS DATA ANALYSIS

Blantyre, Malawi

19-22 July 2010

Report

1

table of conte nts

table of conte nts 2

Objective of the National Seminar 3

Attendance 3

Opening of the Seminar 3

Introduction to the Census Analyses Project 3

Introduction to the Seminar 4

Fertility Analysis 4

fertilty measurement by the P/F Ratio Methods 5

Analyzing Census Data on Age at First Marriage 5

Introduction to CensusInfo 6

General Discussion on Critical Gender Issues in Malawi in the Context of Development 7

Discussion on future plans 7

Closing 9

Annex I: Agenda 10

annex II: list of participants 13

1

Objective of the National Seminar

1.  The Malawi National Seminar on Census Data Analysis was held in Blantyre, Malawi from 19 to 22 July 2010. The main objective of the national seminar is to provide an opportunity for the staff member of the Malawi National Statistical Office in collaboration with the University of Malawi Population Studies Centre, to further collaborate on the analysis of data from the 2008 census of Malawi on selected topics. The national seminar will cover the following topics: (i) fertility and nuptiality from a gender perspective; (ii) fertility indices based on data from the 2008 census; (iii) measures of nuptiality based on data from the 2008 census: and (iv) and use of the CensusInfo software for data presentation and dissemination. See Annex I for Seminar agenda.

Attendance

2.  The Malawi National Seminar on Census Data Analysis was attended by 23 participants (see Annex II) of whom 13 were from the Malawi National Statistical Office, four from the University of Malawi, two from the UNFPA office in Malawi, and four represented UNSD.

Opening of the Seminar

  1. Keiko Osaki Tomita, Chief, Demographic and Social Statistics Branch, delivered an opening remarks on behalf of the UN Statistics Division (UNSD). At the outset, she expressed appreciation for Ms. Mercy Kanyuka, Deputy Commission, and Malawi National Statistics Office (MNSO) for co-organizing the seminar with UNSD. She also thanked the participants for their commitment to the Seminar, despite their busy schedule. Indeed, it is very busy time for the MNSO, as they are currently heavily engaged in census data anlalyses and the enumeration of DHS data at the same time.
  1. Ms. Osaki Tomita described the vital role of the United Nations in advancing the global statistical system. UNSD not only collects, compiles and disseminates statistical information worldwide, but are also responsible for the development of international standards for statistical methods, definitions, classifications of data. Technical cooperation to strengthen the capacity of national statistical systems also constitutes an important element of the Division’s work. UNSD also leads the 2010 World Programme on Population and Housing Censuses in which all but eight countries of the world participate.
  1. She reminded the participants that a population and housing census is usually the largest statistical activity that a country undertakes. The important roles that census data play for planning, policy-making, and administrative and research purposes cannot be overemphasized. But in order for census data be truly useful, the data have to be fully analyzed and widely disseminated. In this connection, she wished that the seminar would provide an excellent opportunity for participants to enhance their knowledge-base for census analyses. She closed the statement wishing a great success in their deliberations.

Introduction to the Census Analyses Project

  1. Ms. Keiko Osaki Tomita explained the objectives of the seminar in the context of the project which UNSD had been implementing. In January 2010, UNSD launched a three-year project entitled “:Strengthening National Capacity to Analyse, Present and Disseminate Census Data for Evidence-based Policy Making”, with a financial support of the Italian Ministry of Foreign Affairs. The project aims to promote higher level of analysis and better dissemination of census data in support of evidence-based decision making.
  1. Organizing national seminar on selected topics where in-depth analyses are needed is one of core activities to be implemented under the project. The first national seminar on census analyses was held in Egypt in May 2010 and the present one in Malawi was the second in a series. Such thematic seminars are expected to take place in nine African countries. Ms. Osaki Tomita acknowledged that two topics – fertility and nuptiality – selected for the seminar by MNSO were of critical importance in Malawi, hence merit deeper analyses. Participants were reminded that they were eventually expected to produce analytical reports, using the knowledge and skills acquired during the seminar.
  1. Ms Osaki Tomita then introduced Dr. Griffith Feeney who would provide lectures on the advanced techniques of estimating fertility and nuptiality during the seminar, and Dr. Jerry Banda who would lecture on how to prepare analytical reports, and lead the discussion on future plans.

Introduction to the Seminar

  1. Dr. Griffith Feeney provided an introduction to the seminar in terms of what was to be covered in the area of measurement of fertility and nuptiality. He explained that the objective of the seminar was to be very practical—to engage in analysis of population census data with the aim of understanding errors and omissions and extracting the best possible information on fertility and nuptiality for the use of participants in finalizing their thematic report on these subjects for the 2008 census. He elaborated that the seminar would of course cover various demographic methods along the way, as they are used to conduct the analysis. He emphasized, however, that the tools only help us with the work. The work that is done using the tools is data analysis, used in the sense of John W. Tukey’s 1997 book Exploratory Data Analysis.

Fertility Analysis

  1. Dr. Griffith Feeney presented on various methods of fertility analysis. The session on reverse survival estimation began with a brief review of basic fertility measures, including the crude birth rate (CBR), age-specific birth rates (ASBRs), the total fertility rate (TFR), and the mean age at childbearing (MAC).
  1. Reverse survival estimation is one of the oldest, but also one of the most underutilized methods for estimating fertility from population census data. The explanation for its under utilization may lie with the United Nations Population Divison’s Manual X: Indirect Techniques for Demographic Estimation, in which reverse survival is presented only for estimation of the crude birth rate for 10 years prior to the census. This session introduced a far more general approach to reverse survival that provides total fertility rates as well as crude birth rates, and for up to 40 years prior to the census date. Total fertility rates are far more useful than crude birth rates, and the longer retrospective period allows comparison with estimates from prior censuses, which is invaluable as a test of data quality and the accuracy of the estimates.
  1. The presentation of the reverse survival method begins with the use of stationary population concepts to solve reverse survival problems. A simple example of a reverse survival problem is: Given the number of persons 0-4 years old at the time of a census, how do we estimate the number of births during the five years prior to the census. The calculation requires life table “big Lx” values, which generally have to be estimated, but the results are robust against departures of assumed from actual life table values.
  1. Once numbers of births during five year periods prior to the census have been estimated, the female age distribution is reverse survived to get the age distribution of reproductive age females at five year intervals prior to the census. Combining information on numbers of births, the age distribution of reproductive age women, and an age pattern of mortality gives estimates of total fertility rates. The age pattern of fertility may come from any number of sources, including census data on births to women during the year prior to the census or data from a recent Demographic and Health Survey. The total fertility estimates are robust against departures of assumed from actual age patterns.
  1. A powerful advantage of the reverse survival method is that it requires as input only the age-sex distribution of the population. This information is available from virtually every population census, and for most countries of the world it is available for a series of past censuses. The example presented in the seminar compared reverse survival estimates of total fertility for India from the censuses of 1951, 1961, 1971, 1981 and 1991. Because estimates are produced for the 40 years prior to each census, each pair of censuses provides a 30 year overlap of estimates. The consistency of estimates from successive censuses provides a reliable diagnostic of the quality of the estimates.
  1. The quality of reverse survival estimates depends first and foremost on the quality of the census age-sex distributions. Problematic age-sex distributions are quickly revealed by examination of reverse survival estimates derived from them, so that reverse survival provides a means of assessing the quality of age-sex distributions as well as a method of estimating the level and trend of total fertility. In some cases, knowledge of the likely trend of fertility, deriving either from estimates produce from other data or general empirical regularities, is used to correct errors in age-sex distributions.

fertilty measurement by the P/F Ratio Methods

  1. The session on P/F ratio methods began by introducing the general idea of the methods, which is to estimate the level of fertility from census data on average numbers of children ever born to young women by using information on the age pattern of fertility derived from reports on numbers of children born to women during the 12 months prior to the census.
  1. There are numerous variants of the P/F ratio method, which was invented by William Brass, but all of them depend on either or both of two basic propositions of formal demography. The first proposition is that, if we know the age-specific fertility rates experienced by a cohort of women, we can calculate the mean number of children ever born to these women at any point in time. The second proposition is that the total fertility of any birth cohort of women estimates the period total fertility rate at the time members of this cohort reach the mean age at childbearing for the cohort.

18.  Several variants of the P/F ratio method were presented and illustrated with Malawi data from the 1998 and 2008 population censuses. Several of the newer variants involve use of the relational Gompertz model for the age pattern of fertility. The relational Gompertz model was defined, a simple computer spreadsheet technique for fitting the model to data was introduced, and the fits to data from the 2008 population census of Malawi illustrated.

Analyzing Census Data on Age at First Marriage

  1. The sessions on analyzing census data on age at first marriage and nuptiality tables began with a brief review of life table concepts, including three basic “columns” of the life table, the probabilities () column, the survival () column, and the frequency distribution () column, and the idea that life table concepts and methods may be applied not just to deaths, but to demographic events of all kinds—including first marriage.
  1. The source data and tabulations used for calculating nuptiality statistics include (a) a tabulation of ever married persons by sex, age at census in single years, and age at first marriage in single years and (b) a tabulation of total population by sex and single years of age. The total population is required to obtain numbers of never married persons, which are essential to the calculations because the define exposure to the risk of marriage.
  1. The simplest approach to age at first marriage statistics from these tables is to first calculate period mean age at first marriage based on numbers of first marriages. The calculation is trivial, requiring nothing but the definition of mean from elementary statistics, once the cells in the census tabulation have been suitably rearranged, but the rearrangement turns out to be moderately complicated, requiring some patience to construct a suitable computer spreadsheet.
  1. Though most simply calculated, mean age at first marriage based on numbers of first marriages is also the least satisfactory indicator of age at first marriage because numbers of marriages of younger women are typically far greater than numbers of marriages of older women simply because the population age distributions are “young,” with large numbers of persons at young ages and much smaller numbers at older ages.
  1. To control for the bias created by the population age distribution, mean ages at first marriages are computer from marriage frequencies, numbers of first marriages at age x to women a cohort age y at the time of the census divided by the total number of women in the cohort. This again is a simple calculation, though spreadsheet implementation is slightly complicated due to the large size of the table—if calculations are made for persons ages 15-79, there are over 4,000 cells in each table.
  1. Examination of the time series of mean ages at first marriage computed from marriage frequencies computer from the 2008 census data for Malawi shows that the series is unsatisfactory because the calculation fails to control for the changing composition of cohorts by marital status. Since only single women are “at risk” of first marriage, the most appropriate way to calculate a mean age at first marriage is to calculate a “nuptiality table” based on “probabilities” of first marriage in the life table sense of the word.

25.  Unlike the situation with life tables, in which probabilities must first be estimated from age-specific death rates, the census tabulations noted above provide for direct calculation of first marriage probabilities. The idea of the calculation is simple and simply illustrated though a full spreadsheet implementation is moderately complicated. Seminar participants were provided with a spreadsheet implementing all of the calculations.

Introduction to CensusInfo

  1. The presentation on “Introduction to CensusInfo” was delivered by Margaret Mbogoni. The objective of the presentation was to give an overview of the CensusInfo software within the context of the United Nations 2010 World Population and Housing Census Programme. The presentation summarized the essential goals of the 2010 World Population and Housing Census Programme as well as related census activities being undertaken by the United Nations Statistics Division. In this context, it was mentioned that the Division is undertaking activities to assist countries to disseminate their census results.
  1. Activities that are undertaken by the UNSD to enhance countries’ ability to disseminate results of their censuses include the project “Strengthening national capacity to analyze, present and disseminate data for evidence-based policy making”. Specific activities within the project include, holding of national seminars on census data analysis and also writing of thematic reports based on census data, in participating countries. In addition to this project, the UNSD is also organizing regional seminars on dissemination and analysis of data and is also planning to host an expert group meeting on strategies and technology for census data dissemination. It is anticipated that from the outcome of the regional seminars and of the expert group meeting, UNSD will compile and disseminate good national practices on strategies and technology for census data dissemination.
  1. The presenter further informed the audience that UNSD, in partnership with UNICEF and UNFPA has developed a software for census data dissemination called CensusInfo. This software is available for free on-line and UNSD is carrying out regional workshops as well as on-sight technical support to countries to enable them to create national adaptations of CensusInfo. It was mentioned that CensusInfo is a tool for disseminating population and housing census results on the web and on CD-ROM. It generates user-defined tables, graphs, maps and reports at different geographic levels, and with accompanying metadata. Furthermore, CensusInfo can be customized to meet country-specific needs in terms of tables and indicators to be generated. Another feature of CensusInfo, it was mentioned, is its ability to import data from other software applications, such as CSPro, SAS, SPSS and Redatam.
  1. In the presentation it was also mentioned that UNSD maintains a website on the 2010 World Population and Housing Census Programme on which there is a webpage on CensusInfo. The presentation also offered information on Global CensusInfo which is a UNSD adaptation of CensusInfo to census tables and indicators based on data that the Division collects from countries as part of the Demographic Yearbook System. The audience was informed that initially, Global CensusInfo will contain data from the 2000 round of population and housing censuses with data from the 2010 round being added as they become available.

General Discussion on Critical Gender Issues in Malawi in the Context of Development