Module 1

 

INTRODUCTION TO SURVEYS

 

TABLE OF CONTENTS

Introduction
What is a survey?
What kinds of questions are asked?
When is a survey conducted?
How is survey information used?
Types of surveys
Concerns when using survey data
BAIS data Information

 

 

 

 

 

 

 

INTRODUCTION

Prior to analyzing survey data, in an attempt to answer questions relating to the social and economic conditions of a given population, we must first understand: where the survey data comes from, how it is collected, and what concerns, if any, a researcher should have in using survey data.

In the following module, you will be given a basic introduction to surveys which will hopefully enable you to use household survey data in future research.

 

WHAT IS A SURVEY?

Have you ever been asked to fill out a customer satisfaction questionnaire? Or asked to give your opinion about a political issue? If you answered "no" to both of these questions, congratulations! You have just taken part in your first survey. If you answered "yes," then you are already an expert.

In general, a survey usually involves the collection of data by interviewing a sample of people. Often times, however, the group of interest is too large to interview completely. It would take too much time and money to question everyone in China, each farmer in South America, or each car owner in Europe. In these cases, a smaller selection, excerpt, or sample of the larger population is selected for questioning. The group of people, households, towns, villages, cities, or countries that participate in the survey is called the survey sample. Each person in the selected sample is asked the same series of questions, and the responses are then organized so that conclusions can be drawn from them.

 

WHAT KINDS OF QUESTIONS ARE ASKED?

Survey questions may ask about a wide range of topics such as people's behavior, attitudes, how and where they live, and information about their economic and social backgrounds (demographic data).

Behavior - Many survey questions relate to people's actions or behavior in various social areas. In economic research, for example, patterns of spending and saving have been studied intensively using surveys. Other types of projects examine such behaviors as voting, reading, visits to the doctor, geographical movements, and recreational habits.

Attitudes - Some survey questions concern people's beliefs, opinions, attitudes, or expectations. These are areas in which it is unlikely that data would be available from non-survey sources.

Environment - In many surveys it is important to have data on the circumstances in which respondents live in order to interpret their responses more accurately. This includes information about the neighborhood, the adequacy of living quarters, membership in groups and organizations, and so on.

Personal Descriptive (demographic) - Surveys always include questions regarding personal-social characteristics of respondents such as sex, age, occupation, income, and education. This kind of information helps researchers interpret the opinions people express in surveys, since people who share particular demographic characteristics also tend to share opinions on social issues. For instance, we can find out how many people aged 20-29 own cars and how car owners in this age group compare with those in other age groups. Certain kinds of demographic data, however, are more significant and useful in some areas of research than in others. For example, in an election study, a respondent's nationality or where he or she grew up would be more important to know than in an economic study.

 

WHEN IS A SURVEY CONDUCTED?

Surveys are conducted when the desired information cannot be obtained in a more accurate and less expensive manner. For example, it would be inefficient and too costly to conduct a sample survey to determine the number of automobiles in use in the United States. This information can usually be obtained more quickly and more reliably from state licensing bureaus. On the other hand, a survey would be required to obtain information about automobile users. For instance, no recorded information is available on the occupations, buying intentions, driving habits, or other characteristics of automobile owners. In this case, interviewing each individual car owner in the United States would be prohibitively expensive and time consuming, thus it would be far more feasible, efficient, and less expensive to collect information on only a small representative sample of car owners.

 

HOW IS SURVEY INFORMATION USED?

After gathering various kinds of information, the study director is then in a position to analyze the data and consider the problem in light of these data. He or she will be concerned with such matters as how many, who, how, and why. A survey on the use of public libraries, for instance, might be designed to answer questions like these:

The answers to such questions are of interest to many people. For example, librarians, educators, and publishers would find them useful when they consider the practical problems of improving facilities. They might also be useful to theoretical researchers interested in studying the manner in which the public is reached and influenced by information from various sources. After the survey data have been analyzed, a paper or report is usually prepared which is generally published and made available to the public. The survey results can then be used as a basis for changing, improving or even discarding existing procedures. Survey data are also used by other researchers to increase their knowledge of a field or to define new areas for investigation.

 

TYPES OF SURVEYS

Today, surveys are being used in public opinion polling and other types of social research, market research, and census taking. These types of surveys and the various ways in which they are used are described briefly below:

Descriptive Statistical Survey - This kind of survey is conducted largely by governments to obtain major descriptive information about their population and its density, the composition of their labor force, national health statistics, and so forth. From early on, governments have taken inventories of their human resources for taxation, military, and legislative purposes. In addition, democratic procedures in government brought forth a need for better communication between people and their officials. Government agencies such as the U.S. Census Bureau conduct or sponsor a wide variety of surveys designed to collect information on many aspects of life in this country.

Survey for Social Research - The social survey is a major research method used by social scientists to gather and analyze information about the social and economic conditions of a population or segments of a population. This information leads to a better understanding of human beings in their social setting.

Market Research Survey - Another type of survey activity is carried on by businesses in consumer market research. Each year, many surveys are conducted by market research groups to determine consumer needs and the effectiveness of marketing programs. Market surveys focus on consumers' attitudes toward current products as well as on possible demand for new products.

 

CONCERNS WHEN USING SURVEY DATA

Up to this point things may seem a bit simple, but there are several aspects related to surveys one must consider carefully prior to using any survey data.

Does the survey design introduce bias into the data?

Both the phrasing and the manner in which questions are asked can introduce bias into a survey. Bias occurs when the results obtained in a survey are systematically skewed and thus do not accurately reflect the real state of the world. With bias in the data, we cannot develop accurate predictions using descriptive or inferential statistics.

A Potential Bias - "What does this question really measure?"

One concern is related to the actual survey questions themselves. It is quite possible that the information respondents are providing for a given question is not what the researcher was looking for. For example, if we are interested in how much a respondent earns in a year, does it matter how we ask our questions? Is it satisfactory to ask someone "How much do you make?"

This again may seem quite straight forward, but given the tremendous variation in the background and circumstances of the respondents, wording of questions can have a significant impact on the qualtiy of information received.

Returning to our example, does asking "How much do you make?" accurately assess what we are trying to measure? That is, by asking this question will the respondent give us the amount he or she earns in a year? What would you guess?

Given our question, it is quite possible that the respondent would provide us with his or her hourly, weekly, or monthly earnings. It is also possible that the respondent would provide us with the earnings from one job, even though he or she is employed at two or three jobs. Finally, the respondent could accurately provide us with the earnings received in the last year, but instead of providing this information for him or herself alone, the answer could reflect the entire household or family earnings for that year.

In general, the wording of a question can play a tremendous role in the information respondents provide, thus when using survey data it is important to understand what the information truly represents. Further, it is important to consider the ordering of questions in a survey, who asks the questions, and how the survey questions are asked, as each of these are also potential sources of bias in the data.

What level of information are you interested in?

Responses can be measured for any unit of analysis. Example units of analysis are: individuals, households, neighborhoods, businesses, provinces, or even countries. For this reason, you need to determine at which level you wish to conduct your survey, or if someone else has conducted the survey, at which level you are going to analyze the data.

For example, a survey may collect information at the household level, meaning the responses are in terms of the respective household. Thus, a question related to income would reflect the income of the entire household; a question related to the amount of food eaten would reflect the amount eaten by the entire household.

It is also possible for a survey to ask questions that will result in numerous levels of information. For example, a survey may have questions relating to the amount of money earned by the entire family (which would produce a "family" level variable), while at the same time have questions relating to how much the individual earns (which would produce an "individual" level variable).

Given this added complexity, it is important to determine early on in the research process - at what level is the survey conducted. From here, one can decide the types of analyses and the level of information to use.

 

BAIS DATA INFORMATION

Throughout this workshop, we are going to use the statistics program Stata and the BAIS data set to examine questions in areas such as education, health care, pregnancy, sexual activity, and STDs. Below is a a detailed description and summary of the Botswana AIDS Impact Survey which should prove useful in your future analyses. The following information comes directly from the Botswana Central Statistics Office Website.

The full document can be reviewed at:   http://www.cso.gov.bw/cso/aids_imp.html

 

Introduction

In most parts of the world, individuals, communities and their leaders struggle with the best way to address concerns about HIV and AIDS. For Botswana, like in many other countries, the HIV/AIDS epidemic represents one of the most serious social problems in recent history, especially for countries in the Sub Saharan Africa. Southern Africa holds the majority of the world’s hard hit countries.

The first cases of HIV in Botswana were diagnosed in 1985. The rapid spread of the infection and AIDS in Botswana over the past 13 years has been tremendous. UNAIDS estimates indicate that by the end of 1999, at least one in four adults in Botswana was living with HIV.

Botswana like other countries, has implemented programs to improve the knowledge about how HIV is transmitted as well as strategies for HIV/AIDS prevention and control. The current strategies include Sentinel surveillance; i.e surveys which are conducted annually or biannually to obtain information on the prevalence of HIV/AIDS, monitor trends of HIV/AIDS infection, provide information for program planning, monitoring and evaluation and assess the impact of intervention programs.

The Central Statistics Office conducted the 2001 Botswana AIDS Impact Survey. The Botswana Government provided funding and Botswana UNDP office assisted with consultancy fees for all consultants engaged at various stages of the survey.

 

Objectives

The Botswana AIDS Impact survey was conducted to obtain more information on topics related to HIV/AIDS. The main objective of this survey was to provide information to:

i. Assess whether programs are operating as intended
ii. Assess performance of intervention programs
iii. Asses whether people are changing their sexual behaviour
iv. Establish the proportion of people in need of care due to HIV infection
v. Establish the proportion of people who are at risk of HIV infection
vi. Assess the impact of the pandemic at household level
vii. Provide information on issues related to the impact of HIV/AIDS on household and communities.

 


Survey Methodology

Frame: The Botswana Multiple Indicator Survey (BMIS) of 2000 collected data on health indicators. The BMIS 2000 sample served as the sampling frame for the BAIS. Ninety eight (98) sample points out of 215 BMIS sampling points were selected. The target population in BAIS is the same as in BMIS.

Sample Design: The sample for the 2001 Botswana AIDS Impact Survey was designed to provide estimates of AIDS indicators at the national level, urban and rural areas, and for the fourteen districts: Gaborone, Francistown, Lobatse, Selebi-Phikwe, Small towns (Jwaneng, Orapa, and Sowa), Southern, South East, Kweneng, Kgatleng, Central, North-East, North West, Ghanzi, and Kgalagadi.

A stratified two-stage probability sample design was utilised for the selection of the sample.

The first stage was the selection of Enumeration Areas (Eas) as Primary Sampling Units (PSUs) selected with probability proportional to measures of size (PPS), where measures of size (MOS) are the number of households in the EAs as listed in MIS 2000. In all 98 EAs were selected with pps out of 215 EAs.

At the second stage of sampling, the households were systematically selected from a fresh list of occupied households prepared at the beginning of the survey's fieldwork (i.e. listing of households for the selected EAs). Overall 2000 valid households were drawn systematically.

The sample is not self-weighting because it was stratified by districts.




Questionnaires

i. Household Questionnaire: A household questionnaire was administered in each selected household, which collected information on various topics related to HIV/AIDS on household members including sex, age, marital status, education, water and sanitation, and orphanhood status.

The Household questionnaire contains the following topics:

Demographic characteristics: Age, sex, marital status, place of usual living and citizenship.
Education
Parental survival and fostering
Economic activity
Deaths in household
Care and support for sick people
Water and sanitation

 

ii. Individual Questionnaire: The individual questionnaire for the 2001 Botswana AIDS Impact Survey was based on the UNAIDS Model Questionnaire with some modifications and additions. Some of the modifications include lowering the age limit of the eligible persons to 10 years, increasing the upper limit to 64 years and modifying some questions in order to adapt the questionnaire to Botswana situation. In addition to the household questionnaire, questionnaires were administered in each household for men and women aged 10-64 years. The individual questionnaire for men and women contains the following topics:

Marriage and cohabiting partnerships.
Sexual history and behaviour.
Sexually transmitted diseases.
Knowledge about HIV/AIDS and exposure to interventions.
Attitudes towards people living with HIV/AIDS and HIV testing.
Child bearing and antenatal care.
 

iii. Community Questionnaire: The community questionnaire aimed to obtain information a) about the perceived impact of AIDS on the community, b) the problems caused by AIDS, and c) the coping mechanism and responses developed by the community.



Fieldwork and Processing

The field-staff was trained from 4th January to 18th January 2001. The data were collected by fifteen teams; each comprising of four interviewers (two males and two females), one driver, and a supervisor. There were a total of 98 blocks (EAs) and each team was assigned an average of 5-8 blocks. The fieldwork started on the 20th January and ended on the 5th January 2001. During the fieldwork two questions related to sexual behaviour especially sexual partners were dropped due to high rate of refusals experienced.

Before data entry was carried out, the questionnaires were edited to check if all the relevant questions have been responded to and coded according to the codes designed for the study. Data entry was carried out between February and March, 2001 by 8 data entry operators under the supervision of one programmer/supervisor. Consistency checks on the data set as per the Computer Edit Specifications designed by the subject matter specialist were performed. Data editing began in April and finished in June, 2001. The data tabulation and analysis was completed at the end of November.



Response Rates

Of the 2,126 households selected for the Botswana AIDS Impact sample, 2,023 were found occupied. Of these, 1,781 were successfully interviewed, yielding a household response rate of 88.0 percent. The response rates in urban and rural areas were 90.3 percent and 85.8 percent respectively. In the interviewed households, 4,728 eligible persons aged 10-64 years were identified. Of these, 4,494 were successfully interviewed, yielding a response rate of 95.1 percent.

 

For more information on the collaborating partners, visit their home pages:

University of Botswana Central Statistics Office of Botswana
UB - Demography Department The World Bank

 

BACK TO TOP