Research Methodology For Childrens Education
4.1 Introduction to the research questions.
Children are the future of a nation. They are the adults of tomorrow and will contribute to the economic development of the nation. It is an established fact that nations that have achieved high rates of economic growth and made the transition from a developing country to a developed one (such as South Korea) have emphasised the development of human capital. People are a resource and can contribute to a nation’s progress by participating in economic activity. Consequently, there is a need to educate the population, remove barriers to the education of girls and encourage the participation of women in the labour force.
The objective of this study is to understand how children allocate their time between work, leisure and schooling. This can provide important insights into the reasons why children are sent to work, girls made to undertake housework and children deprived of schooling. The use of both primary and secondary sources is made. The micro data from the PSLM surveys is used for understanding the bigger picture for Pakistan as a whole. This is complemented with a survey undertaken of different communities in Karachi and analysing the data collected to identify the factors that influence the decision of a household regarding the allocation of their children’s time between different activities.
The research questions that are addressed are:
Why do children work?
Why are children not enrolled in school?
What factors contribute to a child’s educational attainment?
Population and sampling
A survey of children living in some selected poor communities was conducted. The purpose of a survey is to collect data from units who are usually individual respondents referred to as an element. The objective of inferential research is to generalize from the sample to all potential elements, termed a population.
One approach to research is to focus on a specific population or a complete set of units being studied. However, this is not feasible due to time and cost constraints. In such circumstances a sample needs to be drawn from the population. In order to draw a representative sample the ideal situation is to list all the elements of the population. In the context of the present study this would imply listing all the children in the communities selected for the research. In practice to obtain a complete list of the population is not possible. Consequently one needs to work with an incomplete list called the sampling frame.
The sampling frame was the poor communities who have a large number of migrant families representing a diversity of ethnic backgrounds. A sample of five communities was selected for the purpose of this study. A sample of approximately 100 children was drawn from each community. Some of these children were interviewed in the local school and some in the homes. This was deemed necessary to capture the characteristics of both the school going and non-school going children. The elements are the children in the age group of 5 to 17 years. In the first stage the communities were selected. In the second stage schools within these communities were selected. The sampling frame for the second stage consists of the schools. The third stage was to select the children enrolled in the schools and the households to which these school children belonged to interview their siblings who were either working or not attending school. Single stage samples are drawn in one step from a sampling frame. Multistage samples are drawn in two or more steps. In these steps or sample stages only the last one identifies the elements interviewed.
A sampling unit consists of either the element or the group of elements chosen at the sampling stage. A cluster is a sampling unit consisting of groups of elements. Clusters usually involve existing groupings. Multistage cluster sampling designs involve several steps: choosing initial sampling units or clusters, listing the elements within the selected clusters, and picking elements from the chosen clusters. Multistage cluster sampling can save resources. Instead of enumerating all the elements in the population, only the elements in the selected clusters need to be enumerated. The rationale behind the design of this study was to concentrate on interviews in limited clusters to reduce survey travel and personnel time.
Elements can be categorized by stratification. A stratum consists of all the elements that have a common characteristic. For instance in the present study the clusters were stratified according to the characteristics of the communities. Working with strata requires an enumeration of all elements from which a stratum can then be defined. In contrast, one could identify a cluster without knowing the elements it contains. This is the approach adopted in the present study. The reason for using cluster sampling is that the population was large and widely dispersed. This necessitated the selection of communities rather than randomly selecting from the whole population. Stratified sampling on the other hand would involve ordering the sampling frame by one or more characteristics and then selecting the same percentage of people from each subgroup either using simple random or systematic sampling. This was not possible with the limited resources available. Stage sampling which is an extension of cluster sampling was used in which more than one level of grouping is used to generate the sample. In this instance communities were selected then schools and then students within the class and their households.
Since the research goals in this present study do not require generalizations for the entire population of Karachi it was possible to use both clusters and strata. The clusters (i.e. communities) were stratified according to such variables as income, ethnicity of population, presence of government or charitable school. Particular schools and households were drawn from each stratum before drawing children (elements) from the chosen schools and households.
It should be noted that the sampling strategy adopted needs to balance feasibility and purpose. In this study the knowledge about the population was limited and it was not possible to enumerate the entire population. The purpose of this research is to understand the time allocation of children among different activities and does not require the same level of precision as may be necessary in other studies.
At times the objective is to compare and contrast subpopulations rather than to represent all elements equally. In this case the objective is to identify the variables that affect the probability of a child going to school or working. Since one of the variables deemed to influence the household’s choice is ethnicity the subpopulations are based on the ethnicity of the households. This has resulted in oversampling of some communities such as the Christian or Swati community.
Nevertheless, within each subpopulation, stratum or cluster, it is possible to retain probability sampling by using systematic sampling with a random start. The selection of households within a community or the selection of student in a school was systematic.
It is possible to combine the results from non probability proportionate to size (PPS) samples into an estimate for the whole population by statistically adjusting for any over or under sampling through a procedure called weighting.
4.2.1 Sampling Techniques
There are two broad approaches to sampling (i) Probability sampling and (ii) Nonprobability sampling.
(i). Probability sampling.
Probability sampling includes random sampling, systematic sampling and stratified sampling. Sampling bias occurs when sampling procedures consistently miss some kinds of elements while over representing others. To avoid bias every element has to have an equal chance of being sampled. First, an unbiased sampling frame is needed that does not exclude certain kinds of elements. Second the actual selection of elements from the frame must give the elements in the frame an equal probability of selection, a method called probability sampling.
(ii). Nonprobability Sampling. The alternative to probability sampling, called non-probability sampling includes any method in which the elements have unequal chances of being selected. In nonprobability sampling, members are selected from the population in some non-random manner. These include convenience sampling, judgment sampling, quota sampling, and snowball sampling.
Convenience sampling refers to a sample which is selected because it is convenient to do so. The respondents depend on availability. This non-probability method is used during preliminary research to get an estimate of results without incurring too much cost or time that would be needed to select a random sample.
An extension of convenience sampling is judgment sampling where the sample selection is based on the judgement of the researcher. If this method is used the researcher has to ensure that the selected sample represents the entire population.
In purposive sampling respondents are selected because of certain characteristics. In this study the aim is to analyse the characteristics of children going to school and going to work. Some of the migrant communities that the present research was interested in would form a small proportion of the overall population. To find 100 children belonging to one community, a survey using random sampling would have to interview 10,000 children in the general population. Since the researcher could not afford such a large survey, purposive sampling had to be resorted to.
In quota sampling the researcher tries to create a sample that matches some predetermined demographic profile. For instance if it is known that 57 percent of the adults in a community are female and the study requires 100 total respondents then quota sampling would entail interviewing the first 57 females and the first 46 males at a local market. Expanding the quota procedure so that respondents fit several different criteria such as sex, age or ethnicity does not guarantee that the sample will be unbiased on some overlooked but important dimension. Quota sampling can be complex if the sample has to fit several dimensions.
The research methodology is both qualitative and quantitative. Using a purely quantitative approach has its shortcomings since numbers may fail to reflect the true gravity of the problem. A purely quantitative approach on the other hand may not give robust results that can be used for analysing the state of children living in poverty in different communities in Karachi.
When social scientists must go beyond the demographic data of the census to measure more complex constructs, they have to develop their own surveys. Fortunately, samples do not have to be very large relative to the population. [pg 119]
A cross sectional survey collects data at one time point. We can generalize finding from such studies to the sampled population only at the time of the survey.
Medium is the method of gathering data from surveyed respondents – mail, phone, face to face. Each of these media presents special advantages and disadvantages in cost, sampling method, success in gaining cooperation from the respondents, type of content, and format of the questions.
Face to Face. Personal contact maximizes trust and cooperation between interviewer and interviewee. Face to face contact decreases refusals and permits questioning on more intimate topics. It also allows the use of special aids such as cards showing answer options. The interviewer can see and assess the respondent’s nonverbal behavior and habitat. Face to face interviews can include respondents without phones or the ability to read a mailed questionnaire.
A very important consideration in choosing face to face interviews as a medium of conducting the survey was the literacy level of the respondents and the lack of access to a telephone. In designing this survey these constraints had to be taken into consideration which meant adopting a survey method that was more time consuming.
Were it not for limited resources, more researchers would employ face to face interviewing. This medium has proven too expensive for many researchers who have been compelled to turn to other methods such as phone or mail surveys.
Defining Sample Error: Survey researchers cannot avoid all sample error which can be considered as sampling variability. We can easily compute the variance if we have multiple samples drawn from a population. However, we seldom have multiple samples and so must usually estimate sampling variance from a single sample.
We could estimate sampling variance by dividing a single sample into a number of subsamples. The variance among the subsamples describes the variability for a sample of the size of each of the subsamples. However, larger samples (such as the size of the original sample) have less variability than smaller ones.
Sampling Variability and Sample Size. The principle that larger sample sizes give more precise estimates of population values is called the law of large numbers. Larger sample sizes provide more chances for unrepresentative elements (those much higher or lower than the group average) to cancel each other out. This principle helps in estimating sampling error directly from an undivided sample.
Determining Sample Size. Because sample size affects confidence interval, one could, in principle, select the sample size to yield any given degree of confidence. Unfortunately, this approach has several problems. The formulas require information not usually known until after the survey, such as the estimated sampling variability. Moreover, the formulas rest on such assumptions as the normality of the sampling distribution. Practical problems arise as well. Most surveys measure several different variables, each of which may require different levels of measurement precision and thus different minimal sample sizes. Even when researchers know the ideal sample size they may have to work with a smaller sample because of budget limits.
In the absence of precise estimates of required sample size and within the constraints, most researchers use rule of thumb guesswork to set sample size. One approach sets sample size by reference to earlier surveys of a similar kind. If subgroup comparisons seem especially important, the researcher may set overall sample size so as to achieve some minimum number in the smallest subgroup of interest.
Non sampling or data collection error can exceed sampling error and needs to be addressed carefully.
The completion rate measures the extent to which a sample is successfully reached and cooperates. A low completion rate raises concern about possible non-observation bias. The problem arises if the non-contacted and non-cooperative members of the sample differ from those who are contacted and cooperative. This could result in sample bias.
In the present study all interviews were carried out face to face by the researcher with a 100% completion rate. All the households approached for data collection purposes cooperated. A point to note here is the perception of the interviewees is that if someone is taking an interest in them and asking questions then some action may be taken to improve their plight.
The objective of the research is to find the determinants of child labour and school enrolment in the poor migrant communities of Karachi. A migrant community is defined as one that came to Karachi in the last 30 years and still has close links with its town or village of origin. There are two types of migration voluntary and forced. Voluntary migration takes place when families decide to move in search of better job opportunities and higher incomes. Forced migration occurs when households are displaced due to war or natural disasters. The lure of economic betterment acts as a pull factor in the event of voluntary migration whereas war, civil strife or natural disasters act as push factors in case of forced migration.
The selection of communities is based on their characteristics and also accessibility. The following communities have been included in this study.