|
26. Analysis of poverty or inequality in ONE developing country
Description
There is an almost endless list of factors that may be associated with being poor as there are too for the extent of inequality. This project focuses on the microeconomic factors, such as unemployment, sector and occupation, retirement or old age, family composition, illness or disability, skills, differences between geographical areas, between men and women, or people of different races.
You should seek to examine the relative importance of some of these factors in order to analyse the nature of poverty and/or inequality in a particular developing country.
NB: This is one of several development related topics (see 23,24,25 and 26)
Method
You need to first decide which country you would like to analyse. There are many possibilities as most countries have large household survey datasets on-line which we can download for free. Previous students have worked on Brazil, Pakistan, Tanzania and Vietnam. The first step will be to prepare the data so that you can analyze it. This will vary by dataset, and can be a long and time consuming process, so do not underestimate it.
You then need to decide if you are focusing on poverty or inequality.
There are several steps
• What indicator will you use: e.g. income or expenditure? If you analysing poverty, how will you define poverty – in a relative or absolute sense, and where will you draw the poverty line? For either poverty or inequality, how will you take account of economies of scale and differences in need between households? Two households with the same income level but comprised of say, a single adult in one and two adults and four children in the other, should probably not be treated as the same.
• You should then calculate a set of summary measures of poverty/inequality, to give an idea of the extent of the issue, followed by descriptive statistics of the sample: proportions that fall into different categories (employment, occupation, sector, education, family size and composition, age, gender, region etc), as well as poverty/inequality statistics for these groups. This can all be done in STATA using a set of existing programs.
• The next step is to estimate a model that might explain some of the variations in income or expenditure levels. You are aiming to investigate the significance, sign and size of the relationship between some plausible explanatory factors and the welfare indicator you have chosen e.g. income or expenditure per capita.
• If you have chosen poverty you will most likely estimate a probit model where the dependent variable takes values of 0 if not poor, and 1 if poor. Alternatively you could use the poverty gap as your dependent variables.
• For inequality, you might want to use natural log of income (or expenditure) as your dependent variable. log(yi)=+iXi + ei where the X’s are the set of household characteristics. These may include characteristics of the household head: age, gender, race, education, employment status, marital status; broader household characteristics: household size, dependency ratio; and spatial variables: regions, or urban and rural for example. #p#分页标题#e#
• You should investigate whether some continuous variables are better transformed into dummy variables, or whether there may be a non-linear relationship between some, like age and education, and log income.
You might also consider interactions between some variables, such as race and education, or gender and education.
If you find that some variables give significant results you might also consider splitting the sample by this variable, for example gender of household head, or region, or urban v rural. This will allow you to examine whether your explanatory factors affect income differently [i.e. the coefficients vary] for different sub-groups of the population. Take care with degrees of freedom if you decide to split the sample. You can then test formally for equality of the coefficients across the two sub-samples. For example you may find the “return” to education is much lower for ethnic minorities, or women, or rural dwellers – is this evidence of discrimination?
Data
The data used here will be household survey data. There are lots of these available for developing countries: the easiest to work with are the Living Standards Measurement Surveys: www.worldbank.org/lsms for a list of countries and years covered.
Although these are the easiest to work with, they do require quite substantial work in matching and merging variables form across the different modules of the survey, so think carefully about how prepared you are to spend hours/days on data manipulation.
In the department we have useable version of data already compiled for Brazil, Vietnam and Pakistan – and probably more than that. Email some of the development economists about data for countries you might be really keen to work on.
Key readings (some additional readings may be posted on SyD or provided by your supervisor).
Justino, P. and J. Litchfield (2004) Welfare in Vietnam during the 1990s: Poverty, inequality and poverty dynamics, Journal of the Asia Pacific Economy vol9, 145-169
Ray, D., 1998, Development Economics. (Princeton, New Jersey: Princeton University Press).
Parikh, Ashok and Kunal Sen (2006) “Probit with heteroscedasticity: an application to Indian poverty analysis” Applied Economics Letters, Volume 13, Issue 11 September 2006, pages 699 – 707.
Grootaert, Christiaan (1997) “The Determinants of Poverty in Côte d'Ivoire in the 1980s” J Afr Econ (1997) 6(2): 169-196
Poverty in china. Is my topics
|
