Department of Economics, SUFFOLK UNIVERSITY
Jonathan Haughton
ECONOMICS 842: DEVELOPMENT ECONOMICS
ASSIGNMENT 4
Answers to this assignment are due back by Thursday, October 2, 2014.
1.
Basic Measurement of Vulnerability to Poverty
We have the following information on the income of five households. You may assume that income is
normally distributed. Fill in the blanks. [Hint: Excel might help.]
Income Poverty line
100
120
130
160
220



2.
SD of
Probability of
VulnerProbability of poverty at
income poverty next year
ability*
least once in next two years
125
10
125
12
125
22
125
20
125
30
Highly vulnerable:
If probability of poverty next year is > 0.5.
Somewhat vulnerable: If probability of poverty next year is >P0 but <= 0.5.
Not vulnerable:
If probability of poverty next year is <=P 0.
Measuring vulnerability with data from Bangladesh
Measure the proportion of households in Bangladesh who were “highly vulnerable to poverty” in 1998.
Here are the steps that you will probably want to take:
a. Start Stata, and then open the file called final.dta. The idea is to use the data to construct and
estimate a regression model where the dependent variable is the log of consumption per capita.
b. Create a variable called lpcexp that is the (natural) log of per capita expenditure.
c. Regress this variable on those variables that might reasonably be expected to predict a household’s
per capita expenditure level. This would include the age, education, and gender of the head, size of
the household, amount of land owned, assets owned, whether the household has electricity,
distance from the nearest road, and distance from the nearest bank. Include dummy variables for
the districts (thana); a good way to do this is to include in your regressors the term i.thana .
d. Keep the predicted output (lpcexphat) and residuals (resid). For this, use the commands
predict lpcexphat, xb
predict resid, r
e. Create a variable called resid2 that is the square of the residuals.
f. Regress the square of the residuals on the same variables as in iii. and save the predicted value
(estvar).
Ec 842: Economic Development
Assignment 4
Page 1 of 4
g. Replace any negative values of estvar with resid2 .
h. Construct a variable (call it flessc) that is (log of food poverty line – estimated log of
consumption)/(square root of estimated variance).
i. Compute the probability of poverty for each household using normal(flessc).
j. Construct a variable called vul1 that is =1 if the household has at least a 50% probability of being
poor next year.
k. Tabulate and report the results.
l. [Challenging] Redo this exercise on the assumption that the age of the household head has risen by
5 and the household assets have increased by 20%. Hint: You do not need to re-run the regressions,
but you will need to generate new predictions of lpcexphat and estvar1.
3.
i.
An assortment of quick (?) questions
The squared poverty gap index (sometimes referred to as the poverty severity index) is obtained by
computing the square of the poverty gap index, which puts more weight on the very poor.
True
False
ii. The Watts index of poverty is
A
Computed by dividing the poverty line by income for all individuals and taking the average of the
log of this ratio over the sample.
B
Computed by dividing the poverty line by income for all individuals below the poverty line and
taking the average of the log of this ratio over the sample.
C
Computed by dividing the poverty line by income for all individuals and taking the log of the
average of this ratio over the sample.
D
Computed by dividing the poverty line by income for all individuals below the poverty line and
taking the log of the average of this ratio over the sample.
iii. Information on expenditure per capita has been gathered for year 1 and year 2. The poverty rate appears to
be lower in year 2. First order stochastic dominance means that
A
The poverty incidence curve for year 1 is everywhere above the poverty incidence curve for year
2.
B
The headcount poverty rate in year 2 is lower than the headcount poverty rate in year 1, for any
poverty line.
C
For any level of expenditure per capita, the cumulative percentage of the population below the
poverty line is lower in year 2 than year 1.
D
All of the above.
iv. A chronically poor household is defined as a household that
A
Is in poverty in every year covered by the panel survey.
B
Is poor for at least one of the years covered by the panel survey.
C
Is expected to be poor next year.
D
Has an average level of consumption (or income) per capita that is below the poverty line.
v. Which of the following is not a strength of panel data (relative to data from repeated cross-sections)?
A
Panel data remain representative of the population as a whole.
B
Panel data increase the precision with which one can measure changes in poverty over time (for a
given sample size).
C
Only panel data allow one to measure expenditure (or income) transitions over time.
D
Panel data are helpful in measuring the impact of government programs.
Ec 842: Economic Development
Assignment 4
Page 2 of 4
vi. Panel data suffer from attrition bias, which occurs because
A
Over time, the panel does not reflect some parts of the population (such as newly-formed
couples).
B
Some households decline to answer the survey the second time around.
C
Cluster sampling becomes increasingly difficult over time.
D
Durable assets wear out over time.
vii. Here is some information about poverty in Indonesia. According to these data,
Poverty transition matrix, December 1998-August 1999
% of population
December1998
Not poor
Poor
Total
Not poor
67.8
13.9
81.8
August 1999
Poor
6.2
12.1
18.3
Total
74.0
26.0
100.0
Source: Said and Widyanti (2001), Table 7.
A
B
C
D
26.0% of the population was persistently poor.
32.2% of the population was chronically poor.
Poverty rose in Indonesia between 1998 and 1999.
Indonesia rebounded rapidly from the 1997 financial crisis and poverty fell.
viii.
Which of the following is not part of the normal process of determining the poverty rate in a country?
A
It is necessary to impute the rental value of a household’s durable goods when
measuring expenditure.
B
A poverty line needs to be determined.
C
Census data are required to determine the proportion of people who are poor.
D
It is assumed that members of a household have the same level of welfare.
ix.
Which of the following is not generally a problem faced by researchers when measuring vulnerability?
A
The welfare measure (e.g. expenditure per capita, or its log) may not be normally
distributed, leading to mismeasurement of the expected probability of being in
poverty.
B
Measurement error leads one to overstate vulnerability.
C
In the absence of panel or longitudinal data, the variance of expenditure cannot be
estimated fully satisfactorily.
D
The level of expenditure reported by households is understated, which leads to an
underestimate of vulnerability.
x.
According to the World Bank, the “dollar-a-day” poverty rate for less-developed countries approximately
halved between 1981 and 2010, and most of the reduction was due to rapid reductions in poverty in
China and India.
True
False
Uncertain
xi.
Survey data suffer from non-response bias and under-reporting bias. Typically these lead to:
A
An overstatement of the headcount poverty rate and an understatement of the degree
of inequality.
B
An overstatement of the headcount poverty rate and an overstatement of the degree
of inequality.
C
An understatement of the headcount poverty rate and an understatement of the
degree of inequality.
Ec 842: Economic Development
Assignment 4
Page 3 of 4
D
xii.
An understatement of the headcount poverty rate and an overstatement of the degree
of inequality.
In the example set out in Table 1 (reproduced here), suppose that the exchange rate were Rp.50/$
instead of Rp.46/$. Then the value of Indian GDP/capita, using US prices, would be
Computing GDP/capita in purchasing power parity (PPP) terms
US
Price ($)
Quantity
Teachers
30,000
1,000
Computers
250
40,000
GDP
$40 million
Population
GDP/capita
1,000,000
$40
GDP/capita, US prices
GDP/capita, Indian prices
A
$1.90.
B
$1.75.
C
$16.50.
D
$2.07.
$40
Rp.496.8
India
Price (Rp)
Quantity
36,800
1,000
11,500
12,000
Rp. 174.8 million
= $4.6 million at 46 Rp/$
2,000,000
Rp.87.4
= $1.90 at 46 Rp/$
$16.50
Rp.87.4
xiii.
Which of the following steps is not part of the World Bank’s methodology for computing the dollar-a-day
world poverty rate?
A
Pick a poverty line that allows households to buy enough food and other basic needs.
B
Use an average PPP exchange rate to translate the poverty line into domestic currency
terms.
C
Use a domestic consumer price index to find the appropriate poverty line, in domestic
currency, for the years in which household surveys were undertaken.
D
Measure the number of poor in a country by determining, based on fitting Lorenz
curves, how many fall below the poverty line.
xiv.
A household has expected per capita consumption of 95 and the variance of this consumption is 36. The
poverty line is 100. Then the probability that the household will be poor next year is:
A
0.555
B
0.798
C
0.445
D
0.202
Ec 842: Economic Development
Assignment 4
Page 4 of 4