Empirical development economics / Måns Söderbom, and Francis Teal ; with Markus Eberhardt, Simon Quinn and Andrew Zeitlin.

Includes bibliographical references (pages 423-430) and index.

Machine generated contents note: pt. I Linking models to data for development
1.An introduction to empirical development economics
1.1.The objective of the book
1.2.Models and data: the Harris
Todaro model
1.3.Production functions and functional form
1.3.1.The Cobb-Douglas production function
1.3.2.The constant elasticity of substitution (CES) functional form
1.4.A model with human capital
1.5.Data and models
1.5.1.The macro GDP data
1.5.2.Interpreting the data
References
Exercise
SECTION I Cross-section data and the determinants of incomes
2.The linear regression model and the OLS estimator
2.1.Introduction: models and causality
2.2.The linear regression model and the OLS estimators
2.2.1.The linear regression model as a population model
2.2.2.The zero conditional mean assumption
2.2.3.The OLS estimator
2.3.The Mincerian earnings function for the South African data
2.4.Properties of the OLS estimators
Contents note continued: 2.4.1.The assumptions for OLS to be unbiased
2.4.2.The assumptions for OLS to be minimum variance
2.5.Identifying the causal effect of education
3.Using and extending the simple regression model
3.1.Introduction
3.2.Dummy explanatory variables and the return to education
3.3.Multiple regression
3.3.1.Earnings and production functions
3.3.2.The OLS estimators for multiple regression
3.3.3.Omitted variables and the bias they may cause
3.4.Interpreting multiple regressions
3.4.1.How much does investing in education increase earnings? Some micro evidence
3.4.2.How much does investing in education increase productivity? Some macro evidence
4.The distribution of the OLS estimators and hypothesis testing
4.1.Introduction
4.2.The distribution of the OLS estimators
4.2.1.The normality assumption
4.2.2.Why normality?
Contents note continued: 4.3.Testing hypotheses about a single population parameter
4.3.1.The T distribution
4.3.2.The t-test
4.3.3.Confidence intervals
4.4.Testing for the overall significance of a regression
4.5.Testing for heteroskedasticity
4.6.Large sample properties of OLS
4.6.1.Consistency
4.6.2.Asymptotic normality
5.The determinants of earnings and productivity
5.1.Introduction
5.2.Testing the normality assumption
5.3.The earnings function
5.3.1.Bringing the tests together
5.3.2.Robust and clustered standard errors
5.4.The production function
5.4.1.Testing the production function
5.4.2.Extending the production function
5.5.Interpreting our earnings and production functions
5.5.1.Can education be given a causal interpretation?
5.5.2.How much does education raise labour productivity?
SECTION II Time-series data, growth and development
Contents note continued: 6.Modelling growth with time-series data
6.1.Introduction: modelling growth
6.2.An introduction to the Solow model
6.3.A Solow model for Argentina
6.4.OLS estimates under the classical assumptions with time-series data
6.4.1.Assumptions for OLS to be unbiased
6.4.2.The variance of the OLS estimators
6.4.3.Testing for autocorrelation
6.5.Static and dynamic time-series models
6.6.Assumptions to ensure the OLS estimators are consistent
6.7.Spurious regression with nonstationary time-series data
6.8.A brief summary
7.The implications of variables having a unit root
7.1.Introduction and motivation
7.2.Testing for a unit root and the order of integration
7.3.Cointegration
7.4.How are growth and inflation related in Argentina?
7.5.The error-correction model
7.6.Causality in time-series models
7.7.Cross-section and time-series data
Contents note continued: 8.Exogenous and endogenous growth
8.1.The Solow model and the history of development
8.2.Long-term growth and structural change
8.3.The Solow model, structural change and endogenous growth
8.4.Human capital and the dynamic Solow model
8.5.Exogenous and endogenous growth
8.6.A Solow interpretation of development patterns
Appendix: deriving the dynamic Solow model
SECTION III Panel data
9.Panel data: an introduction
9.1.Introduction
9.2.Panel data
9.2.1.The structure of the panel
9.2.2.Panel data and endogeneity
9.3.Panel production functions
9.3.1.A panel macro production function
9.3.2.A panel micro production function
9.4.Interpreting the fixed effect
Appendix: matrix notation
10.Panel estimators: POLS, RE, FE, FD
10.1.Introduction
10.2.Panel estimators
10.2.1.The fixed effects and first difference estimators
Contents note continued: 10.2.2.The random effects estimator
10.3.Key assumptions for consistency
10.4.Model selection
10.4.1.Testing for correlation between the ci and the explanatory variables
10.4.2.Testing for the presence of an unobserved effect
10.5.The micro panel production function extended
10.6.What determines the productivity of Ghanaian firms?
11.Instrumental variables and endogeneity
11.1.Introduction
11.2.Sources of bias in the OLS estimates
11.2.1.Bias from omitted variables
11.2.2.Bias from measurement error
11.2.3.Panel data: omitted variables and measurement error
11.3.Instrumental variables
11.3.1.Valid and informative instruments
11.3.2.Interpreting the IV estimator
11.4.The properties of the IV estimator
11.4.1.The IV and OLS estimators compared
11.4.2.Inference with the IV estimator
11.5.The causes of differences in world incomes
Contents note continued: SECTION IV An introduction to programme evaluation
12.The programme evaluation approach to development policy
12.1.Introduction: causal effects and the counterfactual problem
12.2.Rubin causal model
12.2.1.Potential outcomes
12.2.2.Assignment mechanism
12.2.3.Defining measures of impact
12.2.4.From potential outcomes to regression
12.3.Selection on observables
12.3.1.Ignorability of treatment
12.3.2.Overlap
12.4.Unconditional unconfoundedness and the experimental approach
13.Models, experiments and calibration in development policy analysis
13.1.Introduction
13.2.Empirical estimators under (conditional) unconfoundedness
13.2.1.Multivariate regression
13.2.2.Panel data methods
13.3.A randomised controlled trial (RCT) for conditional cash transfers
13.4.Calibrating technology
13.5.Education, technology and poverty
pt. 2 Modelling development
Contents note continued: 14.Measurement, models and methods for understanding poverty
14.1.Introduction
14.2.The causes of poverty
14.2.1.Poverty and GDP data
14.2.2.Poverty, consumption and incomes
14.2.3.Poverty, inequality and GDP
14.3.The Mincerian earnings function, the price of labour and poverty
14.4.Modelling impacts
14.4.1.A generalised Roy model of selection
14.4.2.Implications of the Roy model for estimation of treatment effects
14.5.An overview: measurement, models and methods
SECTION V Modelling choice
15.Maximum likelihood estimation
75.7.Introduction
15.2.The concept of maximum likelihood
15.3.The concept of population
15.4.Distributional assumptions and the log-likelihood function
15.5.Maximising the (log-) likelihood
15.6.Maximum likelihood in Stata
75.7.Problems and warnings ...
15.7.1.Maximum likelihood and endogeneity
15.7.2.Maximum likelihood and convergence
Contents note continued: 15.8.Properties of maximum likelihood estimates
15.8.1.Consistency
15.8.2.Efficiency
15.8.3.So what?
75.9.Hypothesis testing under maximum likelihood
15.10.Overview
16.Modelling choice: the LPM, probit and logit models
76.7.Introduction
16.2.Binary choices and interpreting the descriptive statistics
16.3.Estimation by OLS: the linear probability model
16.4.The probit and logit models as latent variable models
16.4.1.The probit model
16.4.2.The logit model
76.5.Maximum likelihood estimation of probit and logit models
76.6.Explaining choice
17.Using logit and probit models for unemployment and school choice
17.1.Introduction
17.2.Interpreting the probit model and the logit model
17.2.1.A model of unemployment
17.2.2.Average partial effects and marginal effects at the mean
17.2.3.Age and education as determinants of unemployment in South Africa
Contents note continued: 17.3.Goodness of fit
17.4.Indian private and state schools
17.4.1.How well do private schools perform?
17.4.2.Who attends a private school?
17.4.3.Mother's education and wealth as determinants of attending private school in India
17.5.Models of unemployment and school choice
18.Corner solutions: modelling investing in children and by firms
18.1.Introduction
18.2.OLS estimation of corner response models
18.2.1.Investment in Ghana's manufacturing sector
18.2.2.Gender discrimination in India
18.3.The Tobit model
18.4.Two-part models
18.4.1.Truncated normal hurdle model
18.4.2.The log-normal hurdle model
18.5.Overview
Appendix: the Inverse Mills Ratio (IMR)
SECTION VI Structural modelling
19.An introduction to structural modelling in development economics
19.1.Introduction: the challenge of using microeconomic theory in empirical research
Contents note continued: 19.2.Using a structural model to think about risk-sharing
19.3.Building and solving a microeconomic model
19.4.Thinking about unobservables and choosing an estimator
19.4.1.The model to be estimated
19.4.2.Identification in the model
19.4.3.Testing the model
19.5.Estimating the model
19.5.1.The data
19.5.2.Estimation results
19.6.Conclusion
20.Structural methods and the return to education
20.1.Introduction: Belzil and Hansen go to Africa
20.2.The question
20.3.A model of investment in education
20.4.Thinking about unobservables and choosing an estimator
20.5.Models and data
20.5.1.`Adolescent econometricians'?
20.5.2.Possible applications for structural modelling in development
20.6.Structural models: hubris or humility?
SECTION VII Selection, heterogeneity and programme evaluation
Contents note continued: 21.Sample selection: modelling incomes where occupation is chosen
21.1.Introduction
21.2.Sample selection
21.3.A formal exposition
21.3.1.The regression with sample selection
21.3.2.Modelling the correlation of the unobservables
21.4.When is sample selection a problem?
21.5.Selection and earnings in South Africa
21.6.Corner solution and sample selection models
22.Programme evaluation: regression discontinuity and matching
22.1.Introduction
22.2.Regression discontinuity design
22.3.Propensity score methods
22.3.1.Regression using the propensity score
22.3.2.Weighting by the propensity score
22.3.3.Matching on the propensity score
22.4.Food aid in Ethiopia: propensity-score matching
22.5.Assessing the consequences of property rights: pipeline identification strategies
22.6.Estimating treatment effects (the plot so far)
Contents note continued: 23.Heterogeneity, selection and the marginal treatment effect (MTE)
23.1.Introduction
23.2.Instrumental variables estimates under homogeneous treatment effects
23:3.Instrumental variables estimates under heterogeneous treatment effects
23.3.1.IV for noncompliance and heterogeneous effects: the LATE Theorem
23.3.2.LATE and the compliant subpopulation
23.4.Selection and the marginal treatment effect
23.4.1.Interpreting the LATE in the context of the Roy model
23.4.2.The marginal treatment effect
23.4.3.What does IV identify?
23.5.The return to education once again
23.6.An overview
SECTION VIII Dynamic models for micro and macro data
24.Estimation of dynamic effects with panel data
24.1.Introduction
24.2.Instrumental variable estimation of dynamic panel-data models
24.3.The Arellano
Bond estimator
24.3.1.No serial correlation in the errors
24.3.2.Serially correlated errors
Contents note continued: 24.4.The system GMM estimator
24.5.Estimation of dynamic panel-data models using Stata
24.6.The general case
24.6.1.The regressors are strictly exogenous
24.6.2.The regressors are predetermined
24.6.3.The regressors are contemporaneously endogenous
24.6.4.Implications of serial correlation in the error term
24.7.Using the estimators
Appendix: the bias in the fixed effects estimator of a dynamic panel-data model
25.Modelling the effects of aid and the determinants of growth
25.1.Introduction
25.2.Dynamic reduced-form models
25.2.1.Aid, policy and growth
25.2.2.Dynamics and lags
25.2.3.Differenced and system GMM estimators
25.3.Growth rate effects: a model of endogenous growth
25.3.1.Dynamic and growth rate models
25.3.2.Is there evidence for endogenous growth?
25.4.Aid, policy and growth revisited with annual data
25.4.1.Cross section and time-series uses of macro data
Contents note continued: 25.4.2.Growth and levels effects of aid
25.5.A brief overview: aid, policy and growth
SECTION IX Dynamics and long panels
26.Understanding technology using long panels
26.1.Introduction
26.2.Parameter heterogeneity in long panels
26.3.The mean group estimator
26.4.Cross-section dependence due to common factors
26.5.Conclusion
27.Cross-section dependence and nonstationary data
27.1.Introduction
27.2.Alternative approaches to modelling cross-section dependence
27.2.1.Country fixed effects and year dummies
27.2.2.Estimating unobserved common factors
27.2.3.Constructing weight matrices
27.3.Modelling cross-section dependence using cross-section averages
27.4.Detecting cross-section dependence
27.5.Panel unit root testing
27.5.1.First-generation panel unit root test
27.5.1.1.The Im, Pesaran and Shin test (IPS)
27.5.1.2.The Maddala and Wu test (MW)
Contents note continued: 27.5.2.Second-generation panel unit root test
27.5.2.1.The PANIC approach
27.5.2.2.The CIPS and CIPSM tests
27.6.Cointegration testing in panels
27.6.1.Residual analysis and error-correction models
27.6.2.Tests for panel cointegration
27.7.Parameter heterogeneity, nonstationary data and cross-section dependence
28.Macro production functions for manufacturing and agriculture
28.1.Introduction
28.2.Estimating a production function for manufacturing
28.2.1.The homogeneous models
28.2.2.The heterogeneous models
28.3.Estimating a production function for agriculture
28.3.1.Unit roots
28.3.2.What determines the productivity of agriculture?
28.4.Manufacturing and agriculture and the growth of an economy
SECTION X An overview
29.How can the processes of development best be understood?
29.1.Introduction
29.2.A range of answers as to the causes of poverty
Contents note continued: 29.3.Macro policy, growth and poverty reduction
29.4.Programme evaluation and structural models
29.4.1.Programme evaluation and the `failure' of poverty policies
29.4.2.Structural models and understanding the causes of poverty
29.5.Skills, technology and the returns on investment
29.5.1.The value of skills
29.5.2.The role of technology
29.5.3.Rates of return on investment
29.6.A final word
References.

"Understanding why so many people across the world are so poor is one of the central intellectual challenges of our time. What explains a pattern of extreme destitution for billions combined with plutocratic levels of income for a tiny minority? This book offers a novel approach to addressing those issues, not by providing answers, but seeking to provide the tools and data that will enable the student, the researcher and the professional working in this area to investigate the questions for themselves. Empirical Development Economics has been designed as a hands-on teaching tool to investigate the causes of poverty. The book begins by introducing the basics of the quantitative approach to development economics. All the topics are presented through data that addresses some important policy issue. In Part 1 the focus is on the basics of understanding why incomes differ so much. What is the role of education, technology and institutions in ensuring that where you are born is so important in determining whether you are poor? In Part 2 the focus is on techniques that allow us to address questions which include how firms invest, how households decide how much to spend on education of their children, whether microfinance does help the poor, whether food aid works, who gets private schooling and whether property rights enhance investment.A distinctive feature of the book is its presentation of a range of approaches to studying development questions. Development economics has undergone a major change in focus over the last decade with the rise of experimental methods to address development issues. One of our objectives has been to show how such methods relate to more traditional ones. "--