Econometrics I 1

Download as pdf or txt
Download as pdf or txt
You are on page 1of 40

Econometrics I

Professor William Greene


Stern School of Business
Department of Economics

1-1/40 Part 1: Introduction


http://people.stern.nyu.edu/wgreene/Econometrics/Econometrics.htm

1-2/40 Part 1: Introduction


Abstract
Overview: This is an intermediate level, Ph.D. course in Applied
Econometrics. Topics to be studied include specification,
estimation, and inference in the context of models that include
then extend beyond the standard linear multiple regression
framework. After a review of the linear model, we will develop the
theory necessary for analysis of generalized linear and nonlinear
models.
Topics to be examined:
Regression modeling
Instrumental variables
Robust estimation and inference
Causal inference
Nonlinear modeling
Cross section, time series and panel data
Objective:
Preparation to read and carry out empirical social science
research using modern econometric methods.

1-3/40 Part 1: Introduction


Prerequisites

 A previous course that used linear regression


 Mathematical statistics
 Matrix algebra

1-4/40 Part 1: Introduction


Readings
 Main text: Greene, W.,
Econometric Analysis,
8th Edition,
Prentice Hall, 2017.

 A few articles
 Notes and materials on the course website:

http://people.stern.nyu.edu/wgreene/Econometrics/Econometrics.htm

1-5/40 Part 1: Introduction


Course Schedule
No class on:
Thursday, September 21
*Tuesday, November 21
Thursday, November 23
*Tuesday, November 28
Added class on:
Tuesday, December 12
Midterm: October 24
Final: Take home; posted December 7,
due December 19

1-6/40 Part 1: Introduction


Practicals
 Software
 NLOGIT provided, supported
http://people.stern.nyu.edu/wgreene/Econometrics/NLsetup.exe
 SAS, Stata, EViews optional, your choice
 R, Matlab, Gauss, others
 Questions and review as requested
 Problem Sets: (more details later)

1-7/40 Part 1: Introduction


Course Requirements

 Problem sets: 5 (30%)


 Midterm, in class (30%)
 Final exam, take home (40%)

1-8/40 Part 1: Introduction


Econometrics I

Part 1 - Paradigm

1-9/40 Part 1: Introduction


Econometrics: Paradigm
 Theoretical foundations
 Behavioral modeling: Optimization, labor supply,
demand equations, etc.
 Microeconometrics and macroeconometrics
 Mathematical Elements
 Statistical foundations
 ‘Model’ building – the econometric model
 Mathematical elements
 The underlying truth – is there one?

1-10/40 Part 1: Introduction


Objectives
 Understanding covariation:
 What is the income elasticity of the demand for
health care in an economy?
 Understanding a relationship:
 How will drivers respond to the availability of a
new toll road?
 Causal Inference: The search for
“causal” effects
 Did an antitrust intervention that broke up a
British boarding schools cartel have an impact
on fees charged?
 Prediction of an outcome of interest
 Does internet “buzz” help to predict movie
success?
1-11/40 Part 1: Introduction
Trends in Econometrics
 Small structural models
 Pervasiveness of an econometrics paradigm
 Non- and semiparametric methods vs. parametric
 Robust methods / Estimation and inference
 Nonlinear modeling (the role of software)
 Behavioral and structural modeling vs. “reduced form,”
“covariance analysis”
 Identification and “causal” effects

1-12/40 Part 1: Introduction


Data Structures
 Observation mechanisms
 Passive, nonexperimental (the usual)
 Randomly assigned experiment (wishful)
 Active, experimental (occasional)
 The ‘natural experiment’ (occasional, limited)
 Data types
 Cross section
 Pure time series
 Panel –
 Longitudinal data – NLS
 Country macro data – Penn W.T.
 Financial data
1-13/40 Part 1: Introduction
Paradigm: Classical Inference

Population Measurement

Econometrics
Characteristics
Imprecise inference about Behavior Patterns
the entire population –
sampling theory and Choices
asymptotics

1-14/40 Part 1: Introduction


Paradigm: Bayesian Inference

Population Measurement

Econometrics
Characteristics
Sharp, ‘exact’ inference about Behavior Patterns
only the sample – the ‘posterior’
density. Choices

1-15/40 Part 1: Introduction


An Application:
Cornwell and Rupert Labor Market Data
Is Wage Related to Education?
Cornwell and Rupert Returns to Schooling Data, 595 Individuals, 7 Years
Variables in the file are
EXP = work experience
WKS = weeks worked
OCC = occupation, 1 if blue collar,
IND = 1 if manufacturing industry
SOUTH = 1 if resides in south
SMSA = 1 if resides in a city (SMSA)
MS = 1 if married
FEM = 1 if female
UNION = 1 if wage set by union contract
ED = years of education
LWAGE = log of wage = dependent variable in regressions
These data were analyzed in Cornwell, C. and Rupert, P., "Efficient Estimation with Panel
Data: An Empirical Comparison of Instrumental Variable Estimators," Journal of Applied
Econometrics, 3, 1988, pp. 149-155.

1-16/40 Part 1: Introduction


1-17/40 Part 1: Introduction
Model Building in Econometrics

 Role of the assumptions


 Parameterizing the model
 Nonparametric analysis
 Semiparametric analysis
 Parametric analysis
 Sharpness of inferences

1-18/40 Part 1: Introduction


Application: Is there a relationship Nonparametric Regression
between (log) Wage and Education? Kernel regression of log wage on education

Semiparametric Regression: Parametric Regression: Least squares –


Least absolute deviations regression maximum likelihood – regression

1-19/40 Part 1: Introduction


A First Look at the Data
Descriptive Statistics

 Basic Measures of Location and Dispersion


 Graphical Devices
 Box Plots
 Histogram
 Kernel Density Estimator

1-20/40 Part 1: Introduction


1-21/40 Part 1: Introduction
Box Plots
Shows upward trend in median log wage

1-22/40 Part 1: Introduction


Histogram: Pooled data obscure
within person variation

What is the source of the variance, variation across people or variation over time?
1-23/40 Part 1: Introduction
Kernel density estimator suggests the underlying
distribution for a continuous variable
1 n 1  xi  xm* 
f̂(xm )   i1 K 
*

n B  B 
1 n

n i1
Q(x i | x *
m ,B), for a set of points x *
m

B  "bandwidth"
K  the kernel function
x*  the point at which the density is approximated.
f̂(x*) is an estimator of f(x*)

1-24/40 Part 1: Introduction


The kernel density
estimator is a
histogram (of sorts).

1 n 1  xi  xm* 
f̂(xm )   i1 K 
* *
, for a set of points x m
n B  B 
B  "bandwidth" chosen by the analyst
K  the kernel function, such as the normal
or logistic pdf (or one of several others)
x*  the point at which the density is approximated.
This is essentially a histogram with small bins.

1-25/40 Part 1: Introduction


1-26/40 Part 1: Introduction
1-27/40 Part 1: Introduction
From Jones and Schurer (2011)
Stylized Box Plot

1-28/40 Part 1: Introduction


From Jones and Schurer (2011)

1-29/40 Part 1: Introduction


Objective: Impact of Education
on (log) Wage

 Specification: What is the right model to


use to analyze this association?
 Estimation
 Inference
 Analysis

1-30/40 Part 1: Introduction


Simple Linear Regression

LWAGE = 5.8388 + 0.0652*ED

1-31/40 Part 1: Introduction


Multiple Regression

1-32/40 Part 1: Introduction


Nonlinear Specification:
Quadratic Effect of Experience

1-33/40 Part 1: Introduction


Model Building in Econometrics
A Model Relating (Log)Wage to Gender and Education & Experience

1-34/40 Part 1: Introduction


Partial Effects
Coefficients may not tell the story

Education .05654
Experience .04045 - 2*.00068*Exp
FEM -.38922

1-35/40 Part 1: Introduction


Effect of Experience = .04045 - 2*.00068*Exp
Positive from 1 to 30, negative after.

1-36/40 Part 1: Introduction


Interaction Effect
Gender Difference in Partial Effects

1-37/40 Part 1: Introduction


Partial Effect of a Year of Education
E[logWage]/ED = ED + ED*FEM *FEM
= 0.05458 + 0.02350*FEM
Note, the effect is positive.
The effect 43% is larger for women.

1-38/40 Part 1: Introduction


Gender Effect Varies by Years of Education
E[logWage]/FEM = FEM + ED*FEM *ED

= -0.67961 + 0.0235*ED

1-39/40 Part 1: Introduction


Analysts are interested in interaction effects in models.

1-40/40 Part 1: Introduction

You might also like