The Nature of Econometrics and Economic Data: - Steps in Econometric Analysis

9/21/2021 9/21/2021
Introductory Econometrics: A Modern Approach (7e) Introductory Econometrics: A Modern Approach (7e)
The Nature of Econometrics and Economic Data (2 of 22)

• Steps in econometric analysis
1) Economic model (this step is often skipped)
2) Econometric model
Chapter 1 • Economic models

• Maybe micro- or macromodels
The Nature of Econometrics and • Often use optimizing behaviour, equilibrium modeling, …
Economic Data
• Establish relationships between economic variables
• Examples: demand equations, pricing equations, …
© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a © 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 1 password-protected website or school-approved learning management system for classroom use. 3
1 3
The Nature of Econometrics and Economic Data (1 of 22) The Nature of Econometrics and Economic Data (3 of 22)
• What is econometrics? • Economic model of crime (Becker (1968))
• Econometrics is the use of statistical methods to analyze economic data. • Derives equation for criminal activity based on utility maximization.
• Econometricians typically analyze nonexperimental data.
• Typical goals of econometric analysis:
• Estimating relationships between economic variables.
• Testing economic theories and hypotheses.
• Evaluating and implementing government and business policy.
• Common applications
• Forecasting macroeconomic variables (interest rates, inflation rates,
• Functional form of relationship not specified.
GDP).
• Equation could have been postulated without economic modeling.
• Forecasting non-macro variables (less visible).
2 4
1 2
9/21/2021 9/21/2021
• Model of job training and worker productivity • Econometric model of job training and worker productivity
• What is the effect of additional training on worker productivity?
• Formal economic theory not really needed to derive equation:
• Most of econometrics deals with the specification of the error u.

• Econometric models may be used for hypothesis testing.
• For example, the parameter β3 represents the effect of training on wages.
• Other factors may be relevant, but these are the most important. • How large is this effect? Is it different from zero?
5 7
• Econometric model of criminal activity • Econometric analysis requires data.
• The functional form has to be specified. • There are several different kinds of economic data sets:
• Variables may have to be approximated by other quantities. • Cross-sectional data
• Time series data
• Pooled cross sections
• Panel/Longitudinal data
• Econometric methods depend on the nature of the data used.

• Use of inappropriate methods may lead to misleading results.
6 8
3 4
9/21/2021 9/21/2021
• Cross-sectional data sets • Table 1.2: Cross-sectional data on growth rates and country characteristics
• These may include samples of individuals, households, firms, cities,
states, countries, or other units of interest at a given point of time or in a
given period. obsno
1
country
Argentina
gpcrgdp
0.89
govcons60
9
second60
32
2 Austria 3.32 16 50
• Cross-sectional observations are more or less independent. 3

4
Belgium
Bolivia
2.56
1.24
13
18
69
12
. . . . .
• An example is pure random sampling from a population. .
.
.
.
.
.
.
.
.
.
• Sometimes pure random sampling is violated, for example, people refuse 61 Zimbabwe 2.30 17 6
to respond in surveys, or sampling may be characterized by clustering.

• Cross-sectional data is typically encountered in applied microeconomics.
9 11
• Table 1.1: Cross-sectional data set on wages and other characteristics • Time series data
• This includes observations of a variable or several variables over time.
obsno
1
wage
3.10
educ
11
exper
2
female
1
married
0
• Examples include stock prices, money supply, consumer price index,
2
3
3.24
3.00
12
11
22
2
1
0
1
0
gross domestic product, annual homicide rates, automobile sales, and so
4
5
6.00
5.30
8
12
44
7
0
0
1
1
on.
• Time series observations are typically serially correlated.
. . . . . .
. . . . . .
. . . . . .
525
526
11.56
3.50
16
14
5
5
0
1
1
0
• Ordering of observations conveys important information.
• Data frequency may include daily, weekly, monthly, quarterly, annually,
and so on.
• Typical features of time series include trends and seasonality.
• Typical applications include applied macroeconomics and finance.
10 12
5 6
9/21/2021 9/21/2021
• Table 1.3: Time series data on minimum wage, unemployment, and • Table 1.4: Pooled cross sections on two years of housing prices
related data for Puerto Rico
obsno year hprice proptax sqrft bdrms bthrms
1 1993 85,500 42 1600 3 2
obsno year avgmin avgcov prunemp prgnp 2 1993 67,300 36 1440 3 2
1 1950 0.20 20.1 15.4 878.7 3 1993 134,000 38 2000 4 2
2 1951 0.21 20.7 16.0 925.0 . . . . . . .
3 1952 0.23 22.6 14.8 1015.9 . . . . . . .
. . . . . . . . . . . . .
. . . . . . 250 1993 243,600 41 2600 4 3
. . . . . . 251 1995 65,000 16 1250 2 1
37 1986 3.35 58.1 18.9 4281.6 252 1995 182,400 20 2200 4 2
38 1987 3.35 58.2 16.8 4496.7 253 1995 97,500 15 1540 3 2
. . . . . . .
. . . . . . .
. . . . . . .
520 1995 57,200 16 1100 2 1
13 15
• Pooled cross sections • Panel or longitudinal data
• Two or more cross sections are combined in one data set. • The same cross-sectional units are followed over time.
• Cross sections are drawn independently of each other. • Panel data have a cross-sectional and a time series dimension.
• Pooled cross sections are often used to evaluate policy changes. • Panel data can be used to account for time-invariant unobservables.
• Example: • Panel data can be used to model lagged responses.
• Evaluating effect of change in property taxes on house prices. • Example:
• Random sample of house prices for the year 1993. • City crime statistics; each city is observed in two years.
• A new random sample of house prices for the year 1995. • Time-invariant unobserved city characteristics may be modeled.
• Compare before/after (1993: before reform, 1995: after reform). • Effect of police on crime rates may exhibit time lag.
14 16
7 8
9/21/2021 9/21/2021
• Table 1.5: Two-year panel data set on city crime statistics • Causal effect of fertilizer on crop yield
• “By how much will the production of soybeans increase if one increases
obsno city year murders population unem police the amount of fertilizer applied to the ground.”
• Implicit assumption: all other factors that influence crop yield such as
1 1 1986 5 350,000 8.7 440
2 1 1990 8 359,200 7.2 471
3
4
2
2
1986
1990
2
1
64,300
65,100
5.4
5.5
75
75 quality of land, rainfall, presence of parasites, and so on are held fixed.
. . . . . . .
. . . . . . .
. . . . . . .
297
298
149
149
1986
1990
10
6
260,700
245,000
9.6
9.8
286
334
• Experiment = Feasible
299
300
150
150
1986
1990
25
32
543,000
546,200
4.3
5.2
520
493
• Choose several one-acre plots of land; randomly assign different amounts
of fertilizer to the different plots; compare yields.
• Experiment works because amount of fertilizer applied is unrelated to
other factors influencing crop yields.
17 19
• Causality and the notion of ceteris paribus • Measuring the return to education
• “If a person is chosen from the population and given another year of
education, by how much will his or her wage increase?”
• Implicit assumption: all other factors that influence wages such as
experience, family background, intelligence, and so on are held fixed.
• Ceteris paribus: “other relevant factors being equal.”
• Most economic questions are ceteris paribus questions. • Experiment ≠ Infeasable
• It is important to define which causal effect one is interested in. • Choose a group of people; randomly assign different amounts of
• It is useful to describe how an experiment would have to be designed to education to them (infeasable!); compare wage outcomes.
infer the causal effect in question. • Problem without random assignment, amount of education is related to
other factors that influence wages (e.g. intelligence).
18 20
9 10
9/21/2021 9/21/2021
• Effect of law enforcement on city crime level • Testing predictions of economic theories
• “If a city is randomly chosen and given ten additional police officers, by • Economic theories are not always stated in terms of causal effects.
how much would its crime rate fall?” • For example, the expectations hypothesis states that long-term interest
• Alternatively: “If two cities are the same in all respects, except that city A rates equal compounded expected short-term interest rates.
has ten more police officers than city B, by how much would the two
cities’ crime rates differ?”
• Experiment ≠ Infeasable
• Randomly assign number of police officers to a large number of cities
(virtually impossible, as no two cities are alike in all respects except size • An implicaton is that the interest rate of a three-month T-bill should be equal to
of police force!). the expected interest rate for the first three months of a six-month T-bill; this
• More importantly, in reality, the number of police officers occurs can be tested using econometric methods.
contemoraneously with determination of crime rate.
21 23
Introductory Econometrics: A Modern Approach (7e)
The Nature of Econometrics and Economic Data (21 of 22)

• Effect of the minimum wage on unemployment
• “By how much (if at all) will unemployment increase if the minimum
wage is increased by a certain amount (holding other things fixed)?”
• Experiment ≠ Infeasable
• Government randomly chooses minimum wage each year and observes
unemployment outcomes.
• Experiment would theoretically work because level of minimum wage is
unrelated to other factors determining unemployment.
• In reality, the level of the minimum wage will depend on political and
economic factors that also influence unemployment.
© 2020 Cengage. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part, except for use as permitted in a license distributed with a certain product or service or otherwise on a
password-protected website or school-approved learning management system for classroom use. 22
22
11 12
The Simple Regression Model (2 of 39)

• Interpretation of the simple linear regression model
• Explains how y varies with changes in x
Chapter 2
The Simple Regression Model
• The simple linear regression model is rarely applicable in practice but

its discussion is useful for pedagogical reasons.
The Simple Regression Model (1 of 39) The Simple Regression Model (3 of 39)
• Definition of the simple regression model • Example: Soybean yield and fertilizer
• “Explains variable y in terms of variable x”
• Example: A simple wage equation
• When is there a causal interpretation?
• Conditional mean independence assumption
• Example: wage equation
• Population regression function (PFR) • Deriving the ordinary least squares estimates
• The conditional mean independence assumption implies that • In order to estimate the regression model one needs data
• A random sample of n observations
• This means that the average value of the dependent variable

can be expressed as a linear function of the explanatory variable.
• Deriving the ordinary least squares (OLS) estimators • Example of a simple regression
• Defining regression residuals • CEO salary and return on equity
• Minimize the sum of the squared regression residuals
• Fitted regression
• OLS estimators
• Causal interpretation?
• OLS fits as good as possible a regression line through the data points
• Example of a simple regression • Properties of OLS on any sample of data
• Wage and education • Fitted values and residuals
• Fitted regression • Algebraic properties of OLS regression
• Causal interpretation?
obsno roe salary salaryhat uhat
• Example of a simple regression 1 14.1 1095 1224.058 -129.058 • This table presents fitted
values and residuals for 15
• Voting outcomes and campaign expenditures (two parties) 2
3
10.9
23.5
1001
1122
1164.854
1397.960
-163.854
-275.969 CEOs.
4 5.9 578 1072.348 -494.348
5 13.8 1368 1218.508 149.493 • For example, the 12th CEO’s
6 20.0 1145 1333.215 -188.215 predicted salary is $526,023
7 16.4 1078 1266.611 188.611 higher than their actual
• Fitted regression
8 16.3 1094 1264.761 -170.761
salary.
9 10.5 1237 1157.454 79.546
10 26.3 833 1449.773 -616.773
• By contrast the 5th CEO’s
11 25.9 567 1442.372 -875.372
12 26.8 933 1459.023 -526.023
predicted salary is $149,493
13 14.8 1339 1237.009 101.991
lower than their actual
14 22.3 937 1375.768 -438.768
salary.
• Causal interpretation? 15 56.3 2011 2004.808 6.192
• Goodness of fit • CEO Salary and return on equity
• How well does an explanatory variable explain the dependent variable?
• Measures of variation:
• Voting outcomes and campaign expenditures
• Caution: A high R-squared does not necessarily mean that the regression
has a causal interpretation!
• Decomposition of total variation • Incorporating nonlinearities: Semi-logarithmic form
• Regression of log wages on years of education
• Goodness-of-fit measure (R-squared) • This changes the interpretation of the regression coefficient:
• Fitted regression • CEO salary and firm sales: fitted regression
• The log-log form postulates a constant elasticity model, whereas the

semi-log form assumes a semi-elasticity model.
• Incorporating nonlinearities: Log-logarithmic form • Expected values and variances of the OLS estimators
• CEO salary and firm sales • The estimated regression coefficients are random variables because
they are calculated from a random sample
• This changes the interpretation of the regression coefficient:

• The question is what the estimators will estimate on average and how
large will their variability be in repeated samples
• Standard assumptions for the linear regression model
• Assumption SLR.1 (Linear in parameters)
• Assumption SLR.2 (Random sampling)
• Discussion of random sampling: Wage and education • Assumptions for the linear regression model (cont.)
• The population consists, for example, of all workers of country A • Assumption SLR.3 (Sample variation in the explanatory variable)
• In the population, there is a linear relationship between wages (or log wages)
and years of education.
• Draw completely randomly a worker from the population
• The wage and the years of education of the worker drawn are random because
one does not know beforehand which worker is drawn.
• Throw that worker back into the population and repeat the random draw n
times. • Assumption SLR.4 (Zero conditional mean)
• The wages and years of education of the sampled workers are used to estimate
the linear relationship between wages and education.
• Theorem 2.1 (Unbiasedness of OLS) • Graphical illustration of homoskedasticity
• Interpretation of unbiasedness
• The estimated coefficients may be smaller or larger, depending on the sample
that is the result of a random draw.
• However, on average, they will be equal to the values that characterize the true
relationship between y and x in the population.
• “On average” means if sampling was repeated, i.e. if drawing the random
sample and doing the estimation was repeated many times.
• In a given sample, estimates may differ considerably from true values.
• Variances of the OLS estimators • An example for heteroskedasticity: Wage and education
• Depending on the sample, the estimates will be nearer or farther away from
the true population values.
• How far can we expect our estimates to be away from the true population
values on average (= sampling variability)?
• Sampling variability is measured by the estimator‘s variances
• Assumption SLR.5 (Homoskedasticity)
• Theorem 2.2 (Variances of the OLS estimators) • Theorem 2.3 (Unbiasedness of the error variance)
• Under assumptions SLR.1 – SLR.5:
• Calculation of standard errors for regression coefficients
• Conclusion:
• The sampling variability of the estimated regression coefficients will be the
higher, the larger the variability of the unobserved factors, and the lower, the The estimated standard deviations of the regression coefficients are called “standard errors.” They
higher the variation in the explanatory variable. measure how precisely the regression coefficients are estimated.
• Estimating the error variance • Regression on a binary explanatory variable
• Suppose that x is either equal to 0 or 1
• This regression allows the mean value of y to differ depending on the

state of x
• Note that the statistical properties of OLS are no different when x is

binary
• Counterfactual outcomes, causality and policy analysis • Random assignment
• In policy analysis, define a treatment effect as: • Subjects are randomly assigned into treatment and control groups such that
there are no systematic differences between the two groups other than the
treatment.
• In practice, randomized control trials (RCTs) are expensive to implement and
may raise ethical issues.
• Note that we will never actually observe this since we either observe • Though RCTs are often not feasible in economics, it is useful to think about the
yi(1) or yi(0) for a given i, but never both. kind of experiment you would run if random assignment was a possibility. This
helps in identifying the potential impediments to random assignment (that we
• Let the average treatment effect be defined as: could conceivable control for in a multivariate regression).
• Counterfactual outcomes, causality and policy analysis (contd.) • Example: The effects of a job training program on earnings
• Let xi be a binary policy variable. • Real earnings are regressed on a binary variable indicating
participation in a job training program.
• This can be written as:
• Therefore, regressing y on x will give us an estimate of the (constant) • Those who participated in the training program have earnings $1,790
treatment effect. higher than those who did not participate.
• As long as we have random assignment, OLS will yield an unbiased • This represents a 39.3% increase over the $4,550 average earnings
estimator for the treatment effect τ. from those who did not participate.
Multiple Regression Analysis: Estimation (2 of 37)

• Motivation for multiple regression
• Incorporate more explanatory factors into the model
• Explicitly hold fixed other factors that otherwise would be in
• Allow for more flexible functional forms
Chapter 3 • Example: Wage equation
Multiple Regression Analysis:

Estimation
Multiple Regression Analysis: Estimation (1 of 37) Multiple Regression Analysis: Estimation (3 of 37)
• Definition of the multiple linear regression model • Example: Average test scores and per student spending
• “Explains variable y in terms of variables x1, x2,…, xk”
• Per student spending is likely to be correlated with average family

income at a given high school because of school financing.
• Omitting average family income in regression would lead to biased
estimate of the effect of spending on average test scores.
• In a simple regression model, effect of per student spending would
partly include the effect of family income on test scores.
• Example: Family income and family consumption • OLS Estimation of the multiple regression model
• Random sample
• Regression residuals
• Model has two explanatory variables: inome and income squared
• Consumption is explained as a quadratic function of income
• One has to be very careful when interpreting the coefficients: • Minimize sum of squared residuals
• Example: CEO salary, sales and CEO tenure • Interpretation of the multiple regression model
• Model assumes a constant elasticity relationship between CEO salary and the • The multiple linear regression model manages to hold the values of
sales of his or her firm. other explanatory variables fixed even if, in reality, they are correlated
• Model assumes a quadratic relationship between CEO salary and his or her with the explanatory variable under consideration.
tenure with the firm.
• “Ceteris paribus”-interpretation
• Meaning of “linear” regression • It has still to be assumed that unobserved factors do not change if the
• The model has to be linear in the parameters (not in the variables) explanatory variables are changed.
• Example: Determinants of college GPA • “Partialling out” interpretation of multiple regression
• One can show that the estimated coefficient of an explanatory

variable in a multiple regression can be obtained in two steps:
• 1) Regress the explanatory variable on all other explanatory variables
• Interpretation • 2) Regress on the residuals from this regression
• Holding ACT fixed, another point on high school grade point average is
associated with another .453 points college grade point average • Why does this procedure work?
• Or: If we compare two students with the same ACT, but the hsGPA of student A • The residuals from the first regression is the part of the explanatory variable
is one point higher, we predict student A to have a colGPA that is .453 higher that is uncorrelated with the other explanatory variables.
than that of student B • The slope coefficient of the second regression therefore represents the isolated
• Holding high school grade point average fixed, another 10 points on ACT are effect of the explanatory variable on the dep. variable.
associated with less than one point on college GPA
• Properties of OLS on any sample of data • Goodness-of-Fit
• Fitted values and residuals • Decomposition of total variation
• R squared
• Algebraic properties of OLS regression

• Alternative expression for R squared
• Example: Explaining arrest records • Standard assumptions for the multiple regression model
• Assumption MLR.1 (Linear in parameters)
• Assumption MLR.2 (Random sampling)

• Interpretation:
• If the proportion prior arrests increases by 0.5, the predicted fall in arrests is 7.5
arrests per 100 men.
• If the months in prison increase from 0 to 12, the predicted fall in arrests is
0.408 arrests for a particular man.
• If the quarters employed increase by 1, the predicted fall in arrests is 10.4
arrests per 100 men.
• Example: Explaining arrest records (cont.) • Standard assumptions for the multiple regression model (cont.)
• An additional explanatory variable is added.
• Assumption MLR.3 (No perfect collinearity)
• In the sample (and therefore in the population), none of the independent
variables is constant and there are no exact linear relationships among the
independent variables.
• Interpretation: • Remarks on MLR.3

• Average prior sentence increases number of arrests (?) • The assumption only rules out perfect collinearity/correlation bet-ween
• Limited additional explanatory power as R-squared increases by little explanatory variables; imperfect correlation is allowed.
• If an explanatory variable is a perfect linear combination of other explanatory
• General remark on R-squared variables it is superfluous and may be eliminated.
• Even if R-squared is small (as in the given example), regression may still provide • Constant variables are also ruled out (collinear with intercept).
good estimates of ceteris paribus effects.
• Example for perfect collinearity: small sample • Including irrelevant variables in a regression model
• Example for perfect collinearity: relationships between regressors

• Omitting relevant variables: the simple case
• Standard assumptions for the multiple regression model (cont.) • Omitted variable bias
• Assumption MLR.4 (Zero conditional mean)
• In a multiple regression model, the zero conditional mean assumption is much

more likely to hold because fewer things end up in the error.
• Example: Average test scores
• Conclusion: All estimated coefficients will be biased
• Example: Omitting ability in a wage equation • Standard assumptions for the multiple regression model (cont.)
• Assumption MLR.5 (Homoskedasticity)
• Example: Wage equation
• Short hand notation

• When is there no omitted variable bias?
• If the omitted variable is irrelevant or uncorrelated
• Omitted variable bias: more general cases • Theorem 3.2 (Sampling variances of the OLS slope estimators)
• Under assumptions MLR.1 – MLR.5:
• No general statements possible about direction of bias

• Analysis as in simple case if one regressor uncoreelated with others
• Example: Omitting ability in a wage equation
• Components of OLS Variances: • An example for multicollinearity
• 1) The error variance

• A high error variance increases the sampling variance because there is more
“noise” in the equation.
• A large error variance doesn‘t necessarily make estimates imprecise.
• The error variance does not decrease with sample size. • The different expenditure categories will be strongly correlated because if a school has a lot of
resources it will spend a lot on everything.
• 2) The total sample variation in the explanatory variable • It will be hard to estimate the differential effects of different expenditure categories because all
• More sample variation leads to more precise estimates. expenditures are either high or low. For precise estimates of the differential effects, one would
• Total sample variation automatically increases with the sample size. need information about situations where expenditure categories change differentially.
• Increasing the sample size is thus a way to get more precise estimates.
• As a consequence, sampling variance of the estimated effects will be large.
• Components of OLS Variances (contd.) • Discussion of the multicollinearity problem
• 3) Linear relationships among the independent variables • In the above example, it would probably be better to lump all expen-
• Regress xj on all other independent variables (including constant) diture categories together because effects cannot be disentangled.
• The R-squared of this regression will be the higher when xj can be better • In other cases, dropping some independent variables may reduce
explained by the other independent variables. multicollinearity (but this may lead to omitted variable bias).
• The sampling variance of the slope estimator for xj will be higher when xj can be
better explained by the other independent variables.
• Under perfect multicollinearity, the variance of the slope estimator will

approach infinity.
• Only the sampling variance of the variables involved in • Variances in misspecified models (cont.)
multicollinearity will be inflated; the estimates of other effects may
be very precise.
• Note that multicollinearity is not a violation of MLR.3 in the strict
sense.
• Multicollinearity may be detected through “variance inflation factors.” • Case 1:
• Case 2:
• Variances in misspecified models • Estimating the error variance
• The choice of whether to include a particular variable in a regression can be
made by analyzing the tradeoff between bias and variance.
• An unbiased estimate of the error variance can be obtained by substracting the number of
estimated regression coefficients from the number of observations. The number of observations
minus the number of estimated parameters is also called the degrees of freedom. The n
estimated squared residuals in the sum are not completely independent but related through the
k+1 equations that define the first order conditions of the minimization problem.
• Theorem 3.3 (Unbiased estimator of the error variance)

• It might be the case that the likely omitted variable bias in the
misspecified model 2 is overcompensated by a smaller variance.
• Estimation of the sampling variances of the OLS estimators • Theorem 3.4 (Gauss-Markov Theorem)
• Under assumptions MLR.1 - MLR.5, the OLS estimators are the best
linear unbiased estimators (BLUEs) of the regression coefficients, i.e.
• Note that these formulas are only valid under assumptions MLR.1- • OLS is only the best estimator if MLR.1 – MLR.5 hold; if there is
MLR.5 (in particular, there has to be homoskedasticity) heteroskedasticity for example, there are better estimators.
• Efficiency of OLS: The Gauss-Markov Theorem • Several Scenarios for Applying Multiple Regression
• Under assumptions MLR.1 - MLR.5, OLS is unbiased • Prediction
• However, under these assumptions there may be many other estimators that • The best prediction of y will be its conditional expectation
are unbiased.
• Which one is the unbiased estimator with the smallest variance?
• In order to answer this question one usually limits oneself to linear estimators,
i.e. estimators linear in the dependent variable. • Efficient markets
• Efficient markets theory states that a single variable acts as a sufficient statistic
for predicting y . Once we know this sufficient statistic, then additional
information is not useful in predicting y.

• Several Scenarios for Applying Multiple Regression (contd.)
• Measuring the tradeoff between two variables
• Consider regressing salary on pension compensation and other controls
• Testing for ceteris paribus group differences

• Differences in outcomes between groups can be evaluated with dummy variables

• Several Scenarios for Applying Multiple Regression (contd.)
• Potential outcomes, treatment effects, and policy analysis
• With multiple regression, we can get closer to random assignment by
conditioning on observables.
• Inclusion of the x variables allows us to control for any reasons why

there may not be random assignment.
• For example, if y is earnings and w is participation in a job training program,
then the variables in x would include all of those variables that are likely to be
related to both earnings and participation in job training.
Multiple Regression Analysis: Inference (2 of 37)

• Assumption MLR.6 (Normality of error terms)
Chapter 4
Multiple Regression Analysis:
Inference
Multiple Regression Analysis: Inference (1 of 37) Multiple Regression Analysis: Inference (3 of 37)
• Statistical inference in the regression model • Discussion of the normality assumption
• Hypothesis tests about population parameters • The error term is the sum of “many” different unobserved factors.
• Construction of confidence intervals • Sums of independent factors are normally distributed (CLT).
• Problems:
• Sampling distributions of the OLS estimators • How many different factors? Number large enough?
• The OLS estimators are random variables • Possibly very heterogenuous distributions of individual factors
• We already know their expected values and their variances • How independent are the different factors?
• However, for hypothesis tests we need to know their distribution
• In order to derive their distribution we need additional assumptions • The normality of the error term is an empirical question.
• Assumption about distribution of errors: normal distribution • At least, the error distribution should be “close” to normal.
• In many cases, normality is questionable or impossible by definition.
• Discussion of the normality assumption (cont.) • Testing hypotheses about a single population parameter
• Examples where normality cannot hold: • Theorem 4.2 (t-distribution for the standardized estimators)
• Wages (nonnegative; also: minimum wage) • Under assumptions MLR.1 – MLR.6
• Number of arrests (takes on a small number of integer values)
• Unemployment (indicator variable, takes on only 1 or 0)
• In some cases, normality can be achieved through transformations

of the dependent variable (e.g. use log(wage) instead of wage).
• Under normality, OLS is the best (even nonlinear) unbiased estimator • Null hypothesis (for more general hypotheses, see below)
• Important: For the purposes of statistical inference, the assumption of
normality can be replaced by a large sample size.
• Terminology • t-statistic (or t-ratio)
• Theorem 4.1 (Normal sampling distributions)

• Under assumptions MLR.1 – MLR.6:
• Distribution of the t-statistic if the null hypothesis is true
• Goal: Define a rejection rule so that, if it is true, H0 is rejected only

with a small probability (= significance level, e.g. 5%)
• Testing against one-sided alternatives (greater than zero) • Example: Wage equation (cont.)
• Reject the null hypothesis in favour of the
alternative hypothesis if the estimated
coefficient is “too large” (i.e. larger than a
critical value).
• Construct the critical value so that, if the null

hypothesis is true, it is rejected in, for example,
5% of the cases.
• In the given example, this is the point of the t-

distribution with 28 degrees of freedom that is
exceeded in 5% of the cases.
• The effect of experience on hourly wage is statistically greater than
• Reject if t-statistic is greater than 1.701 zero at the 5% (and even at the 1%) significance level.
• Example: Wage equation • Testing against one-sided alternatives (less than zero)
• Test whether, after controlling for education and tenure, higher work
experience leads to higher hourly wages. alternative hypothesis if the estimated
coefficient is “too small” (i.e. smaller than a
critical value).

5% of the cases.
• In the given example, this is the point of the t-

distribution with 18 degrees of freedom so that
5% of the cases are below the point.
• Reject if t-statistic is less than -1.734
• Example: Student performance and school size • Example: Student performance and school size (cont.)
• Test whether smaller school size leads to better student performance • Alternative specification of functional form:
• Example: Student performance and school size (cont.) • Example: Student performance and school size (cont.)
• The hypothesis that there is no effect of school size on student performance

can be rejected in favor of the hypothesis that the effect is negative.
• How large is the effect?
• One cannot reject the hypothesis that there is no effect of school size
on student performance (not even for a lax significance level of 15%).
• Testing against two-sided alternatives • “Statistically significant” variables in a regression
• If a regression coefficient is different from zero in a two-sided test, the
alternative hypothesis if the absolute value of corresponding variable is said to be “statistically significant”.
the estimated coefficient is too large. • If the number of degrees of freedom is large enough so that the normal
approximation applies, the following rules of thumb apply:
5% of the cases.
• In the given example, these are the points of

the t-distribution so that 5% of the cases lie in
the two tails.
• Reject if absolute value of t-statistic is less than

-2.06 or greater than 2.06
• Example: Determinants of college GPA • Testing more general hypotheses about a regression coefficient
• Null hypothesis
• t-statistic
• The test works exactly as before, except that the hypothesized value is
substracted from the estimate when forming the statistic.
• Example: Campus crime and enrollment • How the p-value is computed (here: two-sided test)?
• An interesting hypothesis is whether crime increases by one percent if • The p-value is the significance level at which one is
enrollment is increased by one percent. indifferent between rejecting and not rejecting the null
hypothesis.
• In the two-sided case, the p-value is thus the

probability that the t-distributed variable takes on a
larger absolute value than the realized value of the test
statistic, e.g.
P(|t|>1.85) = 2*(0.0359) = .0718
• From this, it is clear that a null hypothesis is rejected if

and only if the corresponding p-value is smaller than
the significance level.
• For example, for a significance level of 5% the t-statistic

would not lie in the rejection region.
• Computing p-values for t-tests • Guidelines for discussing economic and statistical significance
• If the significance level is made smaller and smaller, there will be a point where • If a variable is statistically significant, discuss the magnitude of the coefficient to
the null hypothesis cannot be rejected anymore. get an idea of its economic or practical importance.
• The reason is that, by lowering the significance level, one wants to avoid more • The fact that a coefficient is statistically significant does not necessarily mean it
and more to make the error of rejecting a correct H0. is economically or practically significant!
• The smallest significance level at which the null hypothesis is still rejected, is • If a variable is statistically and economically important but has the “wrong”
called the p-value of the hypothesis test. sign, the regression model might be misspecified.
• A small p-value is evidence against the null hypothesis because one would • If a variable is statistically insignificant at the usual levels (10%, 5%, or 1%), one
reject the null hypothesis even at small significance levels. may think of dropping it from the regression.
• A large p-value is evidence in favor of the null hypothesis. • If the sample size is small, effects might be imprecisely estimated so that the
• P-values are more informative than tests at fixed significance levels. case for dropping insignificant variables is less strong.
• Confidence intervals • Example: Model of firms‘ R&D expenditures
• Simple manipulation of the result in Theorem 4.2 implies that
• Interpretation of the confidence interval

• The bounds of the interval are random.
• In repeated samples, the interval that is constructed in the above way will cover
the population regression coefficient in 95% of the cases.
• Confidence intervals for typical confidence levels • Testing hypotheses about a linear combination of the parameters
• Example: Return to education at two-year vs. at four-year colleges
• Relationship between confidence intervals and hypotheses tests • A possible test statistic would be:
• The standard error of the difference in parameters is impossible to • Testing multiple linear restrictions: The F-test
with standard regression output • Testing exclusion restrictions
• An alternative method is to make a substitution in variables.
• Estimation results • Estimation of the unrestricted model
• This method works always for single linear hypotheses
• Estimation of the restricted model • Test decision in example
• Test statistic: • Discussion

• The three variables are “jointly significant”
• They were not significant when tested individually
• The likely reason is multicollinearity between them
• Rejection rule • Test of overall significance of a regression
• A F-distributed variable only takes
on positive values. This corresponds
to the fact that the sum of squared
residuals can only increase if one
moves from H1 to H0.
• Choose the critical value so that we

incorrectly reject the null hypothesis
in, for example, only 5% of the • Discussion
cases. • The three variables are “jointly significant”
• They were not significant when tested individually
• The likely reason is multicollinearity between them

• Testing general linear restrictions with the F-test
• Example: Test whether house price assessments are rational

• Unrestricted regression
• Restricted regression
• Test statistic

The Nature of Econometrics and Economic Data: - Steps in Econometric Analysis

Uploaded by

Copyright:

Available Formats

You might also like

The Nature of Econometrics and Economic Data: - Steps in Econometric Analysis

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Nature of Econometrics and Economic Data: - Steps in Econometric Analysis

Uploaded by

Copyright:

Available Formats

9/21/2021 9/21/2021

The Nature of Econometrics and Economic Data (2 of 22)

Chapter 1 • Economic models

• Most of econometrics deals with the specification of the error u.

• Econometric methods depend on the nature of the data used.

• Cross-sectional observations are more or less independent. 3

to respond in surveys, or sampling may be characterized by clustering.

Introductory Econometrics: A Modern Approach (7e)

The Nature of Econometrics and Economic Data (21 of 22)

The Simple Regression Model (2 of 39)

• The simple linear regression model is rarely applicable in practice but

• Example: A simple wage equation

• Example: wage equation

• This means that the average value of the dependent variable

• Minimize the sum of the squared regression residuals

• Fitted regression • Algebraic properties of OLS regression

• Voting outcomes and campaign expenditures

• The log-log form postulates a constant elasticity model, whereas the

• This changes the interpretation of the regression coefficient:

• Assumption SLR.2 (Random sampling)

• Assumption SLR.5 (Homoskedasticity)

• Calculation of standard errors for regression coefficients

• This regression allows the mean value of y to differ depending on the

• Note that the statistical properties of OLS are no different when x is

• This can be written as:

Multiple Regression Analysis: Estimation (2 of 37)

Chapter 3 • Example: Wage equation

Multiple Regression Analysis:

• Per student spending is likely to be correlated with average family

• One can show that the estimated coefficient of an explanatory

• Algebraic properties of OLS regression

• Assumption MLR.2 (Random sampling)

• Interpretation: • Remarks on MLR.3

• Example for perfect collinearity: relationships between regressors

• In a multiple regression model, the zero conditional mean assumption is much

• Conclusion: All estimated coefficients will be biased

• Example: Wage equation

• Short hand notation

• No general statements possible about direction of bias

• 1) The error variance

• Under perfect multicollinearity, the variance of the slope estimator will

• Theorem 3.3 (Unbiased estimator of the error variance)

Multiple Regression Analysis: Estimation (36 of 37)

• Testing for ceteris paribus group differences

Introductory Econometrics: A Modern Approach (7e)

Multiple Regression Analysis: Estimation (37 of 37)

• Inclusion of the x variables allows us to control for any reasons why

Multiple Regression Analysis: Inference (2 of 37)

• In some cases, normality can be achieved through transformations

• Theorem 4.1 (Normal sampling distributions)

• Goal: Define a rejection rule so that, if it is true, H0 is rejected only

• Construct the critical value so that, if the null

• In the given example, this is the point of the t-

• Construct the critical value so that, if the null

• In the given example, this is the point of the t-

• Reject if t-statistic is less than -1.734

• The hypothesis that there is no effect of school size on student performance

• In the given example, these are the points of