Public Economics K.Peren Arin Contact Information Office Hours:After class! All communication in English please! 1
Introduction The year is 1030 B.C. For decades, Israeli tribes have been living without a central government. The Bible records that people have asked prophet Samuel to make us a king to judge us like all nations Samuel responds: The king shall reign over you, take your sons, and appoint them onto him, to be his horsemen And will take your daughters to be perfumers, and cooks. He will take your fields, wineyards, and give them to his servants Views on Government People s views on how the government should conduct its financial operations are heavily influenced by their political philosophies Organic View: Society is conceived as a natural organism. Each individual is a part of this organism, and community is stressed above the individual. Roots can be traced back to Plato, who argued that an activity of a citizen is desirable only if it leads to a just society. Mechanistic View: The individual, rather than the group is at center stage. There is limited scope for Government 2
Scope for Government for the Mechanistic View Even people who believe in mechanistic view do not deny that we need a government. Why? Are there cases where rational individual behavior leads to non-optimal solution? Size of Government 3
Understanding public policy changes Positive analysis because economists generally cannot perform controlled experiments with the economy, the effects of economic policy are difficult to determine. That is why most economists use econometrics (statistical analysis of economic data) to investigate the effects of government actions Why do we need empirical research? Empirical research helps us to choose between conflicting theories And sometimes, with the use of certain methods, it may lead/inspire theoretical research 4
Introduction Econometrics deals with Stochastic Relations. The simplest form of a stochastic relation takes the form: Y α + β + ε = X i i In this specification Y is called the dependent variable, X is called the independent (explanatory) variable, and and are the unknown regression parameters. The subscript i refers to the ith observation. The values of the variables X and Y are observable, but those for the stochastic disturbance term (which creates the differences in the relation between X and Y) are not. Types of Data Used in Econometric Studies Observations on X and Y can be over time, in which case we have time-series data, or they can be over a group of individuals, firms or countries, in which case we talk about having cross-section data. If we have time-series data for a number of agents (firms or countries) then we talk about having a panel data set. 5
Some tips for Empirical Research Think BIG! A panel study for 120 countries is always more valuable for a single-country study you do for Germany Most single-country studies (especially in macro/finance) are done for the U.S., given that most academic journals are U.S. based and data sources are trusted The only exception is topics that cannot be tested for U.S. case, like privatization, poverty, transition economies..etc More tips for empirical research Present summary statistics, except for time-series studies Never copy+ paste estimation output from a statistical software Remember that presentation and writing is as important as the use of solid econometric techniques (See Cochrane's tips on the internet) 6
The basic regression: OLS Reg Y X (OLS command in STATA) In the below STATA output, both coefficient estimates and corresponding t and p values are reported. Significant levels are determined according to p-values salevad Coef. Std. Err. t P> t bsatisda 1.396239 0.550956 2.53 0.02 pyear 28152.88 14141.8 1.99 0.06 popgro 1849.204 2160.105 0.86 0.402 - _constant 5.60E+07 2.82E+07-1.99 0.061 Choosing the functional form The coefficient of the dependent variable shows the change in dependent variable when there is a one-unit increase in the independent variable Log-Linear Models: Reg LnY LnX One attractive feature of the Log-Linear Model is that the slope coefficient of X measures the percentage change in dependent variable when there is a 1% increase in the independent variable. Therefore, log-linear models measure the elasticity of Y with respect to X. 7
Econometric problems with crosssection data Co-linearity Heteroskedasticity Endogeneity Co-Linearity It is the consequence of experimental data. The classical symptoms include: Wrong Signs Hypersensitivity to data and model changes Important variables being insignificant With Collinearity, OLS is still BLUE, but the confidence intervals will be wider, therefore t-statistics will be smaller, making important variables insignificant. If collinearity affects the variable of interest, but the estimate is statistically significant, we still cannot trust the t-statistic. 8
Dealing with Co-Linearity Example: Both variables have the right signs but they are insignificant. You believe that both are important in explaining the Y. Y = 0 1 1 2 X 2 β + β X + β + ε Solutions Strategy 1: Use and in separate specifications, get the restricted least squares estimator with the smaller variance, but larger bias. Strategy 2: Create a new variable that combines the two variables, with smaller variance and larger bias. 9
Heteroskedasticity This problem occurs when the diagonal elements of the covariance matrix (variances of variables) are not equal. The heteroskedastic error assumption is a reasonable one for cross-section data. In that case, the OLS estimator is unbiased consistent asymptotically normal BUT NOT EFFICIENT (you will have insignificant terms because of the large variance) Diagnosing Heteroskedasticity It is a rule, rather than exception with crosssection data. 10
Dealing with Heteroskedasticity You can correct for heteroskedasticity with White (1980) heteroskedasticity consistent estimator. (It is the most cited paper in both economics and finance!!!) STATA COMMAND: reg Y X, robust Stochastic Regressors (Endogeneity) Endogeneity, or mutual causality is very common in applied public economics. If X causes Y, and Y causes X, then the independent variables will be correlated with the error terms and so, OLS fails Many IO papers published in 80s are not cited anymore as they did not control for this problem of mutual causality 11
Solving endogeneity with crosssection data INSTRUMENTAL VARIABLES (IV) REGRESSION AND TWO-STAGE LEAST SQUARES (2SLS) You have to find some other variable Z (in most cases the lag of X) that is not correlated with the error term, but highly correlated with X to use instrumental variables estimation or the two stage least squares. Regress X on Z Regress Y on estimated coefficients from the first regression. STATA COMMAND: ivreg Y (X=Z) ACHTUNG: R-square is not a good measure of fit with 2SLS. How to Choose Instruments? Check previous literature Make correlation tests STATA COMMAND Pwcorr x y, sig If nothing works, use the lag value of X! 12
..so is there an optimal size of the government? Positive Analyses suggest that optimal size of the government (G/Y) is between 36 and 42 percent. Recent studies showed the importance of composition of government expenditures. Often, the composition is more important than the size. Figure 1.5: Composition of state and local taxes (1965 and 2002) 100% 90% 80% Other Individual tax more important 70% 60% 50% 40% 30% 20% Grants from federal government Corporation tax Inidividual income tax Sales tax Property tax Property tax less important 10% 0% 1965 2002 Source: Economic Report of the President [2006, p. 383]. 13
Normative Analysis We do want to determine not only the effects of government policies, but whether or not they produce results that are good in some sense. This can be done through the use of normative analysis. Welfare Economics Welfare Economics branch of economic theory concerned with the social desirability of alternative economic states 14
r Edgeworth Box y Eve 0 Fig leaves per year u v w 0 Adam x Apples per year s Edgeworth Box Indifference curves in Edgeworth Box r E 1 Eve 0 E 2 Fig leaves per year E 3 A 2 A 3 0 Adam A 1 Apples per year Edgeworth Box s 15
Making Adam better off without Eve becoming worse off Eve r 0 Fig leaves per year E g g h p A Pareto Efficient Allocation A p A h A g 0 Adam Apples per year s Edgeworth Box Making Eve better off without Adam becoming worse off Eve r 0 E g g Fig leaves per year E p1 p 1 p A Pareto Efficient Allocation A g 0 Adam Apples per year s Edgeworth Box 16
Making both Adam and Eve better off r 0 Eve Fig leaves per year E g E p2 g p 1 p 2 p Pareto efficient Pareto improvement A p2 A g 0 Adam Apples per year s Edgeworth Box Starting from a different initial point r 0 Eve Fig leaves per year E g E p2 g p 1 p 2 p k p 3 p 4 A p2 A g 0 Adam Apples per year s Edgeworth Box 17
r The Contract Curve Eve 0 Fig leaves per year E g E p2 g p 1 p 2 p p 3 p 4 The contract curve A p2 A g 0 Adam Apples per year s Edgeworth Box Pareto Efficiency in Consumption MRS af Adam Eve = MRS af 18
Production Possibilities Curve Fig leaves per year w C Slope = marginal rate of transformation y 0 x z C Apples per year Marginal Rate of Transformation MRT af = Marginal rate of transformation of apples for fig leaves MRT af = MC a /MC f 19
Efficiency Conditions with Variable Production MRT af = MRS af Adam MC a /MC f = MRS af = MRS af Adam Eve Eve = MRS af Market Power monopoly Market Failure Nonexistence of Markets asymmetric information externality public good 20
Characteristics of Goods Excludable v Nonexcludable Excludable preventing anyone from consuming the good is relatively easy Nonexcludable preventing anyone from consuming the good is either very expensive or impossible Rival v Nonrival Rival once provided, the additional resource cost of another person consuming the good is positive Nonrival once provided, the additional resource cost of another person consuming the good is zero Types of Goods RIVAL YES NO EXCLUDABLE YES PRIVATE GOODS NATURAL MONOPOLY NO COMMON PUBLIC RESOURCES GOODS 21
Some Other Public Goods Basic research Programs to fight poverty Uncongested nontoll roads Fireworks display Efficient Provision of Private Goods Price $11 Adam (D fa ) 5 Eve (D fa ) 1 Market (D A+E f ) 6 $9 7 3 10 $7 9 5 14 $5 11 7 18 $3 13 9 22 $1 15 11 26 22
Laboratory Experiments and Free- Riding How a typical experiment works Typical results People contribute about 50% of resources to provision of public good Contributions fall the more often the game is repeated Cooperation fostered by prior communication Contribution rates decline when opportunity cost of giving goes up The Privatization Debate Privatization taking services supplied by government and turning them over to the private sector Public v Private Provision: What is the right mix? 23
Externalities Externality An activity on one entity that affects the welfare of another entity in a way that is outside the market mechanism The Nature of Externalities-Graphical Analysis $ MSC = MPC + MD MPC d h g c MD 0 Socially efficient output Q* b a f e Q 1 Actual output MB Q per year 24
What Pollutants Do Harm? Empirical Evidence: What is the Effect of Pollution on Health? What Activities Produce Pollutants? What is the Value of the Damage Done? Empirical Evidence: The Effect of Air Pollution on Housing Values The Coase Theorem Coase Theorem Provided that transaction casts are negligible, an efficient solution to an externality problem is achieved as long as someone is assigned property rights, independent of who is assigned those rights Assumptions necessary for Coase Theorem to work The costs to the parties of bargaining are low The owners of resources can identify the source of damages to their property and legally prevent damages 25
Public Responses to Externalities - Taxes $ MSC = MPC + MD (MPC + cd) Pigouvian tax revenues MPC i j d c MD MB 0 Q* Q 1 Q per year 26