《计量经济学导论》cha.ppt

《《计量经济学导论》cha.ppt》由会员分享，可在线阅读，更多相关《《计量经济学导论》cha.ppt（24页珍藏版）》请在三一办公上搜索。

1、Chapter 19,Carrying Out an Empirical Project,Wooldridge:Introductory Econometrics:A Modern Approach,5e,Goal of this chapterLearn how to complete a term project/write a term paperPosing a questionKnowing precisely what question you want to answer is essentialYou can only collect your data if you exac

2、tly know your questionYou can only know whether you can complete your project in the allotted time if you know whether the necessary data is availableYou can only know if your research question is of interest to someone if you can precisely state it and discuss it with your class mates/instructor,Ca

3、rrying out an Empirical Project,Finding interesting research questionsChoose the area of economics/social sciences you are interested inExamples for typical research questionsLabor Economics:Explaining wage differentialsPublic Economics:Effect of taxes on economic activityEducation Economics:Effect

4、of spending on school performanceMacroeconomics:Effect of investment on GNP growthLook for published papers on the chosen topic using tools such as EconLit,Google Scholar,the Journal of Economic Literature(JEL)etc.,Carrying out an Empirical Project,Your research project should add something newAdd a

5、 new variable whose influence has not been studied beforeExpand economic questions to include factors from other sciencesStudy an existing question for more recent data(may be boring)Use a new data set or study a question for a different countryTry out new/alternative methods to study an old questio

6、nFind a completely new question(hard but possible)It helps if your research question is policy relevant or of local interest,Carrying out an Empirical Project,Literature reviewA literature review is important to place your paper into contextUse online search services to systematically search for lit

7、eratureWhen searching,think of related topics that may also be relevantA literature review can be part of introduction or a separate sectionData collectionMost questions can be addressed using alternative types of data(pure cross-sections,repeated cross-sections,time series,panels),Carrying out an E

8、mpirical Project,Deciding on the appropriate data setMany questions can in principle be studied using a single cross-sectionBut for a reasonable ceteris paribus analysis one needs enough controlsPanel data provides more possibilities for convincing ceteris paribus analyses as one can control for tim

9、e-invariant unobserved effectsExamples for panel data sets:PSID(individuals),Compustat(firms)Panel data for cities,counties,states etc.are often publicly availableData sets are often available online,in journal archives,or from authors,Carrying out an Empirical Project,Entering and storing your data

10、Data formats:1)printed,2)ASCII,3)spreadsheet,4)software specificImportant identifiers:1)observational unit,2)time periodTime series must be ordered according to time periodPanel data are conveniently ordered as blocks of individual dataIt is always important to correctly identify and handle missing

11、valuesNonnummerical data also have to be handled with great careSoftware specific formats often provide good ways of documentation,Carrying out an Empirical Project,Inspecting,cleaning,and summarizing your dataIt is extremely important to become familiar with your data setEven data sets that were us

12、ed before may contain problems/errorsLook at individual entries/try to understand the structure of your dataUnderstand how missing values are coded;if they are coded as“999“or“-1“,this can be extremely dangerous for your analysisIt is better to use nonnummerical values for missing valuesUnderstand t

13、he units of measurement of your variables,Carrying out an Empirical Project,Inspecting,cleaning,and summarizing your dataKnow whether your data is real/nominal,seasonally adjusted/unadjustedCheck if means,std.dev.,mins,and maxs of your data are plausibleClean your data of implausible values and obvi

14、ous coding errorsWhen making data transformations(differencing,growth rates)make sure your data is correctly ordered and no wrong operations resultFor example,in a panel data set,be aware that the first observation of each cross-sectional unit has no predecessor,Carrying out an Empirical Project,Eco

15、nometric AnalysisGiven your research question and the data available,you have to decide on the appropriate econometric methods to useSome general guidelinesOLS is still the most widely used method and often appropriateMake sure the key assumptions are satisfied in your modelAlways check for possible

16、 problems of omitted variables,self-selection,measurement error,and simultaneity,Carrying out an Empirical Project,Some general guidelinesCarefully choose functional form specifications(logs,squares etc.)Beginners mistake:do not include variables that are listed as numerical values but have no quant

17、itative meaning(e.g.,3-digit occupations)Transform such variables to dummy variables representing categoriesHandle ordinal regressors in a similar way(e.g.,job satisfaction)For ordinal dependent variables,there are ordered logit/probit modelsOne can also reduce ordered variables to binary variables,

18、Carrying out an Empirical Project,Some general guidelinesThink of secondary complications such as heteroscedasticitySpecific problems in time series regressions:1)levels vs.differences,2)trends and seasonality,3)unit roots and cointegrationCarry out misspecification tests and think about possible bi

19、asesSensitivity analysis:look at variations of your specification/methodHopefully,results do not change in a substantial wayAre there problems with outliers/influential observations?,Carrying out an Empirical Project,Specific aspects to think of when using panel dataKey assumptionsRandom effects:reg

20、ressors unrelated to individual specific effectsFixed effects:regressors related to individual specific effectsThe fixed effects assumption is often more convincingContemporaneous exogeneity:idiosyncratic errors are uncorrelated with the explanatory variables of the same time periodStrict exogeneity

21、:idiosyncratic errors are uncorrelated with the explanatory variables of all time periods(often problematic),Carrying out an Empirical Project,Specific aspects to think of when using panel dataMethods for panel dataPooled OLS:random effects assumption,serial correlation of error terms,needs only con

22、temporaneous exogeneityRandom effects estimation:random effects assumption,more efficient than pooled OLS,needs strict exogeneityFixed effects estimation:fixed effects assumption,problem with time invariant regressors,needs strict exogeneityFirst differencing:similar to fixed effects,good for longer

23、 time series,Carrying out an Empirical Project,Data mining/specification searchesThe process of looking for the best model is called specification searchOften,one starts with a general model and drops insignificant variablesIf the specification search entails many steps,this is problematicOur assump

24、tions actually require that the model is only estimated onceIf one sequentially estimates a number of models on the same data,the resulting test statistics and p-values cannot be interpreted anymoreThis(difficult)problem is often ignored in practiceOne should keep the number of specification steps t

25、o a minimum,Carrying out an Empirical Project,Writing an empirical paperA succesful empirical paper combines a careful,convincing data analysis with good explanations and a clear expositionIntroductionState basic objectives and explain why the topic is importantLiterature review:What has been done?H

26、ow do you add to this?Grab the readers attention by presenting simple statistics,paradoxical evidence,topical examples,or challenges to common wisdomOne may give a short summary of results in the introduction,Carrying out an Empirical Project,Conceptual(or theoretical)frameworkDescription of general

27、 approach to answering your research questionYou may delevop/use a formal economic model for thisFor example,setting up a utility maximization model of criminal activity clarifies the factors that matter for explaining criminal activity However,often common economic sense suffices to discuss the mai

28、n mechanisms and control variables that have to be taken into accountAs one is in most cases interested in answering a causal question,a convincing discussion of what variables to control for is essential,Carrying out an Empirical Project,Econometric models and estimation methodsSpecify the populati

29、on model you have in mindExample:Effects of alcohol consumption on college GPAExample:Time series model of city-level car theftsExplain your functional form choices,Carrying out an Empirical Project,Econometric models and estimation methodsAfter specifying a population model,discuss estimation metho

30、dsDescribe how you measure the variables in your population modelWhen using OLS:Discuss why exogeneity assumptions hold,and how you deal with heteroscedasticity,serial correlation and the likeWhen using IV/2SLS:Explain why your instrumental variables fulfill the assumptions:1)exclusion,2)exogeneity,

31、3)partial correlationWhen using panel methods:Explain what the unobserved individual specific effects stand for,and how they are removed/accounted for,Carrying out an Empirical Project,DataCarefully describe the data used in your empirical analysisName the sources of your data and how they can be ob

32、tainedTime series data and short data sets may be listed in the appendixIf your data is self-collected,include a copy of the questionnaireDiscuss the units of measurement of the variables of interestPresent summary statistics for the variables used in the analysis For trending variables,growth rates

33、 or graphs are more appropriateAlways state how many observations you use for different estimations,Carrying out an Empirical Project,ResultsPresent estimated equations,or,if there are too many,present tablesAlways include things like R-squared and the number of observationsAre your estimated coeffi

34、cients statistically significant?Are they economically significant?What is their magnitude?If coefficients do not have the expected signs,this may indicate there is a specification problem,for example,omitted variablesRelate differences between the results from different methods to the differences i

35、n the assumptions underlying these methods,Carrying out an Empirical Project,ConclusionSummarize main results and conclusions from themDiscuss caveats to the conclusions drawnSuggest directions for further researchStyle hintsChoose a title that is exciting and reflects the papers topicPapers should

36、be typed and double-spacedNumber equations,graphs and tablesRefer to papers by author and date,for example,White(1980),Carrying out an Empirical Project,Style hintsWhen you introduce an equation,describe important variablesIn order to focus on a particular variable you may write something likePresen

37、ting results in equation form:,Carrying out an Empirical Project,Shorthand for several other explanatory variables,State near the first equation thatstandard errors are in parentheses,Style hints,Carrying out an Empirical Project,Reporting results in tabular form:Clearly indicate dependent and independent variables.Limit the number of digits reported after the decimal point.You may also think of rescaling your variables so that coefficients are not too large or too small.,