PAULINE & MORRIS I ETS USING SAS/ETS SOFTWARE FOR .

PAULINE & MORRIS I ETSUSING SAS/ETS SOFTWARE FOR INSTRUCTION ANDRESEARCH AT PACE UNIVERSITYGerard Pauline and Walter MorrisPace Universit yThis paper will be divided into two sections. The first section will describe how data isextracted from the Citibase and Compustat II databases here at Pace University. Sectiontwo will explain how students use various SAS/ETS procedures to analyze the extracteddata.'SECTION 1Our original intention was to find a way to efficiently access, and more importantly, analyzedata from the Compustat II database. The only means of doing this, before the SAS/ETSproduct was acquired, was to go through a set of FORTRAN programs or a product knownas TROLL. Neither method proved very satisfying for our user population, particularlythose interested in doing research.The 1982 SAS/ETS Guide contained an example of a macro program that accessed oneofthe Compustat files. This was used as a basis to develop individual macros for the otherCompustat files (the files have different record layouts).The ETS product also contained a procedure to access data from Citibase, known asPROC CITIBASE. Although easy to use, it did not provide facilities to specify an exactperiodicity (i.e. 8202 to 8603) or to compact data (transform a time series from one formatto another), as did the TROLL product. The time series transformation could easily beaccomplished in the DATA step, but was confusing for inexperienced users.We decided to try to develop a program that would allow access to any of the 6 Compustatfiles, as well as to Citibase. This program would also have to incorporate exact periodicityspecification and data compacting for Citibase data. Above all, it had to be easy andstraightforward to use. The result of this is a macro program named ACCESS, whichaccomplishes all of the specified criteria.Data extraction is controlled by supplying control parameters to the macro. There are atotal of 11 macro control parameters, of which 4 are required regardless of which databaseis being accessed, and 6 of the 11 are common to both databases. The following is a listof the macro's control parameters:116

PAULINE & MORRIS I ETSDBspecifies the database, eith r CITIBASE or COMPUSTATTYPEdesignates the data type for Citibase (Q-quarterly,M-monthly, A-annual) or the Compustat file (INA-industrial annual, INO·industrial quarterly, RSC-research, OTC-over the counter, etc.)OUTPUTdataset to hold the extracted dataSELECTspecifies the variables for extractionBEGYRthe date/year to start data selectionENDYRthe date/year to end data selectionSELCRITspecifies a condition on which data will be selectedPRTYES/NO toggle, prints the extracted dataCOMPACTYES/NO toggle, specifies if the data is to be compactedCMspecifies the method of data compacting: monthly to annual format by summation, monthly to quarterly format by averaging, quarterly to annual formatby summation, quarterly to annual format by selecting the value of the 4thquarterCOMDATAspecifies the dataset to hold the compacted dataWhen extracting data for a specific period of time, the user may enter that period as a rangeby using both the BEGYR and ENDYR parameters together. As a start date to begin dataselection only BEGYR would be used. To select data up to a certain point only ENDYRwould be used. Not specifying a time period results in all available data being selected.To keep the process of accessing the different databases 'transparent' to the user, asmany of the control parameters as possible are made common to accessing bothdatabases.Regardless of which database is being accessed, the DB, TYPE, OUTPUT and SELECTparameters must be specified. The parameters PRT, COMPACT, CM and COMDATA areused only when accessing Citibase and the SELCRIT parameter is used only whenaccessing Compustat. The macro checks to see if the required control parameters havebeen supplied; if they haven't been, a message is written to the SASLOG.117

PAULINE & MORRIS I ETSThe macro utilizes 'KEYWORD' parameters that have the form KEYWORD XXX. Thisallows the various parameters to appear in any order. The parameters may appear on oneline or across several lines and are separated by commas; spacing is unimportant. Themacro may be invoked as many times as necessary in one program.The macro resides in a file called SASMAC, which also contains other macros. The macrois made available to the user's program by using an Include statement to call the SASMACfile into the user's program, as opposed to using the Autocall facility. This allows for easiermaintenance and access if job submission is through CMSBATCH.To increase execution speed and conserve disk space (the 3 files take up 12 megabytes),the Citibase data has been converted to SAS data format. The data is stored in 3 files;monthly, quarterly and annual data. The conversion to SAS data format has the addedadvantage of storing the variable labels with the data.The files are updated by a maintenance program that is run on the new tape. The programuses PROC CITIBASE to create the individualdatasets and the UPDATE facility to updatethe permanent disk files with the new data. A test program validates the updates.The six Compustat files currently reside in their original format (real binary), although in thefuture they may be converted to SAS data format. The reason for not yet doing so is thatSAS format and real binary format are similar except for the generational information thatSAS stores. Unlike Citibase, the Compustat updates completely replace the old files.The following are examples of using the ACCESS macro. The first example accessesCitibase twice, the first time to extract quarterly data, the second time to extract monthlydata. The second example is a sample of extracting data from Compustat. Note that theInclude statement to call the SASMAC file is not shown (it must be specified before usingthe macro).%ACCESS(DB citibase, TYPE q, OUTPUT xqd1, BEGYR 75q1,ENDYR 79q4, SELECT gnp gyd)%ACCESS(DB citibase, TYPE m, OUTPUT Xmd, SELECT fm1,BEGYR jan75, ENDYR dec79, COMPACT yes, CM M2,COMDATA xqd2)DATA COMBO;MERGE XQD1 XOD2; BY DATE;PROC REG DATA COMBO; MODEL FM1 GNP GYD;118

PAULINE & MORRIS I ETS%ACCESS(DB compustat, TYPE inq, OUTPUT amxdata,SELCRIT CNUM "025816", SELECT CONAME YEAR V22-V26 V14V19)PROC PRINT DATA AMXDATA;In the Citibase example, the first time the macro is invoked, data for the 1st quarter 1975through 4th quarter 1979 is selected for the variables GNP and GYD. The secondinvocation selects monthly data for the period January 1975 through December 1979 forthe variable FM1. Since FM1 is only available in monthly format, it is compacted toquarterly format for compatibility with the variables GNP and GYD. The two datasets,XQD1 (GNP ,GYD) and XOD2 (FM1) are then merged.In the second example, the Compustat Industrial Quarterly (TYPE INQ) is accessed fordata pertaining to American Express (CNUM "025816") for the 9 specified variables.In the near future, we hope to extend the macro to access our other large databases:International Financial Statistics (IFS), Panel Survey of Income Dynamics (PSI D) and theUniversity of Chicago's Center for Research in Security Prices (CRSP) files.Unfortunately, space constraints will not permit a discussion of the coding techniquesused in the Access macro or a listing of the code. For further information, interested partiesshould contact the Department of Academic Computing.SECTION 2\It goes without saying that economics is generally thought of as a difficult and sometimesa rather "dismal" science. Indeed, most students have a negative recollection of their first,which in all likelihood, turns out to be their last economics course taken in college. Whathave teaching economists done to rectify this situation; that is, have teachers been ableto breathe some new life into the dismal science?From a purely pedagogical perspective, even though no new theories have burst upon thescene, the profession has made major advances in promoting existing ideas through theextensive use of computer methodologies. In the areas of data base management as wellas statistical programming, and more importantly, their interactions, have helped movethe teaching of economics into a new and exciting environment. In the areas ofmacroeconomics, microeconomics, managerial economics, labor economics, international economics, and the obvious example of econometrics, the interfacing of largecomplex data bases with canned statistical programs has introduced a fresh teachingapproach into these subjects.119

PAULINE & MORRIS I ETSHere at Pace three major data bases have been linked with the SAS statistical programsproviding users with several unique opportunities to relate economic theory with economics reality (i.e., see ACCESS MACRO at the beginning of the paper). CITIBASE is aneconomic data set that collects approximately 5,000 economic time series from 194 7 topresent. In macroeconomics classes these data together with several SAS procedureshave been useful in helping student estimate and graph consumption functions, investment functions, and money demand functions (i.e., PROC REG). At the conclusion of thecourse, these three functions are linked together in a multiplier-accelerator-liquiditypreference setup, the equations are estimated, simulations are computed, and the resultsare graphed (i.e., PROC SYSNLIN, PROC SIMNLIN).Students are constantly reminded of the interplay of the economic theory with data behindthese equations, and are accorded the capability of playing their results off against variousalternative policies. For examp!e, the models are structured in such a manner as toanswer questions relating to differing approaches to fiscal and monetary policy and howthey impact on the relevant simulated variables. In short, students can place themselvesin the drivers seat with regards to economic policy making and observe the results ofvarious decisions made in these two areas.Pace has access to the longitudinal data base referred to as the Panel Survey of IncomeDynamics (PSID). The PSID is a national probability sample of 7,000 heads of households that includes data gathered from 1968-present. In each survey year approximately300 variables are collected on each household, some of the information is repeated for all16 years and some is collected only for a specific year. The data have been extremelyuseful in econometrics classes as well as classes in labor economics. In the labor classesintense theoretical interest centers around the notion of human capital theory.Since the PSID is linked to the SAS programs students again have a unique opportunityto test theory against practice. Here students typically gather data for one year and testthe relationships of earnings with schooling, training, race, sex, etc. (i.e., PROC FREO,PROC MEANS, PROC TABULATE, PROC REG). Because of the nature of longitudinaldata, the capability exists to test these relationships over an extended time span todetermine if, for example, the returns to a college education have shifted over time (anarea of particular importance to college students). In either case, at the conclusion of thecourse students have a hands on feel for the extremely abstract human capital theory.These are just two examples of how the economics department uses the computermethodologies in a classroom environment. It is clear that the Econometrics courses usea wide variety of statistical programs with these data bases including PROC FACTOR,PROC FORECAST, and PROC ARIMA, in addition to the ones previously described.However, space does not permit us to explore these possibilities. Admittedly, most of theSAS applications discussed here are in the areas of "number crunching", but with ourrecent acquisition of SAS GRAPH the department looks forward to exploring new horizonsin this critical area.120

The 1982 SAS/ETS Guide contained an example of a macro program that accessed one ofthe Compustat files. This was used as a basis to develop individual macros for the other Compustat files (the have different record layouts). The ETS product also contained a procedure to access data from Citibase, known as PROC CITIBASE.