Expenditures and Receipts in State and Local Government Finances ...

1 downloads 889 Views 774KB Size Report
Expenditures and receipts in state and local government finances: Reply*. MICHAEL L. MARLOW. U.S. Department of Treasury, Washington, DC 20220.
Expenditures and and local local government state and and receipts receiptsin state Expenditures government finances: finances: Reply* Reply* MICHAEL L. MARLOW MARLOW MICHAELL.

u.s. Department Department of of Treasury, 20220 U.S. Treasury,Washington, Washington,DC 20220 NEELA MANAGE NEELA MANAGE Department of Economics, Florida Florida Atlantic Atlantic University, Boca Raton, Raton, FL FL 33431 33431 University,Boca Department of Economics,

Chowdhury between alternativetechnique for causality an alternative test for Chowdhury(1988) (1988)adopts techniqueto test causalitybetween adoptsan andlocal local governments governments and claims that that he he obobexpenditures and stateand and claims andreceipts receiptsof state expenditures in tains that are in our 1987 paper (Marlow differentfrom from those those reported our tains results are different 1987 resultsthat reported paper(Marlow and Manage, Manage, 1987). disagree with In this this reply, reply, we we with many his criticisms. and We disagree criticisms.In 1987).We manyof his argue that results and policy implications implications and his our results and policy that he he misinterprets and that that his argue misinterpretsour results are really not much much different fact, sometimes his results results differentfrom from ours; results are really sometimeshis ours;in fact, our hypothesis. provide stronger support of our providestrongersupport hypothesis.

Optimal lag Optimal lag length length Chowdhury criticizes procedure which utilizes symmetric lags our estimation estimationprocedure criticizesour whichutilizes Chowdhury symmetriclags from periods as points. One, from two two to five five periods as arbitrary. We make make two two points. we did did not not One, we arbitrary.We arbitrarily choose a single lag structure but analyzed the causal patterns over choose a structure but the causal over arbitrarily single lag analyzed patterns long range years). In In the the univariate univariate and and multivariate multivariate a relatively five years). relativelylong range (up (up to five equations estimated by Chowdhury, lag length length which which minimizes minimizes the the Final Final estimatedby the lag equations Chowdhury,the Prediction by varying PredictionError is selected Error(FPE) selectedby the lag from 1, 1, ... ..., , nn (FPE) is varyingthe lag length length from in Chowdhury's paper). However, (where the selected selectedlag However,if the (wherenn == 5 in Chowdhury'spaper). lag length lengthequals equals or is the search be nn or is very the literature close to n, literaturesuggests that the searchof lag n, the very close suggeststhat lag length lengthbe in order extended beyond nn in greater than extendedbeyond orderto determine determineif the the optimal is greater than n. optimallag lag is In the In the regressions as the the dependent with local revenuesas local government variable, regressionswith governmentrevenues dependentvariable, Chowdhury selects a four period lag. However, lag lengths peria four selects five periChowdhury periodlag. However,lag lengthsexceeding exceedingfive be lowered by includods ods are are'arbitrarily' not examined FPEcan examinedto check checkif the the FPE can be loweredby includ'arbitrarily'not FPE criterion ing lags greater than five. While Chowdhury claims to use the than five. While FPE claims use the criterion ing lags greater Chowdhury avoidarbitrariness, to avoid his analysis shouldalso also include includethese thesechecks checksfor for optimal arbitrariness,his analysisshould optimal Theviews ** The represent those those of the viewsexpressed arethose thoseof the the authors authorsalone aloneand anddo do not not necessarily the expressedare necessarilyrepresent U.S. U.S. Department Departmentof Treasury. Treasury.

lag justifies his by using provided by by his choice He justifies choiceof nn == 5 by the rationale rationaleprovided lag length. usingthe length.He Blackley (1986). We argue that our choice of restricting the lag lengths up We that our choice the Blackley(1986). argue restricting lag lengthsup to based on similar five periods is also based similarreasoning is not five periods is also and is not any more arbitrary any more reasoningand arbitrary ... the provided than Ram (1988) the experimental than his method. Ram evidenceprovided his method. (1988)argues' argues'... experimentalevidence by Guilkey justify the and Salemi Salemi (1982) seems important the use use of by Guilkeyand (1982) seems importantenough enough to justify in preassigned lag lengths in the vicinity of what gave good results in their exin the what results their exvicinity gave good preassignedlag lengths periments. ' periments.' Two, may result result withrespect thatthey Two, with theymay symmetriclags, lags, Chowdhury Chowdhuryargues respectto symmetric arguesthat in or underspecification of the lag length and thereby in overspecification the and or result lag length therebyresult overspecification underspecification or biased estimates. In order to determine whether this in either inefficient in eitherinefficientor biasedestimates.In order determinewhetherthis is is true, by Chowdhury. our results resultswith withthose those reported we attempted true, we Chowdhury. compareour attemptedto compare reportedby However, between the in three exist aa discrepancy the lag three thereappears However,there discrepancybetween lag lengths lengthsin appearsto exist of the Table4), and the the reader is left wonder readeris left to wonder the four four regressions 4), and (Chowdhury,Table regressions(Chowdhury, in Table For if this based on is based on the the lag forms reported Table2 or or Table Table4. For this comment commentis lag forms reportedin example, it in his with state as the the dependent state expenditure his regression variable,it dependentvariable, expenditureas example,in regressionwith is (Le., the RHS of the equation includes a one period the includes a one the RHS is not not clear clearif a(I,3) a (1,3) lag period lag (i.e., equation lag valuesfor for receipts) or aa (1,4) structure for expenditures andthree threelagged receipts)or (1,4)lag lag structure laggedvalues lag for expendituresand and is selected. between state unidirectionalcausal causalrelation relationbetween staterevenues revenuesand He finds findsaa unidirectional selected.He is the same result we obtained when we used (3,3) and (4,4) expenditures which used and same we obtained when we which is the result (3,3) (4,4) expenditures lag bivariate equations we estimated, in seven sevenof the the eight In fact, structures.In estimated, fact, in eight bivariate equationswe lag structures. we use that our our use as Chowdhury. we suggest resultas we obtain obtainthe the same sameresult Chowdhury.Accordingly, Accordingly,we suggestthat of symmetrical lags does not pose serious problems. not serious does problems. pose symmetricallags

Serial correlation Serialcorrelation

variablesinto into their their Contrary the variables we transformed transformedthe criticism,we Chowdhury'scriticism, Contraryto Chowdhury's for serial serialcorrelation correlation natural used to check we used checkfor testswe naturallogarithms. The diagnostic diagnostictests logarithms.The estifromthe the esticonsisted obtainedfrom on the the residuals residualsobtained consistedof estimating estimatingautoregressions autoregressionson otherresearchresearchmated been used used by by other havebeen Similardiagnostic checkshave matedregressions. diagnosticchecks regressions.Similar While ers andMarlow, Williamset et al., 1976; andManage ers(Mehra, Marlow,1986). 1976;and 1986).While 1979;Williams Manageand (Mehra,1979; trendone point on determinationof trendon the the determination that Chowdhury's one may Chowdhury'spoint may argue argue that use of one one we argue stationarity is t~at use is interesting, and difference-stationarity arguethat difference-stationarity interesting,we stationarityand the is more subjective than objective. Moreover, given approach over the other than over the other is more objective.Moreover,giventhe subjective approach similarity in our results, we find it difficult to claim that Chowdhury's apthat Chowdhury'sapsimilarityin our results, we find difficult claim our transformationof our proach is the correct correcttransformation or necessary for the eithersuperior is either necessaryfor superioror proach data set. data set.

Results at the local level of government government Chowdhury of governresults at the local level of Chowdhury argues argues that he obtains different results

mente and disagree disagree with argument incorrectand with his his argument his comparison We consider considerhis ment. We comparisonincorrect that exerts no potential impact on the local government exerts no the local thatthe the level levelof aggregation potentialimpact government aggregation results. bivariate equation, Chowdhury determines determines the lag length for the lag For each each bivariate results.For equation,Chowdhury lengthfor the RHS RHS and the independent the variableon the and for for the variable. the lagged independentvariable. dependentvariable laggeddependent local level government since estithe local level of government four lag for the since he he estiHe determined determinedfour lengthsfor lag lengths one with mates bivariate equations (see his Table 4) one with expenditures and his Table two bivariate mates two 4) expendituresand equations(see one with these four, four, the the length length of three variable.Of these as the three with receipts the dependent one dependentvariable. receiptsas lags is two periods or less and the fourth lag equals three periods (Table 4). On three or and the fourth two less is lag equals periods(Table4). On lags periods the basis of these he finds a feedback relation between expenditures and finds a feedback relation between and the basis theselags, he lags, expenditures receipts. note that these interactions are observed observed that these it is is important interactionsare However,it importantto note receipts.However, for very structure (two periods) used used in in For the the shortest shortestlag for shortlag lag structure (two periods) lag lengths. very short lengths.For FPE our study (which is close to the lag structure derived the basis of the FPE the structure derived on the is close basis the our study (which lag criterion), we we found that revenues cause expenditures, but expenditures foundthat revenuescause do not not expenditures,but expendituresdo criterion), cause revenues. by ignoring our results results for the shortfor the shortcause revenues.Apparently, Apparently,Chowdhury, Chowdhury,by ignoringour est lag comparison by by considering considering only only the the restrictshis his comparison est structure,incorrectly incorrectlyrestricts lag structure, longer lags where we do not detect causality. We observed at the shortest lag We observed at the where do not detect shortest we causality. lag longerlags structure, ... the one case lends some the one wheresignificant is observed observedlends some case where structure,''... causalityis significantcausality empirical support (p. that tax tax receipts the hypothesis for the receiptssupport supportspending' hypothesisthat spending'(p. empirical supportfor 251). Contrary to Chowdhury's claim, we are not arguing that local expendiwe are not that local claim, arguing expendi251). Contrary Chowdhury's we observed tures independent of one observed causality one another. another.Since Sincewe turesand andrevenues revenuesare areindependent causality at argue that need to be be concerned these interinterthe shorter shorterlag, we argue that we we need concernedabout about these at the lag, we actions. actions. We our results results for longer lag lag strucwithChowdhury's We disagree claimthat thatour for the the longer strucChowdhury'sclaim disagreewith in no our methodology methodology and and are way influenced influenced tures be attributed only to our turescan can be attributedonly arein no way by aggregation problems. We We feel is naive. the many feel that that this this claim claim is naive. Given Given the by aggregationproblems. many diverse the financing of the myriad of local governments, the diverseconstraints constraintsaffecting the local affecting financing myriad governments, it is each locality locality is just just another another 'apple' barrel. difficult to argue is difficult that each the barrel. arguethat 'apple' in the Rather, the barrel is probably loaded with many many apples, bananas .... the barrel is probably loadedwith Rather, apples,oranges, oranges,bananas .... Given problems, we we argue that policy policy implications implications stemming from Giventhese these data data problems, arguethat stemmingfrom our require much much scrutiny. scrutiny. In our results at the the local resultsat local level level of government should require In governmentshould effect, this is just another application the well-known aggregation problem this is another of the well-known effect, just application aggregationproblem in in macroeconomics. macroeconomics.

Conclusion Conclusion It is very complex task. The It is worth worthrepeating that causality The diversity repeatingthat causalitytesting testingis aa very complextask. diversity viewson interpreting of views the definition causality and on the methods for testtestthe definition of and the for methods causality interpreting ing causality highlight this issue. In our study, we cite an excellent article article which of various problems associated with causality tests provides a detailed critique critique of (see Conway, et aI., draws al., 1984). 1984). It would appear appear that Chowdhury's comment draws very strong conclusions and fails to consider some of complex theoretical of the complex

issues problems pertaining pertaining as well well as as important tests as data problems issues regarding causalitytests importantdata regardingcausality to state Moreover, we find it somewhat curious that, we find it somewhat curious local governments. state and and local that, governments.Moreover, in our given prefers to dwell similaritiesin our two two studies, on dwellon the strong studies,Chowdhury Chowdhuryprefers strongsimilarities giventhe the an alternative alternativetechnique ratherthan than concentrate concentrateon on the the fact fact the mechanics mechanicsof an techniquerather that paper on the resultsto our our 1987 the that his his alternative alternativetechnique 1987paper techniqueyields yieldssupporting supportingresults tax-spend hypothesis. tax-spendhypothesis.

References References Blackley, between revenues the federal P.R. (1986). revenuesand and expenditures and the the size size of the federal Blackley,P.R. (1986).Causality expendituresand Causalitybetween budget. Public Public Finance Finance Quarterly 14: 139-156. 139-156. budget. Quarterly14: Chowdhury, and in state A.R. (1988). stateand andlocal localgovernment finances:ComComandreceipts (1988).Expenditures Expenditures receiptsin Chowdhury,A.R. governmentfinances: ment. Public Choice ment. Public Choice59: 59: 277-285. Conway, .A.V.B., Yanagida, .F., and TheimpossiandMuehln, von zur. zur.(1984). Muehln,P. von R.K., Swamy, Swamy,PP.A.V.B., (1984).The impossiConway,R.K., Yanagida,JJ.F., 36 (Summer): bility of causality Agricultural Economics Economics Research Research 36 causalitytesting. testing.Agricultural (Summer):1-19. bility Guilkey, properties of three for Granger-causal threetests andSalemi, M.K.(1982). Smallsample testsfor D.K., and Salemi,M.K. Granger-causal sampleproperties Guilkey,D.K., (1982).Small Review of Economics and ordering process. Review and Statistics Statistics64: 64: 668-680. bivariatestochastic stochasticprocess. orderingaa bivariate of Economics Manage, between federal and and Marlow, M.L. (1986). federalexpenditures The causal causalrelation relationbetween Marlow,M.L. expendituresand (1986). The Manage,N., and receipts. Economic Journal Journal 52 52 (January): SouthernEconomic (January):617-629. receipts.Southern in state Marlow, for causality state and receipts: and Manage, M.L., and Marlow,M.L., receipts:Testing causalityin Manage,N. (1987). (1987).Expenditures Expendituresand Testingfor and Public Choice finances.Public Choice53: 53: 243-255. local government and local governmentfinances. 85: Mehra, prices, and Journal of Political Economy Economy 85: and causality. Y. (1977). Mehra, Y. causality. Journal of Political Money wages, (1977). Money wages, prices, 1227-1244. 1227-1244. Ram, between government revenueand and government on causality R. (1988). Additionalevidence evidenceon Ram, R. causalitybetween governmentrevenue government (1988).Additional Economic Journal Journal 54: expenditure. 54: 763-769. SouthernEconomic expenditure.Southern Williams, and causality. D.H. (1976). income and and Gowland, Gowland,D.H. Goodhart,C.A.E., C.A.E., and Money, income causality. Williams,D., Goodhart, (1976). Money, American Economic Economic Review Review 66: American 66: 417-423.