Local government revenues and expenditures in Uganda: A VAR ...

More documents

Recommendations

Info

{ G t} be affected by the current and past realizations of the { t}T sequence and the let thetime path of the { T t} sequence be affected by current and past realizations of the{ G t} sequence. Our simple bivariate system can be represented as a first order VAR3 asfollows:[1]G = bt10+ b12Tt+ γ G11t−1+ γ T12t−1+ εGt[2]T = bt20+ b21Gt+ γ G21t−1+ γ T22t−1+ εTtWhere (as verifed above), both G and are stationary and thetT { ε } and { ε }tuncorrelated white noise disturbances. Here, is the contemporaneous effect of a unitchange of T on and γ the effect of a unit change in G on T . In other words, thetGt21structure of this system incorporates feedback since Gtand Ttare allowed to affect eachother. After some manipulation, the above structural VAR (equations [1] and [2]) can bereduced to following standard form:b 12t−1tGtTtare[3]G = at10+ a11Gt−1+ a12Tt−1+ e1t[4]T = at20+ a21Gt−1+ a22Tt−1+ e2tSince equations [3] and [4] each have identical right-hand-side variables, estimation viaordinary least squares technique will yield efficient estimates. Table 3 shows results fromthe VAR estimation and were obtained with one lag for each variable. The results inTable 3 confirm the uni-directional causation between expenditures and revenues withcausation running from government revenues to government expenditures. In otherwords, local government revenues and expenditures are related by a feedback causalmechanism, and this is consistent with the tax-spend hypothesis indicating that the localgovernments in Uganda raise revenues first and then identify spending priorities.Table 3. VAR estimation results: 2001-2003Equationln revenues ln expendituresln revenues 0.636(6.77)***0.067(1.68)*ln expenditures -0.077(0.36)0.505(5.54)***Constant 9.292 9.7272Adj. R 0.37 0.40Durbin-Watson 1.83 1.943 The optimal lag length was determined using the Schwarz BIC model selection criterion.5
Prob (F-stat.) 0.0 0.0Number of observations 167 167Notes: Absolute value of z statistics in parentheses;* Significant at 10%; ** significant at 5%; *** significant at 1%.Further, the results reveal a low “tax-spend” elasticity. In particular, a 10 percent increasein revenues results in a 1 percent increment in expenditures. Though not the focus of thisstudy the inelastic response of expenditures to changes in revenues is has importantpublic finance implications. Public finance economists have come to refer to thisphenomenon by different names ranging from the “fly paper” effect to mere abuse ofoffice and/ or corruption. The former school of thought is premised on the argument that“money tends to stick” where is lands since only a fraction of a given dollar of publicfunds utilized in the provision of public goods and services. Proponents of the latterschool of thought argue that the corruption and graft that has often plagued the publicsector is one single most important reason for the absence of a one to one relationshipbetween revenues raised and public expenditures.4. Testing for Autocorrelation in the ResidualsWe conduct a serial autocorrelation test in the residuals via the lagrange-mulitplier testwhich has a null hypothesis of no autocorrelation. The lagrange-multiplier test does notreject the null hypothesis of no autocorrelation in the residuals at the optimal lag order(one) which is consistent with a well specified model. The Chi-square statistic and p-values obtained at lag one (1) are 5.27 and 0.26, respectively, which does not allow forthe rejection the null hypothesis of no autocorrelation.Following Engle and Granger (1987), we regress the first difference of the residuals onthe previous period disturbances. The test for unit root in the disturbances comprisestesting whether the coefficient on the lagged disturbances is not statistically differentfrom zero. Failure to reject the null hypothesis that this coefficient is not statisticallydifferent from zero indicates that the sequence of disturbance terms contains a unit root.Applying this procedure to our data leads to the rejection of the null hypothesis indicatingthat the residuals are stationary which rules out the null hypothesis of spuriousregression. 45. Discussion of the Empirical ResultsThe Granger-causality test and the VAR reveal unilateral causation between governmentrevenues and expenditures with causation running from revenues to expenditures forUganda’s local governments. In light of these findings, we can conclude that localgovernments in Uganda follow the “tax-spend” hypothesis. The “tax-spend” hypothesismeans that local governments in Uganda seem to raise tax revenue and/or request for orreceive grants before engaging in new expenditures. The “tax-spend” hypothesis is in linewith the concept of budget framework papers which requires local governments to4 We reject the null hypothesis of unit root in the disturbances at the 1 percent level of significance.6
Page 1 and 2: LOCAL GOVERNMENT REVENUES AND EXPEN
Page 3 and 4: Payne (1998) have examined the case
Page 5: that our variables are stationary o
Page 9 and 10: ReferencesAraujo, J. T., and M.A.C.

Local government revenues and expenditures in Uganda: A VAR ...

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?