Abstract

Forecasts of central government revenue are of crucial importance in the formulation of a financial program in order to establish the need for use of other financial resources, such as borrowing and depletion of accumulated reserves. The purpose of the present workshop is to illustrate the practical application to the case of Kenya of some of the principles and techniques involved in forecasting government revenue.1

Forecasts of central government revenue are of crucial importance in the formulation of a financial program in order to establish the need for use of other financial resources, such as borrowing and depletion of accumulated reserves. The purpose of the present workshop is to illustrate the practical application to the case of Kenya of some of the principles and techniques involved in forecasting government revenue.1

In principle, revenue forecasts should allow for feedbacks between taxes and their bases. For example, different tax rates on income probably affect the size of income itself. If sufficient data and other resources are available, a comprehensive econometric model can be constructed to take these linkages into account. The revenue implications of a particular change in the tax structure can then be worked out for any time period, whether past or future. For most developing countries—and Kenya is no exception—this degree of sophistication is not possible at present. A more practical approach, usually called the partial equilibrium approach, is to estimate the tax functions for different types of taxes and then project revenues on the basis of these functions, assuming given projections of the bases.

In this workshop, a partial equilibrium approach to revenue forecasting is adopted, treating the bases as predetermined variables. The first section provides data on government revenues and on selected bases for the various taxes and also deals with the problem of adjusting tax revenues for the effects of discretionary changes, using the proportional data adjustment method. The second section considers the specification of functional relationships between taxes and their bases, as well as the statistical estimation of a set of parameters that will reflect the current tax structure. The final section contains a series of exercises and issues for discussion. Appendix II presents a summary of the Kenyan tax system.

PREPARATION OF DATA

The first step in the forecasting exercise is to prepare the data on revenues and the bases for statistical estimation.

Tax Revenues and Bases

Since a tax system frequently consists of a large number of minor taxes, selectivity is needed in their disaggregation in order to avoid excessive detail. As a general rule, those taxes that contribute at least 2–5 per cent of revenue on average over the sample period should be isolated for detailed analysis. While the availability of data will determine how far the disaggregation can be carried, the breakdown should attempt, at a minimum, to distinguish between the major tax categories. In the International Monetary Fund’s Government Finance Statistics Yearbook, taxes are classified into seven major categories2 according to the nature of the base on which the tax is levied. In selecting the appropriate level of disaggregation within each major category, it is necessary to consider the statistical requirements for efficient estimation. A stable functional relationship for a group of taxes will not be established unless the individual taxes in the group exhibit strong correlations. Thus, a group of taxes should be disaggregated if their bases are quite dissimilar, particularly if some of these taxes yield large amounts of revenue. When taxes or groups of taxes involve both specific and ad-valorem rates, it would be appropriate to disaggregate revenue collections accordingly. Greater disaggregation may also help to minimize the number of discretionary changes (legal and administrative) that are present in a single series, because it is quite unlikely that individual taxes will be subject to the same frequency of discretionary changes as will a group of taxes. On the other hand, it is generally true that the more detailed the breakdown of revenue, the greater will be the likelihood that any series will exhibit fluctuations of an increasingly random character. A balance must be struck between these conflicting considerations.

Table 1 presents the time series of Kenya’s central government tax revenue, classified as in the Government Finance Statistics Yearbook, for the ten-year period covering fiscal years 1968—77 (years ending June 30). A relatively few major types of taxes account for the bulk of tax receipts (see Appendix II for a summary of the tax system). In the three-year period covering fiscal years 1975–77, collections from income tax, sales tax, import duties, and excise duties accounted for some 95 per cent of total tax receipts.

TABLE 1.

Kenya: Central Government Tax Revenue, Fiscal Years 1968–77

(In millions of Kenya shillings; year ending June 30)

article image
Sources: International Monetary Fund, Government Finance Statistics Yearbook, Vol. 1 (1977) and Vol. 2 (1978); and Kenya, Statistical Abstract, 1973 and Appropriation Accounts, Other Public Accounts and the Accounts of the Funds for the Year 1967/68 to 1970/71.

The income tax is the most important source of revenue; in fiscal year 1977 it accounted for some 40 per cent of tax revenue. There are essentially three components of the income tax: (1) taxes on companies, corporations, and enterprises; (2) taxes on individuals; and (3) a capital gains tax. The basic rate on net income for resident companies is 45 per cent, while that for nonresident companies is 52.5 per cent. The income tax structure for individuals is a progressive one, with marginal tax rates ranging from 10 per cent to 70 per cent. The capital gains tax, which was introduced in June 1975, is levied on the transfer of land, other property, and marketable securities. Companies are taxed at their corporation tax rate, and individuals are taxed at their income tax rate up to a maximum rate of 35 per cent.

The second most important tax is the sales tax, which was introduced in April-May 1973. It contributed 25 per cent of tax collections in fiscal year 1977. The sales tax is a single-stage tax levied on most manufactured goods, whether produced domestically or imported, and on electricity. The tax consists primarily of ad-valorem rates, ranging from 10 per cent to 30 per cent. Specific rates are levied on electricity as well as on a few commodities, such as petroleum products and beer. Exemptions include most animal and vegetable products, pharmaceuticals, fertilizers, certain raw materials, intermediate goods, and capital goods.

Import duties, the third most important source of tax revenue, were equivalent to about 20 per cent of tax receipts in fiscal year 1977. The import duty structure usually uses ad-valorem rates, ranging from 10 per cent to 150 per cent. However, specific duties are levied on some relatively important imports, including petroleum products, alcoholic beverages, and tobacco products. Duty-free items include live animals, pharmaceuticals, and fertilizers. Also, customs duties are refunded on the inputs of certain domestic industries—a rebate of increasing importance.

Excise duties provided 10 per cent of tax receipts in fiscal year 1977, but their significance has declined sharply since the introduction of the sales tax. Presently, excise taxes, which are levied at specific rates, are applied to domestically produced tobacco products, alcoholic beverages, and sugar, as well as to the purchase of secondhand motor vehicles.

There are four minor taxes or tax categories, none of which contributed more than 2 per cent of total tax revenue on average over the sample period. These include: (1) taxes on use of, or permission to use, goods or to perform activities; (2) taxes on specific services; (3) stamp duties; and (4) estate duty.

Effective July 1, 1977, export duties were imposed on coffee and tea.3 In 1976 exports of these commodities accounted for almost 40 per cent of exports. The basic rate of duty for both coffee and tea is 15 per cent of the export price exceeding KSh 20,000 per metric ton.

After the appropriate revenue classification is selected and the series of tax receipts have been obtained, quantitative information on the bases should be collected. While it would be desirable to obtain data corresponding to the legal tax bases, such information is often either unavailable or available in an unusable form. As a consequence, it may be necessary to select proxies for the legal bases. Frequently, some taxes or tax categories have more than one possible proxy base. In this case, time series should be collected on each alternative and should be studied through the use of statistical and other criteria, so as to facilitate the choice. The selection of the most appropriate proxy base would also depend on whether or not projections of a particular base are available for the period for which revenue forecasts are to be made.

Since the major tax categories often define bases that cover large parts of the economic activity of a country, proxy bases may be found among the variables comprising the national accounts and the balance of payments. These accounts are usually available in both value terms and volume terms. Furthermore, the use of the projections of these accounts makes it possible to examine the revenue forecasts from the viewpoint of overall consistency. Of course, when a particular tax category is highly selective in coverage, it is desirable to find a narrower proxy base that will reflect more adequately fluctuations in the legal base.

Some suggestions on the selection of proxy bases for tax revenues in Kenya appear in Table 2, and data on these are presented in Table 3. These rather aggregative proxy bases reflect to a large extent the lack of relevant disaggregate data for the major categories of revenue. For example, if data had been available on personal and corporate income taxes, it would have been more appropriate to select narrower bases for these taxes. For personal income tax payments, it might have been wages and salaries; and for corporate income tax payments, it might have been the operating surplus of enterprises. But in the case of Kenya, published information on such a breakdown of income taxes is not available for the most recent years.

TABLE 2.

Kenya: Suggested Proxy Bases for Tax Revenues

article image
TABLE 3.

Kenya: Data on Proxy Bases, Calendar Years 1967–771

article image
Sources: Kenya, Economic Survey, Statistical Abstract, and Statistical Digest, various issues.

National accounts and relevant customs data are available only on a calendar-year basis.

In millions of Kenya shillings.

Deflated by the consumer price index.

For import duties and excise duties, revenue collections are grouped by end-use categories and excisable-commodity categories, respectively.4 However, the disaggregate data are not presented on a fiscal-year basis, and, more importantly, no corresponding disaggregate estimates are published for the revenue effects of discretionary tax measures.5

For the sales tax, a breakdown of revenue collections is available on revenue from tax on domestically produced and imported manufactures.6 But this is not a particularly useful grouping from the viewpoint of estimation. And again, no disaggregate estimates are published for the revenue effects of discretionary tax measures.

Adjustment of Tax Revenue to Current Structure Basis

To estimate forecasting equations associated with a given tax structure, it is necessary to separate those changes in revenue that took place automatically in response to growth in the tax base from changes in revenue that occurred as a result of discretionary changes, whether legal or administrative. To do this, methods must be employed that convert the actual revenue data into a series that shows what revenues would have been if there had been no discretionary changes relative to the current year’s tax system.

Two methods may be applied to generate such a “clean” revenue series.7 The first is the constant rate structure method, which consists of applying the current year’s tax rates to the base values in the earlier years, thus simulating a revenue series that corresponds to the current year’s structure.8 Unfortunately, when applied properly, this procedure requires much information; for each tax category, it is necessary to have detailed information on rates, on the base, and on the distribution of the base by brackets or rate categories. Furthermore, the method depends on the assumption that changes in the tax structure do not affect the magnitudes of the subtax bases, which implies that resource allocation is not affected by changes in the relative prices. Because of the very heavy requirement of data and the restrictive assumptions, the constant structure method has found little practical application.

The other method is the proportional data adjustment method, which permits adjustment of a tax revenue series without any information on the tax base. Apart from data on the actual revenue collections, the only additional information required in applying this method is estimates of the quantitative impact on revenue of discretionary changes for the years in which the changes took place.9 Assume the following series of actual revenue collection in years 1 through n:

T1, T2,…, Tn-1, Tn,

where n is the last or the current year.

Let the estimated revenue effect of discretionary changes in the years in which they occurred be as follows: 10

D2,…, Dn-1, Dn.

Let us now construct an adjusted series of revenue equal to those revenues that would pertain if the current year’s tax structure had been in operation throughout the period under analysis:

AT1, AT2,…, ATn-1, ATn.

Since the current year (n) is the reference year, the actual and adjusted revenue for that year would be the same

ATn = Tn.

However, actual revenue collection in years n-1, n-2, …, 2, 1, must be corrected for discretionary changes in subsequent years. Under the proportional adjustment hypothesis, the series of adjusted revenue, with n as the reference year, becomes as follows:11

ATn1=Tn1(TnTnDn)
ATn2=Tn2(TnTnDn)(Tn1Tn1Dn1)
AT2=T2(TnTnDn)(Tn1Tn1Dn1)(T3T3D3)
AT1=T1(TnTnDn)(Tn1Tn1Dn1)(T3T3D3)(T2T2D2)

which can be expressed by the general formula:

ATnj=Tnj(ATnj+1Tnj+1Dnj+1),j=1,2,,n2,n1.

It may be helpful to provide a numerical illustration of the method. Let the actual revenue collections in years 1 through 5 be as follows:

T1 = 100, T2 = 140, T3 = 170, T4 = 250, T5 = 320.

Assume that discretionary changes occurred only in years 2 and 4, with

D2 = 20, D3 = 0, D4 = 30, D5 = 0.

Let us now construct an adjusted revenue series with year 5 as reference year. The effect of the discretionary action in year 4 was to raise revenue collections in that year by 13.64 per cent:

T4T4D4=25025030=1.1364.

If this discretionary action had been taken at the beginning of the data period, the proportional adjustment hypothesis would imply a corresponding proportionate increase in revenue collections in all years preceding year 4. Similarly, the effect of the discretionary action in year 2 was to raise revenue in that year by 16.67 per cent:

T2T2D2=14014020=1.1667.

Only revenue in year 1 needs to be corrected for the change in year 2. Thus, if both discretionary actions had been taken at the beginning of the data period, the revenue series would have become as follows under the proportional adjustment hypothesis:

AT5=T5=320AT4=T4=250AT3=T3(T4T4D4)=170×1.1364=193.2AT2=T2(T4T4D4)=140×1.1364=159.1AT1=T1(T4T4D4)(T2T2D2)=100×1.1364×1.1667=132.6.

This workshop applies the proportional data adjustment method to generate a “clean” revenue series, showing what the revenues would have been if there had been no discretionary changes relative to the current year’s tax structure. In Kenya, estimates of the revenue effects of discretionary measures are prepared at the time the budget is formulated and are published in the budget speech. As can be seen from Table 4, discretionary changes have occurred very frequently for all major tax categories, thus effectively ruling out the application of the dummy variable technique to the raw revenue data. Since the actual revenue effect of discretionary changes may well differ from the ex ante budget estimates, an attempt should be made to update these estimates. In Kenya, however, no ex post revision of the initial estimates is provided. One possible approach to correcting the ex ante budget estimates would be to apply the ratios of actual to estimated overall yield of each tax, on the assumption that the percentage error in estimating the revenue effect of a discretionary change of a particular tax would be the same as the percentage error in forecasting the overall yield of that same tax. This assumption may not be entirely appropriate, but it is reasonable, at least.12 Table 5 provides data on the ratios of actual to estimated tax revenue, which can be applied to the ex ante budget estimates to obtain ex post estimates of the revenue effect of discretionary changes in the years in which they occurred. These revised estimates may be used to construct the adjusted revenue series.

TABLE 4.

Kenya: Budget Estimates of the Revenue Effects of Discretionary Tax Measures, Fiscal Years 1969–77

(In millions of Kenya shillings; year ending June 30)

article image
Source: Kenya, Financial Statements, 1968/69 to 1976/77.

Excluding the graduated personal tax, which was abolished December 31, 1973.

The sales tax was introduced in April-May 1973, and no budget estimate was available for fiscal year 1973.

Excluding the graduated personal tax, which was abolished December 31, 1973, and the export duties that were removed June 15, 1973.

TABLE 5.

Kenya: Ratios of Actual to Estimated Tax Revenue, Fiscal Years 1969–77

(In per cent; year ending June 30)

article image
Source: Kenya, Appropriation Accounts, Other Public Accounts and the Accounts of the Funds for the Year 1968/69 to 1976/77.

Excluding the graduated personal tax, which was abolished December 31, 1973.

The sales tax was introduced in April-May 1973, and no budget estimate was available for fiscal year 1973.

Excluding the graduated personal tax and export duties.

Nontax Revenue

Data on central government nontax revenue, classified as in the Government Finance Statistics Yearbook, are presented in Table 6 for the six-year period of fiscal years 1972–77. In fiscal year 1977, nontax revenue contributed almost 14 per cent of total revenue. The two most important categories are (1) property income, which takes the form of dividends, interest, rents, royalties, or withdrawal from entrepreneurial income; and (2) administrative fees, charges, and nonindustrial sales.

TABLE 6.

Kenya: Central Government Nontax Revenue, Fiscal Years 1972–77

(In millions of Kenya shillings; year ending June 30)

article image
Source: International Monetary Fund, Government Finance Statistics Yearbook, Vol. 1 (1977) and Vol. 2 (1978).

Systematic relationships may be found for certain components of nontax revenue, such as administrative fees, but in other instances, such as profits of the Central Bank, there may be considerable volatility in the time series. Of necessity, the method of projection must be ad hoc in many cases. Sometimes a forecast is made of total nontax revenue by applying the last year’s ratio of total nontax revenue to gross domestic product (GDP) to the projected value of GDP. However, rather than relying on a mechanistic projection of the aggregate, it might be feasible to allow for expected changes in relevant factors through judgmental adjustments in the individual components.

STATISTICAL ESTIMATION

Having adjusted the actual revenue series for the effects of discretionary changes, the major steps in the forecasting procedure consist of specifying functional relationships between the taxes and the selected proxy bases, estimating a set of parameters reproducing the current tax structure, and applying these estimates to forecast tax revenues for given values of the bases.

Specification of Functional Relationships

Choosing the proxy bases suggested in Table 2, the following relationships were specified for the individual taxes or tax categories.

Taxes on Income, Profits, and Capital Gains

ATIt=a+bYt+cDAt(1)
LnATIt=a+bLnYt+cLnDBt(2)

where

ATI = adjusted revenue from taxes on income, profits, and capital gains

Y = gross domestic product at factor cost at current prices.

DA and DB are dummy variables for changes in the timing of the payment of income taxes:13

DAt={0forfiscalyears1968701forfiscalyears197177LnDBt={0forfiscalyears1968701forfiscalyears197177.

Sales Tax

ATSt=a+bCPt(3)
LnATSt=a+bLnCPt(4)

where

ATS = adjusted revenue from sales tax

CP = private consumption expenditure.

Import Duties

ATMt=a+bMt+cM¯t(5)
LnATMt=a+bLnMt+cLnM¯t(6)

where

ATM = adjusted revenue from import duties

M = imports (customs basis) at current prices

M¯ = imports (customs basis) at constant prices.

Excise Duties

ATEt=a+bCP¯t(7)
LnATEt=a+bLnCP¯t(8)

where

ATE = adjusted revenue from excise duties

CP¯ = private consumption expenditure deflated by the CPI.

Minor Taxes

Estimating equations were specified for three of the four minor taxes as follows:

LnATUPt=a+bLnY¯t(9)
LnATSDt=a+bLnY¯t(10)
LnATSSt=a+bLnCPt(11)

where

ATUP = adjusted revenue from taxes on use of, or permission to use, goods or to perform activities

ATSD = adjusted revenue from stamp duties

ATSS = adjusted revenue from taxes on specific services

Y¯ = gross domestic product at factor cost at constant prices.

Estimation Results

Since data on the bases were not available on a fiscal-year basis, it was decided to regress tax revenue in fiscal year t on two alternative bases:14

(a) LBt = lagged, calendar-year base;

for example, LB 1968 is equal to calendar-year 1967 base; and

(b) MBt = mean of two calendar-year bases;

for example, MB 1968 is equal to the mean of the bases in the two calendar years 1967 and 1968.

The parameters were estimated by the method of ordinary least squares. The estimation results are shown below with t values in parentheses.15 Given the limited number of observations, considerable care is needed in the evaluation of the regression estimates; in particular, for the sales tax, there are only four observations to estimate the two parameters of the regression equation.

Taxes on Income, Profits, and Capital Gains

Estimation period: fiscal years 1968–77

ATIt=257.07(6.94)+0.09049(28.44)LYt+165.63(4.45)DAt(1a)R¯2=0.995D.W=2.50
ATIt=154.96(3.33)+0.07570M(21.00)Yt+184.56(3.71)DAt(1b)R¯2=0.990D.W=2.23
LnATIt=3.68010(6.98)+1.09750LnL(18.95)Yt+0.26426Ln(5.78)DBt(2a)R¯2=0.992D.W=3.08
LnATIt=2.99441(5.61)+1.0734LnM(17.42)Yt+0.26577Ln(5.36)DBt(2b)R¯2=0.991D.W=2.79

Sales Tax

Estimation period: fiscal years 1974—77

ATSt=163.23(1.76)+0.06622(10.29)LCPt(3a)R¯2=0.972D.W=2.19
ATSt=172.30(1.52)+0.06002(8.31)MCPt(3b)R¯2=0.958D.W=2.31
LnATSt=1.28589(1.59)+0.86750(10.22)LnLCPt(4a)R¯2=0.972D.W=2.30
LnATSt=1.32644(1.58)+0.86375(9.88)LnMCPt(4b)R¯2=0.970D.W=2.29

Import Duties

Estimation period: fiscal years 1968—77

ATMt=157.59(0.71)+0.07183(5.14)LMt+0.11081(1.87)LM¯t(5a)R¯2=0.769D.W=2.89
ATMt=23.98(0.10)+0.07018(6.51)MMt+0.15341M(2.44)M¯t(5b)R¯2=0.833D.W=2.17
LnATMt=0.38600(0.20)+0.40020(5.40)LnLMt+0.37235LnL(1.53)M¯t(6a)R¯2=0.791D.W=2.74
LnATMt=0.74410(0.37)+0.40859(7.11)LnMMt+0.49701LnM(2.05)M¯t(6b)R¯2=0.862D.W=2.37

Excise Duties

Estimation period: fiscal years 1968—77

ATEt=236.72(3.80)+0.04959(10.67)LCP¯t(7a)R¯2=0.926D.W=2.20
ATEt=269.84(4.11)+0.05083(10.63)MCP¯t(7b)R¯2=0.926D.W=1.74
LnATEt=8.95805(7.19)+1.57886(12.00)LnLCP¯t(8a)R¯2=0.941D.W=2.26
LnATEt=9.70251(7.73)+1.65294(12.52)LnMCP¯t(8b)R¯2=0.945D.W=1.82

Minor Taxes

Taxes on Use of, or Permission to Use, Goods or to Perform Services

Estimation period: fiscal years 1968—77

LnATUPt=10.70870(14.15)+1.58136(19.71)LnLY¯t(9a)R¯2=0.977D.W=1.43
LnATUPt=11.26880(21.91)+1.63591(30.11)LnMY¯t(9b)R¯2=0.990D.W=1.76

Stamp Duties

Estimation period: fiscal years 1968—77

LnATSDt=15.99393(11.88)+2.06526(14.47)LnLY¯t(10a)R¯2=0.959D.W=2.79
LnATSDt=16.54503(11.03)+2.11742(13.36)LnMY¯t(10b)R¯2=0.952D.W=2.71

Taxes on Specific Services

Estimation period: fiscal years 1971—77

LnATSSt=4.33160(3.50)+0.84651(6.36)LnLCPt(11a)R¯2=0.868D.W=0.98
LnATSSt=4.33701(3.70)+0.83950(6.73)LnMCPt(11b)R¯2=0.881D.W=0.82
TABLE 7.

Kenya: Actual and Predicted Values for Selected Equations, Fiscal Years 1968–771

(In millions of Kenya shillings)

article image

Series under the variable name are actual tax revenues adjusted for the effect of discretionary tax measures, as defined in the text. The numbers in parentheses indicate the equations from which the predictions were obtained.

See text for an explanation of the variables and the equations.

EXERCISES AND ISSUES FOR DISCUSSION

Exercises

  • 1. Use the proportional adjustment method to construct adjusted revenue series, with fiscal year 1977 as reference year, for the four major tax categories: (l) taxes on income, profits, and capital gains; (2) sales tax; (3) excise duties; and (4) import duties. The necessary data for the calculations are given in Tables 1, 4, and 5, and a sample worksheet for the proportional adjustment method is provided in Appendix I. Note that revenue from the graduated personal tax, which was abolished on December 31, 1973, must be deducted from taxes on income, profits, and capital gains, as presented in Table 1. For the sales tax, which was introduced in April-May 1973, construct an adjusted series of revenue for fiscal years 1974–77

  • 2. Examine the relationships of tax to base as specified and estimated in the section on Statistical Estimation. In particular:

    (a) What is the significance of the dummy variables in equations (1) and (2)? Use one of the estimated regression equations (2a) or (2b) to test the hypothesis that the elasticity of taxes on income, profits, and capital gains with respect to GDP at factor cost at current prices is equal to unity.

    (b) Comment on the problems that may arise in the estimation of equations (5) and (6). What are the elasticities of import duties in equation (6a) or (6b) with respect to import volume and import price?

    (c) Compare the estimation results with a lagged, calendar-year base (LB) to those with an average of two calendar-year bases (MB).

  • 3. For each tax or tax category, use one of the estimated regression equations with a lagged base to make a forecast for fiscal year 1978. Data on the bases for calendar year 1977 are provided in Table 3 and budget estimates of the revenue effect of discretionary measures in fiscal year 1978 are shown in Table 8. For estate duties that do not exhibit any clear trend, make a forecast by taking the average of the last three years.

  • 4. On the basis of the forecasts made in Exercise 3 for the individual taxes or tax categories for fiscal year 1978, calculate:

    (a) the expected built-in elasticity of total tax revenue with respect to lagged, calendar-year GDP at factor cost at current prices; and

    (b) the expected overall elasticity of total tax revenue with respect to lagged, calendar-year GDP at factor cost at current prices.

  • 5. Compare the forecasts made in Exercise 3 for each tax or tax category with the budget estimates of tax revenue for fiscal year 1978 shown in Table 9. Where do the major differences lie and what factors may account for the differences?

  • 6. Make a forecast of total nontax revenue for fiscal year 1978 by applying the ratio of nontax revenue in fiscal year 1977 to GDP at factor cost at current prices in calendar year 1976 to the value of GDP at factor cost at current prices in calendar year 1977.

TABLE 8.

Kenya: Revenue Estimates of Discretionary Tax Measures, Fiscal Year 1978

(In millions of Kenya shillings; year ending June 30)

article image
Source: Kenya, Financial Statement, 1977/78.
TABLE 9.

Kenya: Budget Estimates of Tax Revenue, Fiscal Year 1978

(In millions of Kenya shillings; year ending June 30)

article image
Source: Kenya, Estimates of Revenu, 1977/78.

Issues for Discussion

  • 1. Comment on the classification of tax revenue data presented in Table 1. Are there tax categories for which a further breakdown would have been desirable? Identify minor taxes or tax categories that may be forecast together rather than on a tax-by-tax basis.

  • 2. Provide reasons for the choices of proxy bases suggested in Table 2. In particular, why would it be inappropriate to choose a common base for domestic taxes on goods and services?

  • 3. Discuss the assumptions and possible limitations of the proportional data adjustment method.

  • 4. Comment on the data on budget estimates and outturns provided in Table 5. In particular, what factors may account for the systematic underestimation of taxes on income, profits, and capital gains?

  • 5. What criteria would you apply in choosing between the various specifications of the tax functions from the point of view of revenue forecasting?

  • 6. Discuss how an increase in domestic prices would affect the built-in elasticity of tax revenue with respect to GDP at factor cost at current prices.

  • 7. Comment on the time series of nontax revenues presented in Table 6. Indicate what kind of information would be desirable for forecasting the major components.

APPENDIX I

Kenya: Worksheet for Proportional Adjustment Method

(Reference year: Fiscat Year 1977)

article image
Notation:t = Fiscal yearTt = Actual revenueBDt = Ex ante budget estimate of the revanue effect of discretionary action in the year in which it was takenRt = Ratio of actual to estimated tax revenueDt = Ex post estimate of the revenue effect of discretionary action in the year in which it was takenAFt = Adjustment factor with 1977 as reference year (Note that AFt = 1 for t = 1977.)ATt = Adjusted revenue with 1977 as reference year (Note that ATt = Tt for t = 1977.)

APPENDIX II

Kenya: Summary of Tax System, July 1977

article image
article image
article image
article image
article image
article image
article image
article image
article image
article image
article image
article image
article image
article image
article image
article image
article image
Source: Information provided by the Kenyan authorities.
1

Much of the work on establishing procedures for revenue forecasting in developing countries has been done by the Fiscal Affairs Department of the International Monetary Fund. This workshop draws from several of these unpublished contributions, especially from Roy W. Bahl, “Alternative Methods for Tax Revenue Forecasting in Developing Countries: A Conceptual Analysis” (October 1972); and Sheetal K. Chand, “Some Procedures for Forecasting Tax Revenue in Developing Countries” (October 1975).

2

See Workshop 2: Government Finance Statistics.

3

A previous system of export duties was terminated, effective June 15, 1973.

4

These data are published in Kenya, Economic Survey and Statistical Abstract, on a calendar-year basis.

5

The adjustment of tax revenue for the effects of discretionary measures is discussed below.

6

The data can be obtained from Kenya, Appropriation Accounts, Other Public Accounts and the Accounts of the Funds for the Year (annual).

7

When discretionary changes are few relative to the number of observations, their effects on the structure can be estimated by the technique of dummy variables.

8

For an application of this method and a discussion of some of the issues involved in using it, see Nurun N. Choudhry, “A Study of the Elasticity of the West Malaysian Income Tax System, 1961–70,” Staff Papers, Vol. 22 (July 1975), pp. 494–509.

9

This method has been applied in a number of Fund studies; see, for example: Charles Y. Mansfield, “Elasticity and Buoyancy of a Tax System: A Method Applied to Paraguay,” Staff Papers, Vol. 19 (July 1972), pp. 425–46; Hessel J. Baas and Daryl A. Dixon, ‘The Elasticity of the British Tax System, 1950/51–1970/71” (unpublished paper, September 1974); and Sheetal K. Chand, “Tax Revenue Forecasting: An Approach Applied to Malaysia” (unpublished paper, March 1975).

10

Note that the first discretionary change occurs in year 2, since year 1 is the first year in the time series.

11

The assumption that discretionary actions affect revenue proportionately implies that the built-in elasticity of the particular tax is unaffected. Thus, the effects of discretionary actions are equivalent to multiplicative shifts in the relationship of the tax to the base.

12

If some of the estimates of the revenue effect of discretionary action seem excessively unreliable, it may be possible, under certain circumstances, to estimate their effects on the tax structure by applying the technique of dummy variables in conjunction with the proportional data adjustment method.

13

The revenue effects of these changes are not taken into account in Table 4. Actually, the changes produced both permanent and windfall gains in tax revenue, but the specification of equations (1) and (2) attempts to estimate the permanent effect on the tax structure.

14

Note that fiscal year 1968 covers the period July 1,1967-June 30,1968.

15

The predicted values for some of the equations are presented in Table 7, together with the actual revenue series adjusted for the effects of discretionary tax measures.