# 0000 YOM INSTITUTE OF ECONOMIC DEVELOPMENT JOINT MASTER’S PROGRAM WITH DEBRA MARKOS UNIVERSITY ESTIMATING ADDIS ABABA TAX OFFICES EFFICIENCY

0000

YOM INSTITUTE OF ECONOMIC DEVELOPMENT

JOINT MASTER’S PROGRAM WITH DEBRA MARKOS UNIVERSITY

ESTIMATING ADDIS ABABA TAX OFFICES EFFICIENCY:

A DATA ENVELOPMENT ANALYSIS APPROACH

JEMAL ESLEMAN JIBRIL

ADVISOR: ZERAZYEHU SIME (PhD)

MSc THESIS SUBMITTED TO DEBRE MARKOS UNIVERSITY AND YOM INSTITUTE OF ECONOMIC DEVELOPMENT

A THESIS SUBMITTED TO THE DEPARTMENT OF ECONOMICS OF DEBRE MARKOS UNIVERSITY AND YOM INSTITUTE OF ECONOMIC DEVELOPMENT, IN PARTIAL FULFILMENT FOR THE REQUIREMENTS OF MASTERS OF SCIENCE DEGREE IN PROJECT PLANNING AND MANAGEMENT

MAY, 2018

ADDIS ABABA, ETHIOPIA

DECLARATIONS

I Jemal Esleman Jibril, registration number/I.D. number GSR/253/09, do hereby declare that this thesis is my original work and that it has not been submitted partially; or in full, by any other person for an award of degree in any other university/institution.

Submitted by:

Full Name Jemal Esleman Jibril Signature——————————-Date———————-

Approved by:

This Thesis has been submitted for examination with my approval.

Name of Advisor: Zerayehu Sime (PhD) Signature————————–Date———————–

APPROVAL

The undersigned certify that they have read and hereby recommend to Debre Markos University and Yom Institute of Economic Development to accept the Thesis submitted by Jemal Esleman Jibril and entitled on Estimating Addis Ababa Tax Offices Efficiency: A Data Envelopment Analysis Approach, in partial fulfilment of the requirements for the award of a Master’s of Science Degree in Project Planning and Management

Submitted by:

Full Name Jemal Esleman Jibril Signature—————————Date———————–Approved by:

Name of Advisor: Zerayehu Sime (PhD) Signature———————– Date——————–

Name of Internal Examiner: ———————————Signature——————-Date————

Name of External Examiner: ——————————–Signature——————-Date————

Name of Head of Department: Birhanu Shanko Signature———————-Date————-

DEDICATION

This research work is dedicated to my dad, sheikh Suleiman Jibril, who passed away on the course of this paper, may Allah be pleased with him and for all Ethiopians, especially for those who lost their indispensable soul asking for peoples’ freedom and democracy on the zenith of the recent struggle that happened in my beloved country.

ACKNOWLEDGMENTS

Allah said “if you count the blessings of Allah, you can’t finish”. He endowed me a lot of things. It’s very minor thing which I got from his blessings. Thank you, my God, for your blessings from start to finish of this paper. Next, there were many people contributed their efforts to the fruitfulness for this research work. The forehead is Ms. Amira Beshir, my beloved wife, thank you all for what you furnished me. my family members in general and, my friends let allow me to extend my appreciation for your invaluable assistance.

My advisor Dr Zerayehu Sime and my Co-advisor Yifru Yirdaw, thank you for your earnest advice, support and tolerance for my queries and hassles.

The data was collected from Ethiopian Revenues and Customs Authority (ERCA) and Addis Ababa Finance and Economic Cooperation Bureau (AAFECB). I am indebted to acknowledge Mr. Tamirat Anise, Mr. Bizualem and Information Technology Staff of ERCA and Mr. Dereje, Mr. Eyasu and Mr. Abiy from AAFECB. Without the strong support of them, the paper would not be successful.

Finally, I am not just to end my gratitude without mentioning Rosa S. for her invaluable backing in doing this thesis. May God reward your good deeds.

TABLE OF CONTENTS

TOC o “1-3” h z u 1. BACKGROUND OF THE STUDY …………………………………………………………..41.1 INTRODUCTION41.2 STATEMENTS OF THE PROBLEM51.3 RESEARCH OBJECTIVES71.3.1 GENERAL OBJECTIVE71.3.2 SPECIFIC OBJECTIVES71.4 RESEARCH QUESTIONS71.5 SCOPE OF THE STUDY71.6 SIGNIFICANCE OF THE STUDY71.7 LIMITATIONS OF THE STUDY81.8 DESCRIPTION OF THE STUDY AREA92. LITERATURE REVIEW…………………………………………………………………….102.1 THEORETICAL LITERATURES…..102.1.1TYPES OF EFFICIENCY MEASUREMENT MODELS112.1.2 HISTORY AND APPLICATIONS OF DEA122.1.3 ADVANTAGES AND LIMITATIONS OF DEA132.2 EMPIRICAL REVIEW132.3 CONCEPTUAL FRAMEWORK163. RESEARCH METHOD……………………………………………………………………..173.1 INTRODUCTION173.2 OPERATIONAL DEFINITION OF VARIABLES173.3 RESEARCH DESIGNS183.4 MODEL SPECIFICATION193.5 RESEARCH METHODS223.6 SAMPLE DESIGN223.7 POPULATION OR UNIVERSE233.8 SAMPLING FRAME233.9 SAMPLING UNIT/UNIT OF ANALYSIS233.10 SAMPLE SIZE233.11 SOURCES OF DATA233.12 DATA ANALYSIS AND INTERPRETATION:233.13 DATA PRESENTATION:234. REFERENCES ………………………………………………………………………………245. APPENDIX1 RESEARCH ACTIVITY AND TIME SCHEDULE ………………………266. APPENDIX2 BUDGET PLAN FOR THE THESIS ………………………………………27

LIST OF TABLES

Table 1: Descriptive statistics for inputs, outputs and other variables of Addis Ababa tax offices

Table 2: Comparison of the characteristics of the variables in each tax offices

Table 3 CRS, VRS and Scale Efficiency results of fourteen DMUs

Table 4: Efficiency Results of twelve DMUs under CRS Frontier

Table 5: Rank of Tax Offices under CRS-Input Orientation Frontier

Table 6: Efficiency Results of twelve DMUs under VRS Frontier

LIST OF FIGURES

Figure 1 CRS input orientation efficiency frontier line with two inputs X1, X2 and one output Y2

LIST OF APPENDICES

Appendix A1 Efficiency Result Summary in CRS Model

Appendix A2 Summary of Output and Input Slacks

Appendix A3 Summary of Peers and Peer Weights

Appendix A4 Peer Count Summary

Appendix A5 Summary of Output and Input Targets

Appendix B1 Firm by Firm Results

Appendix B2 Firm by Firm Results

Appendix B3 Firm by Firm Results

Appendix B4 Firm by Firm Results

Appendix B5 Firm by Firm Results

Appendix B6 Firm by Firm Results

Appendix C1 Efficiency Result Summary in VRS Model

Appendix C2 Summary of Input and Output Slacks

Appendix C3 Summary of Peers and Peer Counts

Appendix C4 Peer Count Summary

Appendix C5 Summary of Output and Input Targets

Appendix D1 Firm by Firm Results

Appendix D2 Firm by Firm Results

Appendix D3 Firm by Firm Results

Appendix D4 Firm by Firm Results

Appendix D5 Firm by Firm Results

Appendix D6 Firm by Firm Results

Appendix E1 CRS, VRS And Scale Efficiency Results of Fourteen DMUs

Appendix E2 EVIEWS Results for Descriptive Statistics for Inputs, Outputs and Other Variables Of 12 Addis Ababa Tax Offices

Appendix E3 Comparison of The Characteristics of The Variables in Each Tax Offices

Appendix E4 Empirical Data of Input, Output and Other Variables for the year 2015/16

Appendix F1 Research Activity and Time Schedules

Appendix F2 Budgets

LIST OF ACRONYMS

TNTs = Total Number of Taxpayers

ORE = Office Rent Expense

TNEs = Total Number of Employees

DTs = Direct Taxes

IDTs = Indirect Taxes

AAM1 = Addis Ababa Number 1 Medium Tax Payers Branch Office

AAM2 = Addis Ababa Number 2 Medium Taxpayers Branch Office

MR1 = Merkato Number 1 Project Office

MR2 = Merkato Number 2 Project Office

ADI = Addis Ketema Sub City Small Taxpayers Branch Office

AKK = Akaki Kaliti Sub City Small Taxpayers Branch OfficeARA = Arada Sub City Small Taxpayers Branch Office

BOL = Bole Sub City Small Taxpayers Branch Office

GUL = Gulele Sub City Small Taxpayers Branch Office

KIR = Kirkos Sub City Small Taxpayers Branch OfficeKOK = Kolfe Keranyo Sub City Small Taxpayers Branch Office

LID = Lideta Sub City Small Taxpayers Branch Office

NIL = Nifassilk Lafto Sub City Small Taxpayers Branch Office

YEK = Yeka Sub City Small Taxpayers Branch Office

AAFECB = Addis Ababa Finance and Economic Cooperation Bureau

AARA = Addis Ababa Revenues Authority

ERCA = Ethiopian Revenues and Customs Authority

DMUs = Decision Making Units

DEA = Data Envelopment Analysis

CRS = Constant Returns to Scale

VRS = Variable Returns to Scale

BCC = Banker, Charnes And Cooper

CCR = Charnes, Cooper and Rhodes

DRS = Decreasing Returns to Scale

IRS = Increasing Returns to Scale

TE = Technical Efficiency

SE = Scale Efficiency

ABSTRACT

Governments have chosen to implement new taxes and increased tax rates in the expectation of finding a solution to its diminishing tax revenue, while improving the efficacy and efficiency of the body responsible with managing taxation might offset this need. Efficiency can be simply defined as the ratio of output to input. Efficiency is evaluated against a best practice frontier using CRS and VRS input-oriented, Data Envelopment Analysis (DEA) approach.

The paper analyses the technical and Scale Efficiency of tax offices, that is, the offices which deal with declarations, assessments, collection and disputed tax demands of individuals and companies at Addis Ababa City level, using data from ERCA and AAFECB. The objective of this paper is to study the technical and scale efficiency of Addis Ababa City Administration tax offices in order to establish how far the City Government’s policy is achieving its aims.

The study tries to quantify efficiency measures for 12 small and medium tax offices, in charge of direct and indirect taxes within the jurisdiction of Addis Ababa City Administration.

The study used Total Number of Taxpayers, Office Rent Expense and Total Number of Employee as input and Direct and Indirect Taxes as output for the year 2015/16.

Overall, the average pure efficiency scores indicate that tax offices should be able to collect their current level of revenues with approximately 9% percent less inputs. Moreover, adequate chance exists for tax offices to increase their efficiency through management of reported data.

The analysis shows that the main source of overall technical inefficiency in the tax collecting offices is due to pure technical inefficiencies than scale inefficiencies. This suggest that tax collecting offices managers must improve their operational planning and management practices in an efficient way. This could be via adopting best practices of other DMUs and an optimal combination of factors of production, adequate investment in, and adaption of new technologies relevant to modernization of tax offices and further training and education in the adaption and use of the new technology. The next step would then be to improve their scale efficiencies. About 75% of the tax offices are exhibiting increasing returns to scale. These tax offices should increase its scale operations through internal growth or consolidation in the sector.

Keywords: Technical Efficiency, Scale Efficiency, Data Envelopment Analysis (DEA), DMU’s, Tax Offices

CHAPTER ONE

INTRODUCTION BACKGROUND

Tax is a medium which country across the globe depend upon so as to carry out the mandate of their citizens. Tax administrations exist largely to ensure compliance with the tax laws, and the effectiveness with which tax agencies fulfill their mission has always been a high priority for governments. However, the administrative dimension of taxation has long been recognized by tax administrators, especially those working on tax policy in developing countries, (Goode 1981; Bird and de Jantsche, 1993), there has been little systematic analysis of this administrative dimension, at least by economists. The available, but often mainly unreliable, evidence from government budgetary information clearly indicates that the budget cost of collecting individual income, business income, and sales taxes is generally in excess of 1 percent of the revenues from these taxes, and can sometimes be substantially higher, Sandford (1995).

Unfortunately, there is little information on how “efficient” any tax administration may actually be in using administrative “inputs” (e.g., personnel, materials, information, laws, procedures) to generate “outputs” like tax revenues.

Recent world-wide fiscal trends of rising government deficits and mounting debt have added considerable pressure to the revenue collection agencies on at least two fronts. There is noticeable pressure to increase tax collections, which under current tax laws can only occur through increased enforcement. Simultaneously, the fiscal hurt is forcing cutbacks in resources allocated to the tax agencies. Administrative agencies are therefore being asked – or forced – to do more, and to do more with fewer resources. These developments also mean that inefficient agencies will need to take steps to increase the efficacy of their tax collection operations. While public spending efficiency has received a great deal of attention, tax collection efficiency has received considerably less notice, largely because the absence of comparable data across tax administrations has made the comparative analysis of tax offices impossible.

Governments have opted to implement new taxes and increased tax rates in the hope of finding a solution to increase tax revenue, whereas improving the efficacy and efficiency of the body charged with managing taxation might offset this need.

After the bilateral agreement which have been signed in 2011, all tax collecting offices of Addis Ababa city government are came under direct control of ERCA.

The reason for the merger is to improve efficiency and enhance quality in taxation works as a means of increasing revenues and customer satisfaction resulting from a more efficient tax administration system.

The tax collection trend in the Addis Ababa City Tax Authority jurisdiction shows an increasing trend. In 2011, it was collected 4.3 billion Birr. But after the merger with ERCA, in 2017, by showing 20% average increment, it is reached 23 Billion Birr and it increases the capacity of covering their spending budget from 51% to 75%. There is a potential to collect more than 40 Billion Birr per year in the city (Tax for Development newspaper, 9th year, no. 108, Nov, 2017).

The purpose of the paper is to address these issues. Specifically, it attempts to determine the relative efficiency of tax collection offices that are offices which deal with declarations, assessments, collection and doubtful tax demands of individuals and companies at the city government level.

STATEMENTS OF THE PROBLEMThe study aims to measure the relative efficiency of 12 Addis Ababa city government tax offices which were under the administration of ERCA. Tax administration should be effective in the sense of ensuring high compliance by taxpayers, and efficient in the sense that administrative costs are low relative to revenue collected.

Since income from taxation fluctuates according to the prevailing economic trends, there is little that a tax office can do to increase its tax base, beyond increasing its efficiency. Secondly, there have been protracted and bitter labour disputes on the issue of updating the managerial procedures of tax offices, particularly with regard to increasing the use of computers.

Thirdly (and to some extent related to the previous point), the sector suffers from chronic low levels of education among employees, bureaucracy and endemic petty corruption, due to antiquated employment policies. Fourthly, the appointment of tax office managers has, until recently, been prone to politically dominant interest groups and to their private agendas, Olson (1965), which has impeded modernization and development on the basis of economic, democratic and professional prerogatives and promotion on the basis of merit. Finally, there is a historically determined deficit of capital, which is one contemporary illustration of how and how far, the country and the city lags behind the nations of the industrialized world. The Addis Ababa City Government funds the costs of the tax offices. The allocation of public resources to each tax office is based on the budget of the city government for the inland revenues authority. Despite the follow-up inspection procedures carried out by the ERCA, auditing or feedback regarding the tax offices’ activities is still not enough, as can be inferred from the yearly reports presented by tax office. These reports focus mainly on the financial performance of the unit that is the collected amount of revenue. However, information on operational activity is not standardized and in the case of some offices, even neglected. This, in turn, implies that the City Government is, at best, under-informed as to the return on its policy. From this, it can be inferred that the tax offices are free to set their own, private agendas, bypassing the public objectives which they are assumed to pursue and which would be in common with the Government’s declared policy goals. Since a worldwide process of modernization of public administration is being carried out throughout State institutions, Keehley et al. (1997), the paper emphasizes the need to apply the DEA benchmark procedure to analyze the efficiency of the tax offices’ activities. This would be at least one step in the right direction in the attempt to reduce the lag behind other countries and cities tax offices.

1.3 RESEARCH OBJECTIVES

1.3.1 GENERAL OBJECTIVEThe main objective of the study is to assess the relative efficiency of tax collecting offices of the Addis Ababa City Government using DEA model.

1.3.2 SPECIFIC OBJECTIVES

To estimate relative technical efficiencies(TE) and scale efficiency (SE) of individual Decision-Making Units (DMUs)

To measure the magnitude of inefficiency behind this tax collecting offices

Ranking tax collecting offices based on their efficiency scores

RESEARCH QUESTIONSHow efficiently do tax offices utilize their resources and to their scale size?

How much inefficiency is existed in each tax offices?

Which tax office is relatively better efficient and inefficient?1.5 SIGNIFICANCE OF THE STUDYThe paper contributed to the literature in different important ways. To best of my knowledge, this is the first known attempt to estimate relative efficiency scores of tax offices by employing DEA model in the city. The measurement of efficiency the research employed were also different from other common measures, such as simple tax ratios (e.g., cost-to-revenue), often used to measure the efficiency with which tax revenues are collected. A major drawback in using these measures is that they fail to account for the fact that tax collection is a production process that uses multiple inputs to produce multiple outputs. Estimating DEA efficiency scores makes it possible for us to account for these inputs and outputs and also for the environmental factors that affect how the inputs are combined in the production process (James Alm ; Denvil Duncan, 2013).

The analysis provided information on suitable political and administrative measures that can lead to improvements in the general setup behind the tax offices. Finally, it become useful for further study to those whom want to undertake more research in this area.

1.6 SCOPE OF THE STUDY

As mentioned in the main objectives, the study assesses the relative efficiency (Technical and Scale only) of Inland revenue collecting offices of Addis Ababa City Administration. The study has covered 12 Addis Ababa city Administration tax collecting offices for the period 2015/16 only. And the variables like Total Number of Tax Payers, Office Rent Expense and Total Number of Employees was used as input and Direct Taxes and Indirect Taxes are taken as output variables. Other input and output variables of tax offices was not considered in this empirical study.

1.7 LIMITATIONS OF THE STUDYThe paper has two limitations: first, limitations related to the data set and second, those related to the DEA method.

With reference to the data set, the homogeneity of the tax offices used in the analysis is questionable, since the study had compared offices with different size (small to medium), may face different restrictions and therefore, might not be considered to be directly comparable. Furthermore, the data obtained was not as wide as could be expected. If a larger amount of data were available, an analysis using other corresponding methods such as the Tobit regression could be performed which would greatly improve the conclusions. In order for the conclusions to be generalized, the study would need to have a larger data set. Reducing the number of observations in DEA variables increases the likelihood that a given observation will be judged relatively efficient, Banker (1993).

Some of the limitations of the DEA model are the following: the DEA does not impose any functional form on the data, neither does it make distributional assumptions for the inefficiency term, nor does it make a prior distinction between the relative importance of any combination of inputs and outputs.

1.8 DESCRIPTION OF THE STUDY AREAThe study was conducted in twelve (12) of fourteen (14) Addis Ababa City Administration tax offices which are currently operating in the city. Their classification of tax collection responsibility is based on the physical geography of the sub cities which are classified for appropriateness of the city administration and the ability of taxpayers. The city has 10 micro taxpayer offices, 2 medium taxpayer offices and 2 project offices. The names of the offices are Bole Sub City Micro Taxpayers Branch Office, Arada Sub City Micro Taxpayers Branch Office, Yeka Sub City Micro Taxpayers Branch Office, Kirkos Sub City Micro Taxpayers Branch Office, Gulele Sub City Micro Taxpayers Branch Office, Akaki Kaliti Sub City Micro Taxpayers Branch Office, Nifas Silk Lafto Sub City Micro Taxpayers Branch Office, Kolfe Keranyo Sub City Micro Taxpayers Branch Office, Lideta Sub City Micro Taxpayers Branch Office, Addis Ketema Sub City Micro Taxpayers Branch Office. The two medium tax payer branches are Addis Ababa Number 1 Medium Taxpayers Office and Addis Ababa Number 2 Medium Taxpayers Office. Lastly Merkato Number 1 and Merkato Number 2 project offices are project offices which are established recently and responsible for collecting tax on the Merkato trade area which is said to be Africa’s biggest trading area. From these tax offices, Addis Ababa Number 2 Medium Taxpayers Office and Arada Sub City Micro Taxpayers Branch Offices were excluded from the data set since they are supposed to be outliers.

1.9 ORGANIZATION OF THE PAPER

The paper is organized as follows. The next chapter contains literature reviews with theoretical and empirical literatures of DEA measure of efficiency. Chapter three presents the method of the research work it has been conducted. Chapter four communicates with data analysis and presentation of findings. Final part of the paper summarizes presented results and discusses possible directions for future research work.

CHAPTER TWO

LITERATURE REVIEW

2.1 INTRODUCTION

2.2 THEORETICAL LITERATURE REVIEW

Afonso et al. (2006) highlighted the importance behind the rational use of resources available to the public sector and the need for high-caliber fiscal policies. Both these aspects are considered crucial for economic growth and stability as well as for individual well-being. In recent years various attempts have been made at measuring the degree of public expenditure efficiency via quantitative analysis methods, including composite indicators and non-parametric approaches. In particular, Tanzi and Schuknecht (1997, 2000) in trying to assess total public spending-associated benefits in 18 industrialized countries, used data on various socio-economic indicators, available for groups of countries. Afonso et al. (2006) refined this approach and defined composite indicators of public sector performance. Odeck (2005) illustrated the usefulness of non-parametric methods on exploring causes of productivity growth of public sector.

However, tax revenue is the key funding source behind public spending; therefore, taxation system efficiency is of pivotal importance to the longevity and well-being of any public-sector activity. Increasing the revenues obtained from taxation can be achieved only at progressively higher marginal costs (Afonso et al., 2006)

Efficiency can be simply defined as the ratio of output to input. More output per unit of input reflects relatively greater efficiency. If the greatest possible output per unit of input is achieved, a state of absolute or optimum efficiency has been achieved and it is not possible to become more efficient without new technology or other changes in the production process. Efficiency can be measured in different ways.

Technical efficiency of a tax office is a comparative measure of how well it processes inputs to achieve its outputs, as compared to its maximum potential for doing so, as represented by its production possibility frontier. Scale efficiency refers to the amount by which efficiency can be increased by moving to the most efficient scale size. Scale efficiency measures can be obtained for each firm by conducting both a CRS and a VRS DEA, and then decomposing the TE scores obtained from the CRS DEA in to two components, one due to scale inefficiency and one due to “pure” technical inefficiency (i.e. VRS TE).

If there is a difference in the CRS and VRS TE scores for a particular firm, then this indicates that the firm has scale inefficiency. If inputs and/or outputs are measured in dollars rather than physical units, the efficiency differences we observe can be due to price efficiency as well as scale and technical efficiency.

Allocative efficiency results from an efficient mix of inputs used to produce the mix of outputs. It refers to whether inputs, for a given level of output and set of input prices, are chosen to minimize the cost of production, assuming that the organization being examined is already fully technically efficient and global (economic) efficiency, which reflects the production of goods and services that afford the greatest benefits to society at the lowest possible social cost.

2.2.1 TYPES OF EFFICIENCY MEASUREMENT MODELSRegression and Stochastic frontier analysis have been the popular methods of measuring efficiency. Data Envelopment Analysis (DEA) is one of the latest additions to the bracket of these techniques. The econometric approach to the construction of frontiers and the estimation of efficiency relative to the constructed frontiers has similarities and differences with the mathematical programming approach. Both are analytically rigorous benchmarking exercises that exploit the distance functions introduced to measure efficiency relative to a frontier. However, the two approaches use different techniques to envelop data more or less tightly in different ways. At the risk of oversimplification, the differences between the two approaches boil down to two essential features.

• The econometric approach is stochastic. This enables it to attempt to distinguish the effects of noise from those of inefficiency, thereby providing the basis for statistical inference.

• The programming approach is nonparametric. This enables it to avoid confounding the effects of misspecification of the functional form (of both technology and inefficiency) with those of inefficiency.

A. Parametric Models (Econometric Models)

Econometric models can be categorized according to the type of data they use (cross-section or panel), the type of variables they use (quantities only, or quantities and prices), and the number of equations in the model. The most widely used model, the single equation cross-section model, and the panel data models. In both contexts the efficiency being estimated can be either technical or economic. And also, multiple equation models and shadow price models, which typically involve multiple equations.B. Non-Parametric Model (Data Envelopment Analysis)

Data envelopment analysis (DEA) is a mathematical programming-based technique to evaluate the relative performance of organizations. While the main applications have been in the evaluation of not-for-profit organizations, the technique can be successfully applied to other situations competing with other techniques as cost benefit analysis and multi criteria decision making as can be seen, for instance, in a recent study about the best choice for traffic planning, namely, the design and location of a highway in Memphis, Bougnol et al. (2005).

2.2.2 HISTORY AND APPLICATIONS OF DEADEA is a mathematical programming technique presented in 1978 by Charnes et al. (1978), although its roots may be found as early as 1957 in Farrel’s seminal work, Farrell (1957) or even to Debreu’s, which introduced in the early fifties the “coefficient of resource utilization”, Debreu (1951). It deserves special attention and also the work of the Dutch Nobel-prized Tjalling Koopmans and his “activity analysis concepts”, Koopmans (1951).

The DEA technique is usually introduced as a non-parametric one, but in fact it rests on the assumption of linearity, Chang and Guh (1991) and for the original constant returns to scale (CRS) models even in the more stringent assumption of proportionality. Its application has been focused mainly on the efficiency assessment of not for- profit organizations, since these cannot be evaluated on the basis of traditional economic and financial indicators used for commercial companies.

The first application of DEA was in the agriculture field; as a matter of fact, Farrell applied it to 1950 data of 48 states in the United States of America, considering 4 inputs and 2 outputs.

At that time, the DEA term was not yet created, so in fact the first time the term DEA and that technique was applied in the area of education, specifically in the analysis of Program Follow Through, conducted in the USA, in the late seventies, Rhodes (1978). Since then it has been used to assess efficiency in areas such as health, Wilson et al. (2012), county goals, Seiford and Zhu (2002), courts, Schneider (2005), universities, Bougnol et al. (2010) and many other not-for-profit sectors. Nowadays DEA can be seen to have spread to other fields such as transit, Chiu et al. (2011), mining, Chen et al. (2010), air transportation, Pestanae Dieke (2007) and even banking, Emrouznejad and Anouze (2010).

2.2.3 ADVANTAGES AND LIMITATIONS OF DEAThe main advantage of DEA is that it can readily incorporate multiple inputs and outputs to calculate technical efficiency. By identifying the “peers” for organizations that are not observed to be efficient, it provides a set of potential role models that an organization can look to, in the first instance, for ways of improving its operations. However, like any empirical technique, DEA is based on a number of simplifying assumptions that need to be acknowledged when interpreting the results of DEA studies.

Its main limitations include but not limited to the following: being a deterministic rather than statistical technique, DEA produces results that are particularly sensitive to measurement error. DEA only measures efficiency relative to best practice within the particular sample. Thus, it is not meaningful to compare the scores between two different studies. DEA scores are sensitive to input and output specification and the size of the sample. Despite these limitations, data envelopment analysis is a useful tool for examining the efficiency of government service providers. Just as these limitations must be recognized, so must the potential benefits of using DEA (in conjunction with other measures) be explored to increase our understanding of public sector performance and potential ways of improving it?

INPUT AND OUTPUT SLACKS

This is a value which shows the discrepancy in the constant or proportional change of input and output variables (Coelli, 2008). It also represents the amount of value for improvement in both input and output. Slacks only show the variable discrepancy between the constant output and input. It is only perceived that the value must be either increased or decreased (Cooper, Seiford, & Zhu, 2011). Efficient tax offices did not have input and output slack because they are expected to use the inputs to produce their output efficiently. Thus, no need of changing inputs and outputs for the efficient tax offices. But input or output slacks exist in inefficient tax offices.

To illustrate this, we consider a simple numerical example used in Zhu (2002) as shown in Table 1-3 where we have five DMUs representing five supply chain operations. Within a week, each DMU generates the same profit of $2,000 with a different combination of supply chain cost and response time.

Table 1-3. Supply Chain Operations Within a Week Inputs Output

DMU Cost ($100) Response time (days) Profit ($1,000)

1 1 5 2

2 2 2 2

3 4 1 2

4 6 1 2

5 4 4 2

Source Zhu (2002)

We now turn to the BCC model for which Figure 1-4 presents the five DMUs and the piecewise linear DEA frontier. DMUs 1, 2, 3, and 4 are on the frontier. If we adjoin the constraint ? = 1 to model (1.4) for DMU5, we get from the data of Table 1-3, =njj1?

Min ?

Subject to

1 ?1 + 2?2 +4?3 +6?4 +4?5 < 4?

5 ?1 + 2?2 +1?3 + 1?4 +4?5 < 4?

2 ?1 + 2?2 +2?3 +2?4 +2?5 > 2

?1 + ?2 +?3 +?4 +?5 = 1

?1, ?2, ?3,?4, ?5 > 0

This model has the unique optimal solution of = 0.5, = 1, and = 0 (j ? 2), indicating that DMU5 needs to reduce its cost and response time to the amounts used by DMU2 if it is to be efficient This example indicates that technical efficiency for DMU5 is achieved at DMU*?*2?*j?2 on the boundary.

Figure 1-4. Five Supply Chain Operations

Source: Zhu (2002).

Now, if we similarly use model (1.4) with = 1 for DMU4, we obtain = 1, = 1, and = 0 (j ? 4), indicating that DMU4 is on the frontier and is a boundary point. However, Figure 1-4 indicates that DMU4 can still reduce its response time by 2 days to achieve coincidence with DMU3 on the efficiency frontier. This input reduction is the input slack and the constraint with which it is associated is satisfied as a strict inequality in this solution. Hence, DMU4 is weakly efficient.

The nonzero slack can be found by using model (1.5). With the constraint ?nj= 1 ?j = 1adjoined and setting = 1 yields the following model,

Max S-1+S-2+S1

Subject to

1 ?1 + 2?2 +4?3 +6?4 +4?5 + = 6 = 6 ?1s*?

5 ?1 + 2?2 +1?3 + 1?4 +4?5 + = 1 = 1 ?2s*?

2 ?1 + 2?2 +2?3 +2?4 +2?5 – = 2 +1s

?1 + ?2 +?3 +?4 +?5 = 1

?1, ?2, ?3,?4, ?5, , , ?1s?2s+1s> 0

2.3 EMPIRICAL LITERATURE REVIEWAt the empirical level, DEA methods have been extensively applied for the estimation of the efficiency scores of individual production units (cross-section) in various areas such as: health services, education, public transportation, post offices, municipalities, banking, insurance, …

On the contrary, at the revenue side of the government sector, empirical studies on the productive efficiency of tax offices seem to be rather rare. Although efficiency of the tax administration is one of the four “canons” recommended by Adam Smith (1776) in his long-standing treatise on taxation. Moreover, one notices that economists are more interested in the empirical investigation of individual compliance costs of taxation rather than the measurement of the operational efficiency of existing tax administrations (see Sandford, Godwin and Hardwick (1989); Slemrod (1992).

Analysis of efficiency of tax offices is scarce, probably because of the non-disclosure policy of public entities such as tax offices everywhere including in Ethiopia. The study refers to the following papers: James Alm and Denvil Duncan (2013), Moesen and Persoons (2002), Gonzalez and Miles (2000), Ramón Fuentes (2014), Sang-Lyul Ryu and Seok-Young Lee (2013), Katharaki and Tsakas, (2010) and C.P Barros (2007).

James Alm and Denvil Duncan (2013) attempted to determine the relative efficiency of tax collection agencies of 30 OECD (Organization of Economic Co-operation and Development) countries for the period 2005 to 2009 using input-oriented variable-return-to-scale three stages DEA model and econometric model. They used salary and information technology administrative costs related to tax function as inputs and corporate income tax (CIT), personal income tax (PIT), and value-added tax (VAT) revenues separately, in total, and in various combinations as outputs. Their findings indicate that 12 of the 30 countries in the sample are relatively efficient at collecting any of the three types of tax revenues (personal income, corporate income, and value added taxes). Overall, the average efficiency scores indicate that countries should be able to collect their current level of revenues with approximately 10 to 13 percent less inputs.

Moesen and Persoons (2002) analyzed the efficiency of 289 regional tax offices dealing with personal income tax matters in Belgium for the fiscal year 1991, with Free Disposal Hull (FDH) and DEA taking labour as input and number of audited returns with a different complexity as output.

Gonzalez and Miles (2000) analyze the efficiency of 15 Spanish regional tax offices for the year 1995 with DEA, using a bootstrap technique. The single input used was the ratio of the number of tax inspectors to the total personnel; the two outputs used were the ratio, the number of actions performed by the tax office to total tax payers and the ratio of debt to gross added value. Their conclusions were that the average efficiency was 0.81 and that only third of the 15 offices analyzed were efficient.

Ramón Fuentes, (2014) analyzed how the productivity growth of the tax offices located in the province of Alicante (Spain) evolved between 2004 and 2006. The methodology employed was Output-oriented Data Envelopment Analysis (DEA) based Malmquist Indices to compute the levels of productivity and smoothed bootstrap to avoid the sensitivity of the results to sampling variations. The inputs were office area and number of employees and the outputs were number of tax returns and number of taxpayers. The finding shows the aggregated development of Malmquist productivity index and its components for the whole of the period 2004-2006, total average productivity increased by 5.73%, with a simultaneous improvement in all components.

Sang-Lyul Ryu and Seok-Young Lee (2013) Using data envelopment analysis discovers how the efficiency of collecting national tax in tax jurisdictions has been changed in Korea over the period 1998-2011 for six tax jurisdictions. The input variables taken were direct taxpayers, indirect taxpayers and real GDP (RGDP). Direct tax and other taxes were taken as output variables. They got the mean efficiency scores 0.62 for pooled sample of 84 observations, suggesting that there exists a significant level of waste in national tax collection activity. Furthermore, trend analysis indicates that the aggregate efficiency for tax jurisdictions has declined steadily over time since the currency crisis of 1997.

Katharaki and Tsakas, (2010) studied the technical and scale efficiency of 27 tax offices in Greece during the period 2001-2006 utilizing DEA and Window analysis. They used labor in each tax office, measured by the number of employees; the number of computers operating in each tax office; the number of natural persons (NP) paying taxes; and the number of legal entities (LE) paying taxes as inputs and incoming taxation funds related to natural persons and the incoming taxation funds related to legal entities as outputs. And find that ”scale size” and the structure of regional economy where tax offices operate are important factors affecting their efficiency.

C.P Barros (2007) estimates the technical and allocative efficiency of 41 tax offices in central and great Lisbon from 1999 to 2002, utilizing DEA. The inputs used were labour, measured by the number of workers; capital, measured by the rents paid for the premises and the tax population (population registered in the office’s files) and the outputs used were the total amount of personal income tax collected; value of corporate income tax collected; value of VAT, value of inheritance and donations taxes and value of other taxes and found the results are, at best, mixed, leading to conclude that the Government’s regulatory body, the Central Tax Authority (CTA), falls short of achieving its stated aims. DEA has been used in a number of public finance studies to assess the relative efficiency of public spending, Adam, Delis, and Kammas (2011) and taxation, Thanassoulis, Dyson, and Foster (1987).

RESEARCH GAPS

Various research works have been conducted in relation to the tax administration issue of Ethiopia. But little studies that focus on efficiency of tax offices are conducted. To the best of the researcher’s knowledge, no studies have been conducted so far to estimate the relative efficiency of tax offices using DEA model in Ethiopia. However, the researcher has got a study by Yifru (2016), on Efficiency of Private Commercial Banks in Ethiopia, using input-oriented CRS and VRS single stage DEA model. To fill the gap, the researcher makes a reference on studies which are relevant to tax offices efficiency conducted in Europe and Asia. The study tried to bring an answer for the question of relative efficiency of tax collecting offices under investigation.

2.5 CONCEPTUAL FRAMEWORKThe proposed model which explains the input and output variables used for the estimation of technical and scale efficiency of 12 tax collecting offices of Addis Ababa City Administration is as follows:

INPUTS

OUTPUTS

Total Number of Employees

Total Number of Tax Payers

Office

Rent

Expense

Indirect Taxes

Direct Taxes

Transformation

MEASURE OF EFFICIENCY

Technical Efficiency

Scale Efficiency

CHAPTER THREE

3. RESEARCH METHODS

3.1 INTRODUCTIONGenerally, the research method part includes the operational definition of variables, research design, model specification, research method, sample design such as: population, sampling frame, sampling unit, the sample size, sources of data. In addition, methods of data analysis and presentations are also part of these section.

The paper was concerned with the concept of technical and scale efficiency, given that analyzing other types of efficiency entails the need for awareness of market prices, or, where necessary, social cost, values that, in the case of the public sector, remain largely unknownCITATION bar l 1033 (barrilao-gonzalez).

3.2 OPERATIONAL DEFINITION OF VARIABLES The following variables are operationalized as follows:

Tax offices mean an Inland Revenue small ; medium size branch office which are established by Addis Ababa City Administration for collecting tax revenues.

Decision making units (DMUs) are entities responsible for converting inputs in to outputs, such as firms. In this study, DMU refer to 12 Addis Ababa city administration tax collecting offices.

Technical Efficiency: is comparative measure of how well it processes inputs to achieve its outputs, as compared to its maximum potential for doing so.

Pure Technical Efficiency: is the residual efficiency element that refers to resource conservation or output enhancement efficiency.

Scale Efficiency: is the ability of each DMU’s to operate as close to its most productive scale size as possible.

Ranking: giving grade and setting precedence for the efficiency scores of each DMU’s.

Total Number of Employees: the total number of permanent employees who had been working for each tax offices in the sample period.

Total Number of Taxpayers: all types of taxpayers who were registered and active in each DMU’s in the sample period excluding employee and students.

Office Rent expense: is an expense which were disbursed for office building rent of each DMU’s in the sample period

Direct Taxes: It includes all types of inland revenue direct taxes collected by each tax offices in the sample period and it does not include salary income tax.

Indirect Taxes: It includes all types of inland revenue indirect taxes collected by each tax offices in the sample period.

3.3 RESEARCH DESIGNTo estimate the technical and scale efficiency of the DMU’s, this analytical study was used a database from ERCA and AFECO for the year 2015/16, containing information on inputs and outputs. The study was employed DEA methodology, based on the article by Charnes et al. (1978), set up as a non-parametric, deterministic approach that enables us to obtain a measurement of relative efficiency between tax offices, in order to identify those that present optimal performance when compared with the remainder.

In many studies the analysts have tended to select input-oriented models because many DMU’s have particular orders to fill (e.g. electricity generation) and hence the input quantities appear to be the primary decision variables, although this argument may not be as strong in all industries. In some industries the DMUs may be given a fixed quantity of resources and asked to produce as much output as possible. In this case an output orientation would be more appropriate.

Several criteria can be applied to the selection of inputs and outputs. Usually, the criterion of available archival data is used. Next, the literature survey is a way to ensure the validity of the research and as such, another criterion to take into account. The last criterion for measurement selection is the professional opinion of managers in the area concerned and the realistic approach required when analyzing these statistics. For these study, it follows all four mentioned criteria.

In a general production function, labor, land and capital are included as inputs. Prior studies for example James Alm and Denvil Duncan (2013); Barros (2007); Katharaki and Tsakas, (2010); Sang-Lyul Ryu and Seok-Young Lee (2013); were used total amount of tax collected by regional tax office in tax jurisdictions as one of outputs.

James Alm and Denvil Duncan (2013) uses salary expense and information technology costs as input variables and tax revenues as output variable.

Barros (2007) takes number of tax population, rent expense and number of employee as input variables and direct and indirect taxes with different segregation as output variables.

Katharaki and Tsakas, (2010) used labour in each tax office, measured by the number of employees; the number of computers operating in each tax office; the number of natural persons (NP) paying taxes; and the number of legal entities (LE) paying taxes as inputs and incoming taxation funds related to natural persons and the incoming taxation funds related to legal entities as outputs.

Sang-Lyul Ryu and Seok-Young Lee (2013) Using data envelopment analysis taken input variables like direct taxpayers, indirect taxpayers and real GDP (RGDP). Direct tax and other taxes were taken as output variables.

Taking into account the existing literature and availability of data, the study includes the Total Number of Taxpayers, Office Rent Expense and Total Number of Employees as input and Direct Taxes and Indirect Taxes as output in estimating the efficiency for each observation.

3.4 MODEL SPECIFICATIONThe study was used single stage input-oriented CRS and VRS DEA approach. The linear programming formulation is employed to estimate efficiency scores with the input-oriented CRS model is presented below as follows.

Assume there are data on N inputs and M outputs for each of I firms. The input and output for DMUj be (x1j, x2j, x3j,……, xNj) and (q1j, q2j, q3j,….., qMj), respectively. The N×I input matrix, X, and the M×I output matrix, Q, represent the data for all I firms.

Where j=1, 2,…, I,

Input (x): N number of input items: xi where i=1…, N

Output (q): M number of output items: qr where r=1…, M

Thus, the input X and the output Q data matrixes can be arranged as;

X (input) = x11 x12 ……….x1I

x21 x22 ……….x2 I

x31 x32 ……….x3 I

xxx xxx ………xxx

xxx xxx ………xxx

xN1 xN2 ……….xN 12

Q (output) = q11 q12 ……….q1 I

q21 q22 ……….q2 I

qxx qxx ………qxx qxx qxx ……….qxx qM1 qM2 ……..qM12

So, in the case of this study, there are three inputs x1j, Total Number of Taxpayers (TNT), x2j, Office Rent Expense (ORE), Total Number of Employees and the two outputs q1j, Direct Taxes (DT), q2j, Indirect Taxes and the number of DMU’s (tax offices in our case) I are 12.

The input-output matrix represents the data of all DMU’s can be presented as follows:

X (Input) = TNT1 TNT2 TNT3 TNT4…………TNT12

ORE1 ORE2 ORE3 ORE4…………TNE12

TNE1 TNE2 TNE3 TNE4…………TNE12

Q (Output) = DT1 DT2 DT3 DT4………………DT12

IDT1 IDT2 IDT3 IDT4……………IDT12

Efficiency is measured as a ratio of all outputs to all inputs. For each DMU’s a measure of the ratio of all outputs over all inputs, such as u’qi/v’xi, where u is an M×1 vector of output weights and v is a N×1 vector of input weights. Then to select the optimal weights the mathematical programming problem is as follows:

Max u,v (u’qi/ v’xi), (1)

Subject to u’qj/ v’xj?1 J = 1,2,…..,I,

u,v ? 0

This involves finding values for (u) and (v) that maximize the ratio of DMUj, the DMU being evaluated, subject to the constraint that all efficiency measures must be less than or equal to one. Efficiency value takes the values between zero and one.The problem with this particular ratio formulation is that it has an infinite number of solutions. To avoid this, one can impose the constraint v’xi =1, which provides:Max µ, ? (µ’qi), (2)Subject to ?’xi =1, µ’qj-?’xj?0, j=1,2,….,I µ, ? ? 0,Where u’s and v’s are replaced by µ’s and ?’s. This change reflects the transformation, called the multiplier form of the linear programming problem. Using the duality in linear programming, one can derive an equivalent envelopment form of this problem:Min ?,? ?, (3)Subject to –qi + Q ? ? 0, ?xi – X ? ?0, ? ?0,Where ? is a scalar and ? is a I×1 vector of constants. Q is a vector of outputs of every DMU and X is the vector of all inputs of every DMU. In the envelopment model, the number of degrees of freedom will increase with the number of DMUs and decrease with the number of inputs and outputs. A rule of thumb that can provide guidance is as follows (Cooper, Seiford, ; Tone, 2000): n ? max {m × s, 3× (m+s)}; where n is number of DMUs, m is number of inputs and s is number of outputs. This pre-condition has been fulfilled by the analysis in this paper Gordana Savi?, A. D. et al. (2015).

The value of ? obtained will be the efficiency score for the ith DMU. It satisfies ? ? 1, with a value of 1 indicating a point on the frontier and hence a technically efficient firm, according to the Farrell (1957) definition. This linear programming problem will be solved I times, once for each firm in the sample. A value of ? is then obtained for each DMU’s.

The above approach takes in to account the CRS that is applicable when all DMU’s are operating at an optimal scale. In other situations, this is not the case. The scale of operations may have an impact on the outputs, creating “economies” or “diseconomies” of scale. The BCC model, developed by Banker et al. (1984), can deal with variable returns to scale, Jorge Santos et al. (2013). Banker, Charnes and Cooper (1984) suggested an extension of the CRS DEA model to account for VRS. The use of CRS specification when not all DMU’s are operating at the optimal scale, will result in measures of TE which are confounded by scale efficiencies (SE). the use of VRS specification will permit the calculation of TE devoid of these SE effects. The CRS can be easily modified to account for VRS by adding the convexity constraint ?’ ? =1 to equation (4) to provide:

Min ?, ? ?, (4)Subject to –qi + Q? ? 0, ?xi – X ? ? 0, ?’ ? =1, ? ?1,Where ?’ is an I×1 vector of ones. This condition ensures that an inefficient tax office is “benchmarked” against similar size tax offices. As a result, VRS method envelops the data more closely than CRS method. So that VRS efficiency scores are greater than or equal to CRS efficiency scores.3.5 RESEARCH METHODSIn order to achieve the main research objectives quantitative research method is used. DEA is applied to measure the relative efficiency of decision making units. DEA has been used in a number of public finance studies to assess the relative efficiency of public spending, Adam, Delis, and Kammas (2011) and taxation, Thanassoulis, Dyson, and Foster (1987); Barros (2007); Katharaki and Tsakas (2010).

The output produced by tax agencies can be measured in several ways. One could examine the rate at which contested cases are resolved, Barros (2007), the number of actions that are taken against delinquent accounts, Katharaki and Tsakas (2010), or the number of returns that are audited, Moesen and Persoon (2002). An ideal analysis would also distinguish between revenues collected via voluntary payments and revenues collected through explicit enforcement activities. Unfortunately, it was hardly easy to have enough information on the sources of revenues to account for this distinction. As a result, the output measures focus exclusively on tax revenues measured separately as Direct Tax and Indirect Tax. We take these as the appropriate measures of output since the core objective of tax agencies is to collect revenues (James Alm ; Denvil Duncan, 2013). The efficiency scores were obtained by DEAP statistical software and, descriptive statistics was run by EVIEWS 8 software packages.

3.6 SAMPLE DESIGNA sample design is a definite plan for obtaining a sample from a given population. The sample was included 12 tax collecting offices of Addis Ababa city administration, for the period 2015/16. It is possible to measure efficiency using cross sectional data in the case of DEA. Shwu-Huei Huang, Ming-Miin Yu et al employed DEA to evaluate operating efficiency in departments of 20 Taiwanese local tax offices for 2013, Hoque and Rayhan used DEA to measure efficiency of 24 banks in Bangladesh for the year 2010. Similarly, Yifru Yirdaw used DEA to measure efficiency of 17 private commercial banks in Ethiopia.

A. POPULATION OR UNIVERSEThe total number of small and medium sized tax collecting offices administered by Addis Ababa City Government is 14. So, the research was taken all tax offices of small and medium size except two tax offices which were supposed to be outliers.

B. SAMPLING FRAMEIt is the list of populations and which consist of all categories of sampling units or units of analysis. Hence the sampling frame in this study was the 14 tax collecting offices of Addis Ababa City Administration.

C. SAMPLING UNIT/UNIT OF ANALYSISThe sampling units in the study are the tax collecting offices of Addis Ababa City Administration. D.SAMPLING TECHNIQUE

E. SAMPLE SIZEThe study was based on the efficiency evaluation of 12 Addis Ababa City Administration tax offices located here in Addis Ababa. The sample used in the analysis represents characteristically similar tax offices in the city administration.

3.7 SOURCES OF DATAAll data required for the study were collected from secondary sources. It was collected from published and unpublished documents, plans, books, and other related materials collected from each tax offices, AAFECB and ERCA. CHAPTER FOUR

RESULT AND DISCUSSION

INTRODUCTION

This chapter presents the results of the DEA efficiency estimation of each tax offices.

Of the 14 existing tax offices, 12 were taken into consideration for this study, excluding the two tax offices of Arada sub city small tax payers branch office and Addis Ababa number two (2) medium taxpayers branch office, which considered to be outliers, and whereby their data is not homogeneous and therefore not subject to extrapolation to the remaining offices. Indeed, their inclusion would have undermined the coherence of the results to a considerable extent. The DEA models run in this study is under the assumption of input minimization (also known as input orientation).

Descriptive Statistics

It is necessary to start the analysis with descriptive statistics so that we can get information about the inputs, outputs and other variables status.

The descriptive statistics for variables used to calculate the DEA efficiency estimates and other variables using EVIEWS 8 package are as shown in table 1.

Table 1: Descriptive statistics for inputs, outputs and other variables of Addis Ababa tax offices

Variables Name ObsMean Std. Dev Min Max Sum

Total Number of Taxpayers 12 29544.17

17110.54 4965 60482 354530

Office Rent Expense 12 9363434

5271047 2554092 20516058 112361210

Total Number of Employees 12 322.5

49.95907 221 398 3870

Direct Taxes 12 324.6658

206.6814 129.4200760.483895.99

Indirect Taxes 12 285.3025

214.5336 95.47 838.693423.63

Establishment year 12 2010.417

0.900337 2010 2013 –

Service year 12 5.583333

0.900337 3 6 67

Enterprise Taxpayers 12 1620

1107.586 72 3185 19440

Individual Taxpayers 12 27924.17

16224.24 4893 57646 335090

With regard to tax collection inputs, on average, 29,544 active taxpayers from all tax offices, with a maximum of 60,482 in Nifassilk Lafto sub city and a minimum of 4965 in Addis Ababa Medium no.1 are existed in discharging their tax responsibilities. From these taxpayers 19,440 are companies registered as enterprise taxpayers and 335,090 are individual company taxpayers. On average, 9363434 Birr office rent expense is disbursed with a maximum of Birr 20,516,058 in Merkato no.1 branch office and a minimum of Birr 2,554,092 Merkato no.2 branch office. On average, 323 persons were allocated as employee, with a maximum of 398 in Bole sub city and a minimum of 221 in Lideta sub city. Similarly, in relation to outputs on average, Birr 324.6658 million direct taxes were collected, with a maximum of Birr 760.48 Million in Addis Ababa Medium no.1 and a minimum of Birr 129.4200 million in Lideta.

On average, Birr 285.3025 million indirect taxes were collected with a maximum of Birr 838.69 million in Addis Ababa Medium no.1 and a minimum of Birr 95.47 million in Lideta sub city. when we see other variables, the mean establishment year is 2010 with a maximum of 2013 in AAM1 and a minimum of 2010 in ADK, AKK, BOL, GUL, KIR, KOK, LID, NIL ;YEK. On average the 5.58 service year is served with a maximum of 6 years in ADK, AKK, BOL, GUL, KIR, KOK, LID, NIL ;YEK and a minimum of 3 years in AAM1. On average, 1620 enterprise taxpayers were operated with a maximum of 3185 in Yeka sub city and a minimum of 72 in Addis Ababa medium no.1. and also, on average, 27924 individual taxpayers were operated with a maximum of 57646 in Nifassilk Lafto sub city and a minimum of 4893 in Addis Ababa medium no.1.Generally, tax offices were found to be hardly different in those inputs they procure and outputs they produce.

In table 2, for the purpose of comparison of the characteristics of the variables in each tax offices are listed with a ratio. It is clear that the mean value of the employee per taxpayer ratio is 0.01697 and its variance is lower than the mean and equal to 0.00023. The mean value of the ratio, taxpayer/direct tax is 0.0001166 and its variance is lower than the mean and equal to 4.17e-09. The mean value of the ratio, rent/direct tax is 0.0366547 and its variance is lower than the mean and equal to 0.0007018. The mean value of the ratio, employee/direct tax is 1.28e-06 and its variance is lower than the mean and equal to 3.07e-13. The mean value of the ratio, taxpayer/indirect tax is 0.0001513 and its variance is lower than the mean and equal to 0. 8.33e-09. The mean value of the ratio, rent/direct tax is 0.0456616 and its variance is lower than the mean and equal to 0.0011192. The mean value of the ratio, taxpayer/indirect tax is 1.58e-06 and its variance is lower than the mean and equal to 5.17e-13. Therefore, we can proclaim that the tax offices analyzed have a degree of homogeneity which is adequate for the DEA analysis.

Table 2: Comparison of the characteristics of the variables in each tax offices

No. Tax Offices Employee

Per taxpayer Taxpayer per direct tax Rent per direct tax Employee per direct tax Taxpayers per indirect tax Rent per indirect tax Employee per indirect tax

1 AAM1 .0588117 6.53e-06 .0185639 3.84e-07 5.92e-06 .0168327 3.48e-07

2 MR1 .0294642 .0000442 .0770701 1.30e-06 .0000449 .0782847 1.32e-06

3 MR2 .0272385 .0000425 .008338 1.16e-06 .0000329 .0064538 8.97e-07

4 ADIK .0132448 .0001734 .0226918 2.30e-06 .0001815 .0237631 2.40e-06

5 AKK .0089289 .000201 .0484527 1.79e-06 .0002606 .0628204 2.33e-06

6 BOL .0074925 .0000765 .0158899 5.73e-07 .0001041 .0216331 7.80e-07

7 GUL .0146091 .0001271 .092787 1.86e-06 .0001674 .1222394 2.45e-06

8 KIR .0131316 .0000748 .0333059 9.83e-07 .0000932 .0414701 1.22e-06

9 KOK .0071045 .000169 .0284526 1.20e-06 .0002454 .0413104 1.74e-06

10 LID .0097002 .000176 .0542071 1.71e-06 .0002386 .0734836 2.31e-06

11 NIF .0055884 .0001604 .0128895 8.96e-07 .0002062 .0165712 1.15e-06

12 YEK .0082935 .0001482 .0272082 1.23e-06 .0002346 .0430773 1.95e-06

Mean .0169673 .0001166 .0366547 1.28e-06.0001513 .0456616 1.58e-06Variance .0002324 4.17e-09 .00070183.07e-13. 8.33e-09.0011192 5.17e-13

INPUT OUTPUT RELATIONSHIP

All the inputs taken in this study such as Total Number of Taxpayers, Office Rent Expense and Total Number of Employees have positive relationship with the output direct taxes, and except Total Number of Taxpayers, all inputs have positive relationship with the output indirect taxes.

The following graphs generated using EVIEWS 8 show the input-output relationship of each input and output variables under consideration.

COMPARATIVE ANALYSIS OF TAX OFFICES EFFICIENCY IN CRS AND VRS MODEL

The DEA efficiency scores in this study are estimated using the computer program, DEAP, version 2.1, developed by Tim Coelli, The University of Queensland.

Table 3 reveals the results of the input-oriented DEA analysis in CRS and VRS scores, scale efficiency and the status of returns to scale of 14 DMUs (by including the two DMUs i.e. AAM2 and ARA which were supposed to be outliers), helps us to compare with the 12 DMUs efficiency scores that were run by excluding the two outliers.

Table 3 CRS, VRS and Scale Efficiency results of fourteen DMUs

No. DMUs Code DEA CRS model

Overall Technical Efficiency DEA VRS model

Pure Technical Efficiency Scale Efficiency Returns to Scale

1 AAM1 0.823 1.000 0.823 IRS

2 AAM2 1.000 1.000 1.000 CRS

3 MR1 0.163 0.791 0.206 IRS

4 MR2 0.652 1.000 0.652 IRS

5 ADIK 0.177 0.938 0.189 IRS

6 AKK 0.081 0.946 0.086 IRS

7 ARA 1.000 1.000 1.000 CRS

8 BOL 0.248 0.650 0.381 IRS

9 GUL 0.084 0.836 0.101 IRS

10 KIR 0.155 0.685 0.226 IRS

11 KOK 0.138 0.802 0.173 IRS

12 LID 0.079 1.000 0.079 IRS

13 NIL 0.306 0.915 0.334 IRS

14 YEK 0.145 0.780 0.186 IRS

mean 0.361 0.882 0.388 As we see the efficiency results from Table 3, the mean overall technical efficiency (CRS), pure technical efficiency (VRS) and scale efficiency scores of these tax offices are 0.361, 0.882 and 0.388 respectively.

If we take the CRS technical efficiency, only AAM2, and ARA tax offices are efficient with a score of 1.000, the rest twelve (12) DMUs are inefficient. And unfortunately, ten (10) tax offices efficiency scores are below the average 0.36. whereas the VRS assumption taken in to consideration, five (5) of the DMUs are technically efficient with a score of 1.000, namely AAM1, AAM2, MR2, ARA and LID. The rest nine (9) are inefficient. And six (6) tax offices efficiency scores are below the average 0.882. Moreover, only AM2 and ARA tax offices are scale efficient whereas the rest are scale inefficient with increasing returns to scale nature. Ten (10) of the fourteen (14) DMUs are operating below the average 0.388 scale efficiency.

Surprisingly only AAM2 and ARA (which assumed to be outliers and excluded from the model) are efficient in CRS, VRS and SE. Though, AAM1 and LID is technically inefficient at CRS and their scale is inefficient, they turn to efficient at VRS assumption.

The least technically inefficient DMU is LID with a score of 0.079 and the highest technically inefficient DMU is AAM1 scoring 0.823, under CRS assumption. When we assume VRS technology, the least technically inefficient DMU is BOL with a score of 0.650 and the highest technically inefficient DMU is AKK scoring 0.946. If we see SE, the least inefficient DMU is LID with a score of 0.079 and the highest inefficient DMU is AAM1 with a score of 0.823.CRS EFFICIENCY DISCUSSION OF THE TWELVE DMUs

Table 4 presents the results of the input-oriented DEA analysis of 12 DMUs, under the assumption of CRS. In contrast to Table 3 (36.1%), this DEA scores makes increase the Overall Technical Efficiency average to 59.3%. Of the 12 DMUs, 8 DMUs are below the average efficiency score of 0.593, under CRS assumption. The technical inefficiency scores presented in table 1.3 indicates the potential for DMU’s to decrease their usage of inputs while maintaining outputs constant. This average efficiency score suggests that the DMUs input can be minimized by 40.75%, given the outputs.

Three of the twelve (25%) of tax offices, namely A.A medium no. 1, Merkato no.2 and Bole sub city tax payers branch offices are fully efficient with a score of 1.000. Whereas, nine (75%) of tax offices are inefficient and have efficiency scores ranging from the least technically inefficient for example 0.207 by Gulele sub city to the highest technically inefficient DMU Nifassilk Lafto sub city with a scores of 0.952 in CRS DEA model.

Table 4: Efficiency Results of twelve DMUs under CRS Frontier

No. DMUs Code DEA CRS model

Technical Efficiency

1 AAM1 1.000

2 MR1 0.295

3 MR2 1.000

4 ADK 0.445

5 AKK 0.326

6 BOL 1.000

7 GUL 0.207

8 KIR 0.509

9 KOK 0.530

10 LID 0.306

11 NIL 0.952

12 YEK 0.541

mean 0.593

SUMMARY OF SLACKS

This is a value which shows the discrepancy in the constant or proportional change of input and output variables (Coelli, 2008). It also represents the amount of value for improvement in both input and output. Slacks only show the variable discrepancy between the constant output and input. It is only perceived that the value must be either increased or decreased (Cooper, Seiford, & Zhu, 2011). Efficient tax offices did not have input and output slack because they are expected to use the inputs to produce their output efficiently. Thus, no need of changing inputs and outputs for the three (3) efficient tax offices. But input or output slacks exist in inefficient tax offices. Among inefficient tax offices, six (6) tax offices i.e. MR1, ADIK, GUL, KOK, NIL & YEK have input slacks except for total number of employees and, all inefficient DMUs i.e. MR1, ADIK, AKK, GUL, KIR, KOK, LID, NIL & YEK have slacks for indirect taxes. But MR1, ADIK, GUL, KOK, NIL & YEK showed both input and output slacks. There is no DMU is existed having slack for direct taxes.

From Appendix A2, it is understood that DMU MR2 should decrease the number of taxpayers by 1,731 and 1,101,510 Birr rent expense to become efficient. Similarly, DMU ADIK can decrease their input of number of taxpayers by 4,129 to become efficient. DMU GUL also should decrease its input usage of number of taxpayers by 3,118 and rent expense Birr 99,226 to become efficient. DMUs KOK, NIL and YEK need to reduce their number of taxpayers by 4,461, 34,884 and 2,465 respectively to become efficient.

SUMMARY OF PEERS

The term Peer or Reference set indicates an inefficient firm to follow or as a reference (Coelli, 2008). It is a point that an inefficient DMUs targets to move from the Farrell efficient point to Pareto-Koopman efficient point based on the two-stage DEA solution or Pareto optimal solution (Yong-bae Ji and Choonjoo Lee). From table 5 here below, the 2ND and 7th inefficient DMUs such as MR1 and GUL can follow the input-output trend of AAM1. Like this DMU ADIK, KOK, NIL can follow the input-output trend of any of BOL or MR2. DMU AKK and LID can follow the input-output trend of any of MR2, BOL or AAM1. Similarly, DMU KIR can follow the input-output trends of any of BOL, AAM1, or MR2 and, DMU YEK can follow the input-output trends of any of MR2 or BOL. But the efficient tax offices don’t have to follow any tax office and are the peer of themselves only. Thus, it is necessary to rank the tax offices according to the Efficiency Scores and the number of Peer Counts.

Table 5: Rank of Tax Offices under CRS-Input Orientation Frontier

No. DMUs Code CRS

Technical Efficiency Peers Peer Weights Peer Counts Rank

1 AAM1 1.000 AAM1 1.000 5 3

2 MR1 0.295 AAM1 0.350 0 11

3 MR2 1.000 MR2 1.000 7 1

4 ADIK 0.445 BOL, MR2 0.052 0.387 0 8

5 AKK 0.326 MR2, BOL, AAM1 0.028, 0.164, 0.024 0 9

6 BOL 1.000 BOL 1.000 7 2

7 GUL 0.207 AAM1 0.208 0 12

8 KIR 0.509 BOL, AAM1, MR2 0.226, 0.254, 0.055 0 7

9 KOK 0.530 BOL, MR2 0.342, 0.094 0 6

10 LID 0.306 MR2, BOL, AAM1 0.007, 0.125, 0.053 0 10

11 NIL 0.952 BOL, MR2 0.283, 0.590 0 4

12 YEK 0.541 MR2, BOL 0.140, 0.334 0 5

Furthermore, peer weight is a number of variables that can be followed for the respective reference firm (Cooper et al., 2011). For the 2nd DMU i.e. MR1, since it has peers value for DMU1 (AAM1). Thus, it can follow 35% of the number of taxpayers, rent expense and number of employee of AAM1 to become efficient. For the 4th, DMU i.e. ADIK, since it has peers value for DMU6 ; 3 (BOL ; MR2). Thus, it can follow 5.2%, 38.7% of the number of taxpayers, rent expense and number of employee of BOL and MR2 respectively to become efficient. For the 5th, DMU i.e. AKK, since it has peers value for DMU3, 6 ; 1 (MR2, BOL ; AAM1). Thus, it can follow 2.8%, 16.4% and 2.4% of the number of taxpayers, rent expense and number of employee of MR2, BOL ; AAM1 respectively to become efficient. For the 7th DMU i.e. GUL, since it has peers value for DMU1 (AAM1). Thus, it can follow 20.8% of the number of taxpayers, rent expense and number of employee of AAM1. For the 8th DMU i.e. KIR, since it has peers value for DMU 6, 1 ; 3 (BOL, AAM1 ; MR2). Thus, it can follow 22.6%, 25.4% and 5.5% of the number of taxpayers, rent expense and number of employee of BOL, AAM1 and MR2 respectively to become efficient. For the 9th, DMU i.e. KOK, since it has peers value for DMU6 ; 3 (BOL ; MR2). Thus, it can follow 34.2% and 9.4% of the number of taxpayers, rent expense and number of employee of BOL ; MR2 respectively to become efficient. For the 10th, DMU i.e. LID, since it has peers value for DMU3, 6 ; 1 (MR2, BOL ; AAM1). Thus, it can follow 7%, 12.5%, ; 5.3% of the number of taxpayers, rent expense and number of employee of MR2, BOL ; AAM1 respectively to become efficient. For the 11th, DMU i.e. NIL, since it has peers value for DMU6 ; 3 (BOL ; MR2). Thus, it can follow 28.3% and 59% of the number of taxpayers, rent expense and number of employee of BOL ; MR2 respectively to become efficient. For the 12th, DMU i.e. YEK, since it has peers value for DMU3 ; 6 (MR2 ; BOL). Thus, it can follow 14% ; 33.4% of the number of taxpayers, rent expense and number of employee of MR2 ; BOL respectively to become efficient.

Finally, peer count shows the number of times a peer firm is being used as the reference unit. The 1st DMU i.e. AAM1 has been used as a peer 5 times, while DMU 3 ; 6 i.e. is MR2 ; BOL is used as a peer 7 time each. Therefore, DMUs MR2 ; BOL is the most efficient DMUs followed by AAM1 and are appropriate examples to share for other inefficient DMUs.

SUMMARY OF INPUT TARGETS

Target values are different from slack values. Slacks show discrepancies while targets show predicted value. This shows the model that is to be followed for the preferred and perceived inputs, in our case, total number of taxpayers, office rent expense and total number of employees that will help in making tax offices efficient. However, these inputs i.e. the number of taxpayers, rent expense and number of employee (input targets) remain the same for efficient organizations. Appendix A5 shows the input target for nine (9) inefficient DMUs. DMU2 i.e.MR1; for constant output of direct and indirect taxes, it must have an efficiently proportional total number of taxpayer of 1738, office rent expense of Birr 4,941,698 and total number of employee of 102. So, MR1 can increase the numbers of taxpayers by 1738, financials of rent expense by 4,941,698 and number of employee by 102. Likewise, for DMU4 i.e. ADIK; for constant output of direct and indirect taxes, it must have an efficiently proportional number of taxpayer of 7802, rent expense of Birr 1,561,730 and number of employee of 158. So, ADIK can increase the numbers of taxpayers by 7802, financials of rent expense by 1,561,730 and number of employee by 158. For DMU5 i.e. AKK; for constant output of direct and indirect taxes, it must have an efficiently proportional number of taxpayer of 9188, rent expense of Birr 2,215,385 and number of employee of 82. So, AKK can increase the numbers of taxpayers by 9188, financials of rent expense by 2,215,385 and number of employee by 82. For DMU7 i.e. GUL; for constant output of direct and indirect taxes, it must have an efficiently proportional number of taxpayer of 1031, rent expense of Birr 2,930,119 and number of employee of 61. So, GUL can increase the numbers of taxpayers by 1031, financials of rent expense by 2,930,119 and number of employee by 61.

For DMU8 i.e. KIR; for constant output of direct and indirect taxes, it must have an efficiently proportional number of taxpayer of 13992, rent expense of Birr 6,226,409 and number of employee of 184. So, KIR can increase the numbers of taxpayers by 13992, financials of rent expense by 6,226,409 and number of employee by 184.For DMU9 i.e. KOK; for constant output of direct and indirect taxes, it must have an efficiently proportional number of taxpayer of 19403, rent expense of Birr 4,018,067 and number of employee of 170. So, KOK can increase the numbers of taxpayers by 19403, financials of rent expense by 4,018,067and number of employee by 170.

For DMU10 i.e. LID; for constant output of direct and indirect taxes, it must have an efficiently proportional number of taxpayer of 6979, rent expense of Birr 2,149,159 and number of employee of 68. So, LID can increase the numbers of taxpayers by 6979, financials of rent expense by 2,149,159 and number of employee by 68.For DMU11 i.e. NIL; for constant output of direct and indirect taxes, it must have an efficiently proportional number of taxpayer of 22704, rent expense of Birr 4,627,493 and number of employee of 322. So, NIL can increase the numbers of taxpayers by 22704, financials of rent expense by 4,627,493 and number of employee by 322.

For DMU12 i.e. YEK; for constant output of direct and indirect taxes, it must have an efficiently proportional number of taxpayer of 19568, rent expense of Birr 4,045,249 and number of employee of 183. So, YEK can increase the numbers of taxpayers by 19568, financials of rent expense by 4,045,249 and number of employee by 183.

RADIAL MOVEMENT, SLACK MOVEMENT, AND PROJECTED VALUE

Radial Movement gives the value for input-output variables for improvement (Coelli, 2008). The Radial Movement shows the adjusted proportionality of input and output variables (Coelli, 2008). Projected value shows the proportionality of input and output for every radial and slack movement.

The radial values for the DMU2 i.e. MR1 are -8,307 for Number of Taxpayers, Rent Expense is Birr -14,472,850.000 and -245 for Number of Employees. The projected value of input for the MR1 against each output is 1,737.959 (Number of Taxpayers), Rent Expense is Birr 4,941,697.693 and 102.212 (Number of Employee).

MR1 can adopt a Direct Tax of 266.200 million and Indirect Tax of 262.070 million output values to achieve efficiency. The values are to be decreased by -8307 for Number of Taxpayers, Birr -14,472,850.000 for Rent Expense and -245 for Number of Employee.

The radial values for the DMU4 i.e. ADIK are -14871.850 for Number of Taxpayers, Rent Expense is Birr -1946653.240 and -196.974 for Number of Employees. The projected value of input for the ADIK against each output is 7802.162 (Number of Taxpayers), Rent Expense is Birr 1561729.760 and 158.026 (Number of Employee).

ADIK can adopt a Direct Tax of 154.610 million and Indirect Tax of 147.640 million output values to achieve efficiency. The values are to be decreased by -14871.850 for Number of Taxpayers, Birr -1946653.240 for Rent Expense and -196.974 for Number of Employee. Similarly, the highest inefficient DMU is NIL and the radial values for the DMU are -2893.516 for Number of Taxpayers, Rent Expense is Birr -232506.963 and -16.170 for Number of Employees. The projected value of input for the NIL against each output is 22704.091 (Number of Taxpayers), Rent Expense is Birr 4627493.037 and 321.830 (Number of Employee).

NIL can adopt a Direct Tax of 377.050 million and Indirect Tax of 293.280 million output values to achieve efficiency. The values are to be decreased by -2893.516 for Number of Taxpayers, Birr -232506.963 for Rent Expense and -16.170 for Number of Employee. Likewise, if we see the least inefficient DMU i.e. GUL, its radial values are -15907.522 for Number of Taxpayers, Rent Expense is Birr -11616152.807 and -232.394 for Number of Employees. The projected value of input for the GUL against each output is 1030.501 (Number of Taxpayers), Rent Expense is Birr 2930118.572 and 60.606 (Number of Employee).

GUL can adopt a Direct Tax of 157.840 million and Indirect Tax of 119.810 million output values to achieve efficiency. The values are to be decreased by -15907.522 for Number of Taxpayers, Birr -11616152.807 for Rent Expense and -232.394 for Number of Employee. Similar interpretations can be made for the remaining five inefficient DMUs (see Appendix B1, B2, B3, B4, B5 ; B6).

VARIABLE RETURNS TO SCALE IN DEA AND SUMMARY OF EFFICIENCY AND SLACKS

Variable returns to scale (VRS) is a type of frontier scale used in DEA. It helps to estimate efficiencies whether an increase or decrease in input or outputs does not result in a proportional change in the outputs or inputs respectively (Cooper, Seiford, ; Zhu, 2011). This method includes both increasing and decreasing returns to scale. Hence, VRS may exhibit increasing, constant and decreasing returns to scale when working in Data Envelopment Analysis Program (DEAP).

Before discussing the interpretations and findings, look into the previous interpretations, where the procedure to extract, apply and execution of data to perform the analysis is discussed. However, here this section includes interpretations and findings and how VRS differs from CRS.

Table 6: Efficiency Results of twelve DMUs under VRS Frontier

No. DMUs Code DEA CRS model

Overall Technical Efficiency DEA VRS model

Pure Technical Efficiency Scale Efficiency Returns to Scale

1 AAM1 1.000 1.000 1.000 CRS

2 MR1 0.295 0.791 0.372 IRS

3 MR2 1.000 1.000 1.000 CRS

4 ADK 0.445 0.938 0.475 IRS

5 AKK 0.326 0.946 0.344 IRS

6 BOL 1.000 1.000 1.000 CRS

7 GUL 0.207 0.836 0.247 IRS

8 KIR 0.509 0.726 0.701 IRS

9 KOK 0.530 0.854 0.620 IRS

10 LID 0.306 1.000 0.306 IRS

11 NIL 0.952 0.996 0.956 IRS

12 YEK 0.541 0.830 0.651 IRS

mean 0.593 0.910 0.639 Table 6 above shows efficiency summary for input oriented VRS DEA. The above table shows the difference in technical efficiency between CRS and VRS frontier. The overall Technical Efficiency is broken down into Pure Technical Efficiency specified by the VRS DEA score (average = 91%) and Scale Efficiency (average = 63.9%). In contrast to CRS model, the VRS model placed four DMUs (33.33%) as efficient, namely A.A medium no. 1, Merkato no.2, Bole sub city and Lideta sub city tax payers branch offices but the remaining eight DMUs (66.66%) namely MR1, ADK, AKK, GUL, KIR, KOK, NIL and YEK are remained inefficient. The VRS DEA model made Lideta sub city tax payers branch office technically efficient in opposite to CRS DEA model. Therefore, LID attains pure technical efficiency but remained scale inefficient with increasing returns to scale nature.

The least technically inefficient DMU is KIR with a score of 0.726 and the highest technically inefficient DMU is NIL scoring 0.996, under VRS assumption.

The efficiency scores for the inefficient DMU’s indicates the presence of and extent of inefficiency of inputs in the production process. For example, on average the tax offices in Addis Ababa tax jurisdiction have a scope for reduction of inputs by 9%.

Furthermore, the scale efficiency is the “unit where the size of operations is optimal so that any modifications on its size will render the unit less efficient” (Kao & Liu, 2011, p.225). Scale efficiency can be achieved by dividing the total efficiency by the technical efficiency. Moreover, scale efficiency shows whether the returns are increasing or decreasing by identifying the tax offices by increasing returns to scale and decreasing returns to scale. If output increases by less than proportional change in inputs, there is decreasing returns to scale. Similarly, increasing returns to scale occurs when the output increases by a larger proportion than the increase in inputs during the production process (Banker et al., 2004, p.346).

Thereby, it can be interpreted that the direct and indirect taxes of the offices have increased by a larger proportion of the given amount of number of tax payers, rent expense and number of employee, and thus they need to increase its number of taxpayers, rent expense and number of employee to become more efficient.

Moreover, Kao & Liu, (2011) says that scale efficiency value is also given by:

Scale Efficiency = CRS TE/VRS TE

Where:

CRS = constant returns to scale

VRS = variable returns to scale

TE = Technical efficiency

The average Scale Efficiency of the tax offices was 63.9% and six (50%) of the twelve DMUs are below this average SE score. Of the twelve DMUs, three (25%) of DMUs namely AAM1, MR2 and BOL are efficient in both technical efficiency measurements and scale efficient and have six peer counts except BOL which have zero peer counts in VRS. Such seldom appearance of efficient DMU in the reference set of inefficient DMU is likely to possess a very uncommon input/output mix. Thus, it is not suitable example to share for other inefficient DMUs. The rest nine (75%) of DMUs namely MR1, ADK, AKK, GUL, KIR, KOK, LID, NIL and YEK are scale inefficient suggesting that they are not operating at an optimal scale. Of this nine-scale inefficient DMUs, only one DMU i.e. Lideta sub city is technically efficient at VRS and have eight peer counts. The rest eight DMUs are both scale and technically inefficient. The least scale inefficient DMU is GUL by scoring 0.247 and the highest scale inefficient DMU is again NIL by with a score of 0.956.

The investigation further investigates the nature of returns to scale of each DMUs. The right most column of Table 6 indicates the nature of returns to scale of each DMUs calculated based on the DEA scores. Here nine (75%) of DMUs namely MR1, ADK, AKK, GUL, KIR, KOK, LID, NIL and YEK shows an increasing return to scale and the rest three (25%) of DMUs like A.A no.1, Merkato no. 2 and Bole sub city exhibits constant returns to scale. Here there is no DMUs shown decreasing returns to scale. For all DMUs that exhibits increasing returns to scale in their operation, an increase in input will result in a more than proportionate increase in output. Thus, the DMUs that work with IRS could achieve substantial efficiency gains by increasing its scales of operation. This could be attained through internal progress and consolidation in the sector.

NIL becomes the only highest inefficient DMU in all the three efficiency measurements of CRS, VRS and SE by scoring 0.952, 0.996 and 0.956 respectively. GUL is the least inefficient firm in CRS and SE with an equivalent score of 0.207 and 0.247 and KIR is the least inefficient DMU at VRS having 0.726 efficiency score.

The highest inefficient DMU NIL needs only 0.048, 0.004 and 0.044 of reduction to its input at CRS, VRS and SE respectively to become efficient. Similarly, the least inefficient DMU at CRS and SE is GUL. It needs as much as 0.793 and 0.753 less reduction in their input usage to become inefficient at CRS and SE. and also the least inefficient DMU at VRS is KIR. It needs 0.274 reduction in its input usage to become efficient.

SUMMARY OF SLACKS

All inefficient tax offices, i.e. MR1, ADIK, AKK, GUL, KIR, KOK, NIL & YEK have both input and output slacks except for the total number of employees. Moreover, the findings from the summary of slacks can further be interpreted the same way as it was interpreted for the CRS DEA model. As here, from Appendix C2, DMU 2 have increasing returns to scale, so MR1 has scale returns discrepancy of Birr 3856809.369 for office rent expense. DMU4 have increasing returns to scale, so ADIK has scale returns discrepancy of 10493.102 for total number of taxpayers. DMU 5 have increasing returns to scale, so AKK has scale returns discrepancy of 5178.917 for total number of taxpayers. DMU 7 have increasing returns to scale, so GUL has scale returns discrepancy of 2831716.511 for office rent expense. DMU 8 have increasing returns to scale, so KIR has scale returns discrepancy of 4490.617 for total number of taxpayers. DMU 9 have increasing returns to scale, so KOK has scale returns discrepancy of 21085.980 for total number of taxpayers. DMU 11 have increasing returns to scale, so NIL has scale returns discrepancy of 48099.391 for total number of taxpayers. DMU 12 have increasing returns to scale, so YEK has scale returns discrepancy of 16958.833for total number of taxpayers.

EFFICIENCY AND THE FRONTIER LINE

Figure 1 illustrates the concepts of efficiency, slacks and references or peers in an intuitive manner using two inputs and one output. The concept of frontier is especially important for the analysis of efficiency, because we measure efficiency as the relative distance to the frontier. For example, firms that are technically inefficient operate at points in the interior of the frontier, while those that are technically efficient operate somewhere along the technology defined by the frontier. The DMU is called efficient when the DEA score equals to one and all slacks are zero (Cooper, Seiford, and Tone, 2006). If only the first condition is satisfied, the DMU is called as efficient in terms of “radial”, “technical”, and “weak” efficiency. If these two conditions are satisfied, the DMU is called efficient in terms of “Pareto-Koopmans” or “strong” efficiency (Yong-bae Ji and Choonjoo Lee).

It is portrayed graphically using the two inputs and one output utilized. In this case we will have 3 variables, and we cannot represent them on the XY plane. Thus, Total Number of Taxpayers i.e. X1 and Office Rent Expense i.e. X2 are used as input variable with Indirect Taxes i.e. Y2 as output variable. since we assumed that constant returns to scale to exist, we can normalize the inputs by the single output Y2. That time, we get the Fig.1, where we can easily contract three efficient DMUs: AAM1, MR2 & BOL remains efficient and DMU s MR1, ADIK, AKK, GUL, KIR, KOK, LID, NIL & YEK are remained inefficient that lies off the efficiency frontier line. Since the input Total Number of Employee and the output Direct Taxes have no non-zero slack value, they are not taken as an input and output variables. The technical efficiency of DMUs 2 and 7 are defined as the distance O2’/O2 and O7’/O7, respectively.

The highest inefficient DMU i.e. NIL had 0.952 efficiency score and the lowest inefficient DMU GUL scored 0.207 efficiency score. Thus, given the current output, NIL has excess input utilization of only 4.8%. whereas GUL exhibited 79.3% excess input utilization given the output unchanged as compared to efficient DMUs. Inefficiency can be seen as how much the inputs must contract along a ray from the origin until it crosses the frontier. For example, for DMU 2, the measure of technical efficiency is the distance O2’/O2. Point O2′ is the Farrell efficient point, however, input X2 could be further reduced and still produce the same output. For this case, firm G has input slack CG1. Reference or peer is a point that an inefficient DMU such as DMU2 targets to move from the Farrell efficient point such as point G1 to Pareto-Koopman efficient point such as point C in Figure 2 based on the two-stage DEA solution or Pareto optimal solution.

X2/Y2

Figure 1 CRS input orientation efficiency frontier line with two inputs X1, X2 and one output Y2

EFFICIENCY TO ESTABLISHMENT PERIOD AND SERVICE YEAR

When we see the efficiency of DMUs with respect to establishment period, two of efficient DMUs (AAM1, MR2) in CRS, VRS & Scale efficiency have established after the mean year 2010. LID, VRS efficient DMU and CRS, VRS and Scale efficient DMU i.e. BOL and all inefficient DMUs except MR1 are established at the mean year 2010. Whereas MR1 is established after the average year of establishment 2010. Therefore, years of establishment doesn’t affect the efficiency of DMUs.

Likewise, the effect of service year on efficiency of each DMU is presented as follows:

Two of efficient DMUs (AAM1, MR2) in CRS, VRS ; Scale efficiency have below the mean service years of 5.58. LID, VRS efficient DMU and BOL, CRS, VRS and Scale efficient DMU and all inefficient DMUs except MR1 have service years above the mean service years of 5.58. While MR1 have served below the average 5.58 service years. Therefore, service year doesn’t have impact on the efficiency of DMUs. The statistics is compiled in Appendix E2.

EFFICIENCY TO ENTERPRISE COMPANY AND INDIVIDUAL COMPANY TAXPAYERS

From table 1, the average number of Enterprise Taxpayers and Individual Taxpayers is computed. The average number of Enterprise Taxpayers is 1620 and the average number of Individual Taxpayers is 27924.17. All efficient DMUs (AAM1 & MR2) in CRS, VRS & Scale efficiency measurement except BOL which have more than the average number of Enterprise Taxpayers and the VRS efficient DMU i.e. LID have below the average number of Enterprise Taxpayers. All inefficient DMUs except MR1, ADIK & KIR which have below the average number of Enterprise Taxpayers, have above average number of Enterprise Taxpayers.

Similarly, if we see efficiency with respect to individual company, all efficient DMUs (AAM1 & MR2) in CRS, VRS & Scale efficiency measurement except BOL (which have more than the average number of Individual Taxpayers) and the VRS efficient DMU i.e. LID have below the average number of Individual Taxpayers. Part of Inefficient DMUs like KOK, NIL & YEK have above the average number of individual taxpayers. The remaining inefficient DMUs MR1, ADIK, AKK, GUL & KIR have below the average number of Individual Taxpayers. Thus, efficiency is not affected by the type of the taxpayers whether they are enterprise taxpayers or individual taxpayers. Appendix E2 is compiled the descriptive statistic results for these variables.

CHAPTER FIVE: CONCLUSIONS AND RECOMMENDATIONS

CONCLUSIONS

Given the current challenge of budget deficit in government, tax offices performance is of major importance for government revenue generation. This study analyzed the technical efficiency and scale efficiency of twelve (12) Addis Ababa City Administration tax collecting offices using cross-sectional data in the year 2015/16 with input- oriented CRS and VRS DEA approach. The inputs used were Total Number Taxpayers, Office Rent Expense and Total Number of Employees. Whereas, the two outputs are direct tax and indirect taxes. From the finding, the main source of overall technical inefficiency in the tax collecting offices is due to pure technical inefficiencies than scale inefficiencies. This suggest that tax collecting offices managers must improve their operational planning and management practices in an efficient way. This could be via an optimal combination of factors of production, improving tax compliance, adequate investment in, and adaption of new technologies relevant to modernization of tax offices, and further training and education in the adaption and use of the new technology. The next step would then be to improve their scale efficiencies. About 75% of the tax offices are exhibiting increasing returns to scale. These tax offices should increase its scale operations through internal growth or consolidation in the sector. RECOMMENDATIONS

Considering the results, the managerial implications of this paper are as follows: The findings show inefficient DMUs have excess number of taxpayers and office rent expense. Therefore, the management should give due concern in their rent office expense plan and recruitment of employees.

Most of the inefficient DMUs are operating below their optimal scale size. Those DMUs should therefore, investigate their operation level and regulate their scale of operation. To undertake the measures firstly, the AARA must upgrade its follow-up inspection procedures regarding the tax offices’ activities, in order to provide more clearly binding incentives for increasing productive efficiency; secondly, the regulatory office must expand the scope of the data obtained in the follow-up inspection, to include contextual factors beyond managerial control, since it is not clear if different offices have the same operating environment; thirdly, a benchmark analysis should be carried out with the data and also published, in order to enforce an efficient adjustment of the least-performing offices; fourthly, the office rent expense should be minimized by having their own buildings or through reducing office area or changing from ground floors to upper floors; fifthly, reducing tax compliance cost so as to get voluntary taxpayers and finally, the salaries should be more directly related to performance.

These measures will define an organizational governance environment with transparency, accountability and efficiency improvements which explicitly force the tax offices to achieve efficiency in their operational activities. These recommendations seek to establish a governance framework within the tax offices, with the aim of improving organizational efficiency. Moreover, the savings that will result from the improved efficiency will reduce waste, allowing more resources to be diverted into alternative public uses. The policy implication of this research is the restructure of the AARA, which is now being implemented proclaim independency of the authority from ERCA, may result in efficiency improvement through a better coordination of managerial tasks. The role of the tax agencies in this improvement will be the allocation of objectives to all tax agencies and to the workers, as a field activity inside the transformation of the public administration. The anticipated result is the improvement of the efficiency. Based in these policies we should expect an improvement in efficiency. Surely, future DEA research should see additional methods for treatment of the data variation and variable selection problems reported herein, as well as empirically testing conclusions in a variety of management control settings.

REFERENCES Barros, C.P. (2007) ‘Technical and allocative efficiency of tax offices: a case study’, Int. J. Public Sector Performance Management, Vol. 1, No. 1, pp.41–61.

Banker, R., Charnes, A. and Cooper, W. (1984) ‘Some models for estimating technical and scale inefficiencies in data envelopment analysis’, Man Sci, 30: 1078–1092.

Barrilao-gonzalez, e. a. (n.d.). relative efficiency within a tax administration: the effects of result improvement

Charnes, A., W. W. Cooper, and E. Rhodes. 1978. Measuring the Efficiency of Decision Making Units. European Journal of Operational Research 2: 429-444.

Coeli, T.J. (2008). A Guide to DEAP Version 2.1: A Data Envelopment Analysis (Computer) Program. CEPA Working Paper 96/08, 1-49.

Cooper, W.W., Seiford, L.M., ; Zhu, J. (2011). Handbook on Data Envelopment Analysis.

Ethiopian Revenues ; Customs Authority’s establishment proclamation no. 587/2008

Ethiopian Revenues & Customs Authority’s five year (2016-2020) strategic plan

Farrell, M.J. (1957) ‘The measurement of productive efficiency’, Journal of the Royal Statistical Society, Series A, Vol. 120, No. 3, pp.253–290.

F. Y. Edgeworth (Mar., 1897), the Pure Theory of Taxation, the Economic Journal, Vol. 7, No. 25. (Mar., 1897), pp. 46-70.

Federal Tax Administration Proclamation no. 983/2008, of Federal Democratic Republic of Ethiopia, Federal Negarit Gazeta Special Edition

Gonzalez, M.X. and Miles, D. (2000) ‘Eficiencia en la Inspección de Hacienda’, Revista de Economia Aplicada, Vol. 8, No. 24, pp.203–219.

Gordana Savi?, Aleksandar Dragojlovi?, Mirko Vujoševi?, Milojko Arsi? ; Milan Marti? (2015), Impact of the efficiency of the tax administration. Economic Research-Ekonomska Istraživanja, 28:1, 1138-1148.

Harold O. Fried, C. A. Knox Lovell and Shelton S. Schmidt, Efficiency and Productivity

Income Tax Proclamation no. 286/2002, Federal Democratic Republic of Ethiopia, Federal Negarit Gazeta, 8th year no. 34 Addis Aababa-4th July, 2002

Income Tax Proclamation no. 979/2008, Federal Democratic Republic of Ethiopia, Federal Negarit Gazeta Special Edition

James Alm and Denvil Duncan (2013), Estimating Tax Agency Efficiency.

John W. Creswell (2003), Research Design Qualitative, Quantitative, and mixed methods approach, 2ndEd.

Jorge Santos, Elsa Ros´ario Negas, and Lu´?s Cavique Santos, A.B. Mendes et al. (eds.) (2013), Introduction to Data Envelopment Analysis, Efficiency Measures in the Agricultural Sector: With Applications, DOI 10.1007/978-94-007-5739-4 3, © Springer Science +Business Media Dordrecht

Kao, C., ; Liu, S.-T. (2011). Scale Efficiency Measurement in Data Envelopment Analysis with Interval Data: A Two-Level Programming Approach. Journal of CENTRUM Cathedra: The Business and Economics, Research Journal, 4(2), 224–235. https://doi.org/10.7835/jcc-berj-2011-0060 Maria Katharaki, Marios Tsakas, (2010) “Assessing the efficiency and managing the performance ofGreek tax offices”, Journal of Advances in Management Research, Vol. 7 Issue: 1, pp.58-75, https://doi.org/10.1108/09727981011042856 Moesen, W. and Persoon, A. (2002) ‘Measuring and explaining the productive efficiency of tax offices: a non-parametric best practice frontier approach’, Tijdschrift Voor Economie en Management, Vol. 47, No. 3, pp.399–416.

Park, J. E., Ryu, S. L. (2012), Efficiency and Productivity Analysis- The Case of Korean Regional Tax Offices. Korean Journal of Taxation Research, 29(3), pp. 223-252

Ramón Fuentes (2014), Productivity at SUMA tax offices: a step further, Conference Proceedings 15th Toulon-Verona Conference “Excellence in Services”, College of Management Academic Studies, Rishon Lezion, Israel, 3-4

Sang-Lyul Ryu, Seok-Young Lee (2013) “An Exploratory Study of Efficiency in Tax Jurisdictions”, Advanced Science and Technology Letters Vol.34, pp.46-49, http://dx.doi.org/10.14257/astl.2013.34.12 Tim Coelli, A guide to DEAP version 2.1: A Data Envelopment Analysis (Computer) Program, http://www.une.edu.au/econometrics/cepa.htm CEPA Working Paper 96/08 Timothy J. Coelli, D.S. Prassada Rao, Christopher J. O’Donnell and George E. Battese (1998), an introduction to efficiency and productivity analysis, Springer Science + Business Media, inc., 2nd edition.

Venkatesh Bhagavath, Technical Efficiency Measurement by Data Envelopment Analysis: An Application in Transportation

Yifru Yirdaw (2016), Efficiency of private commercial banks in Ethiopia: A data envelopment analysis approach, Yom Institute of Economic Development and Debra Markos University, Ethiopia.

Yong-bae Ji, Choonjoo Lee, Data Envelopment Analysis in Stata, The Stata Journal (yyyy) vv, Number ii, pp. 1-13.

Appendix A1 efficiency result summary in CRS model

Results from DEAP Version 2.1

Instruction file = eg1-ins.txt

Data file = eg1-dta.txt

Input orientated DEA

Scale assumption: CRS

Slacks calculated using multi-stage method

EFFICIENCY SUMMARY:

firm te 1 1.000 2 0.295 3 1.000 4 0.445 5 0.326 6 1.000 7 0.207 8 0.509 9 0.530 10 0.306 11 0.952 12 0.541 mean 0.593Appendix A2 Summary of output and input slacks

summary of output slacks:

firm output: 1 2

1 0.000 0.000

2 0.000 31.507

3 0.000 0.000

4 0.000 31.987

5 0.000 6.224

6 0.000 0.000

7 0.000 54.263

8 0.000 55.383

9 0.000 28.234

10 0.000 15.753

11 0.000 84.337

12 0.000 52.263

mean 0.000 29.996

summary of input slacks:

firm input: 1 2 3

1 0.000 0.000 0.000

2 1731.073 1101510.399 0.000

3 0.000 0.000 0.000

4 4128.987 0.000 0.000

5 0.000 0.000 0.000

6 0.000 0.000 0.000

7 3117.977 99225.621 0.000

8 0.000 0.000 0.000

9 4461.252 0.000 0.000

10 0.000 0.000 0.000

11 34884.393 0.000 0.000

12 2464.730 0.000 0.000

mean 4232.368 100061.335 0.000

Appendix A3 summary of peers and peer weights

summary of peers:

firm peers:

1 1

2 1

3 3

4 6 3

5 3 6 1

6 6

7 1

8 6 1 3

9 6 3

10 3 6 1

11 6 3

12 3 6

summary of peer weights:

(in same order as above)

firm peer weights:

1 1.000

2 0.350

3 1.000

4 0.052 0.387

5 0.028 0.164 0.024

6 1.000

7 0.208

8 0.226 0.254 0.055

9 0.342 0.094

10 0.007 0.125 0.053

11 0.283 0.590

12 0.140 0.334

Appendix A4 peer count summary:

(i.e., no. times each firm is a peer for another)

firm peer count:

1 5

2 0

3 7

4 0

5 0

6 7

7 0

8 0

9 0

10 0

11 0

12 0

Appendix A5 summary of output and input targets

summary of output targets:

firm output: 1 2

1 760.480 838.690

2 266.200 293.577

3 306.320 395.750

4 154.610 179.627

5 140.440 114.544

6 694.780 510.330

7 157.840 174.073

8 367.300 350.373

9 266.540 211.814

10 129.420 111.223

11 377.050 377.617

12 275.010 225.963

SUMMARY OF INPUT TARGETS:

firm input: 1 2 3

1 4965.000 14117439.000 292.000

2 1737.959 4941697.693 102.212

3 13033.000 2554092.000 355.000

4 7802.162 1561729.760 158.026

5 9188.470 2215384.689 82.043

6 53120.000 11040000.000 398.000

7 1030.501 2930118.572 60.606

8 13992.184 6226409.056 183.739

9 19403.151 4018067.243 169.544

10 6979.465 2149158.881 67.702

11 22704.091 4627493.037 321.830

12 19568.471 4045249.171 182.731

Appendix B1 Firm by Firm ResultsResults for firm: 1

Technical efficiency = 1.000

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 760.480 0.000 0.000 760.480

output 2 838.690 0.000 0.000 838.690

input 1 4965.000 0.000 0.000 4965.000

input 2 14117439.000 0.000 0.000 14117439.000

input 3 292.000 0.000 0.000 292.000

LISTING OF PEERS:

peer lambda weight

1 1.000

Results for firm: 2

Technical efficiency = 0.295

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 266.200 0.000 0.000 266.200

output 2 262.070 0.000 31.507 293.577

input 1 11777.000 -8307.968 -1731.073 1737.959

input 2 20516058.000 -14472849.908 -1101510.399 4941697.693

input 3 347.000 -244.788 0.000 102.212

LISTING OF PEERS:

peer lambda weight

1 0.350

Appendix B2 Firm by Firm Results

Results for firm: 3

Technical efficiency = 1.000

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 306.320 0.000 0.000 306.320

output 2 395.750 0.000 0.000 395.750

input 1 13033.000 0.000 0.000 13033.000

input 2 2554092.000 0.000 0.000 2554092.000

input 3 355.000 0.000 0.000 355.000

LISTING OF PEERS:

peer lambda weight

3 1.000

Results for firm: 4

Technical efficiency = 0.445

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 154.610 0.000 0.000 154.610

output 2 147.640 0.000 31.987 179.627

input 1 26803.000 -14871.850 -4128.987 7802.162

input 2 3508383.000 -1946653.240 0.000 1561729.760

input 3 355.000 -196.974 0.000 158.026

LISTING OF PEERS:

peer lambda weight

6 0.052

3 0.387

Appendix B3 Firm by Firm Results

Results for firm: 5

Technical efficiency = 0.326

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 140.440 0.000 0.000 140.440

output 2 108.320 0.000 6.224 114.544

input 1 28223.000 -19034.530 0.000 9188.470

input 2 6804702.000 -4589317.311 0.000 2215384.689

input 3 252.000 -169.957 0.000 82.043

LISTING OF PEERS:

peer lambda weight

3 0.028

6 0.164

1 0.024

Results for firm: 6

Technical efficiency = 1.000

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 694.780 0.000 0.000 694.780

output 2 510.330 0.000 0.000 510.330

input 1 53120.000 0.000 0.000 53120.000

input 2 11040000.000 0.000 0.000 11040000.000

input 3 398.000 0.000 0.000 398.000

LISTING OF PEERS:

peer lambda weight

6 1.000

Appendix B4 Firm by Firm Results

Results for firm: 7

Technical efficiency = 0.207

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 157.840 0.000 0.000 157.840

output 2 119.810 0.000 54.263 174.073

input 1 20056.000 -15907.522 -3117.977 1030.501

input 2 14645497.000 -11616152.807 -99225.621 2930118.572

input 3 293.000 -232.394 0.000 60.606

LISTING OF PEERS:

peer lambda weight

1 0.208

Results for firm: 8

Technical efficiency = 0.509

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 367.300 0.000 0.000 367.300

output 2 294.990 0.000 55.383 350.373

input 1 27491.000 -13498.816 0.000 13992.184

input 2 12233273.000 -6006863.944 0.000 6226409.056

input 3 361.000 -177.261 0.000 183.739

LISTING OF PEERS:

peer lambda weight

6 0.226

1 0.254

3 0.055

Appendix B5 Firm by Firm Results

Results for firm: 9

Technical efficiency = 0.530

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 266.540 0.000 0.000 266.540

output 2 183.580 0.000 28.234 211.814

input 1 45042.000 -21177.597 -4461.252 19403.151

input 2 7583755.000 -3565687.757 0.000 4018067.243

input 3 320.000 -150.456 0.000 169.544

LISTING OF PEERS:

peer lambda weight

6 0.342

3 0.094

Results for firm: 10

Technical efficiency = 0.306

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 129.420 0.000 0.000 129.420

output 2 95.470 0.000 15.753 111.223

input 1 22783.000 -15803.535 0.000 6979.465

input 2 7015479.000 -4866320.119 0.000 2149158.881

input 3 221.000 -153.298 0.000 67.702

LISTING OF PEERS:

peer lambda weight

3 0.007

6 0.125

1 0.053

Appendix B6 Firm by Firm Results

Results for firm: 11

Technical efficiency = 0.952

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 377.050 0.000 0.000 377.050

output 2 293.280 0.000 84.337 377.617

input 1 60482.000 -2893.516 -34884.393 22704.091

input 2 4860000.000 -232506.963 0.000 4627493.037

input 3 338.000 -16.170 0.000 321.830

LISTING OF PEERS:

peer lambda weight

6 0.283

3 0.590

Results for firm: 12

Technical efficiency = 0.541

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 275.010 0.000 0.000 275.010

output 2 173.700 0.000 52.263 225.963

input 1 40755.000 -18721.799 -2464.730 19568.471

input 2 7482532.000 -3437282.829 0.000 4045249.171

input 3 338.000 -155.269 0.000 182.731

LISTING OF PEERS:

peer lambda weight

3 0.140

6 0.334

Appendix C1 efficiency result summary in VRS model

Results from DEAP Version 2.1

Instruction file = eg1-ins.txt

Data file = eg1-dta.txt

Input orientated DE

Scale assumption: VRS

Slacks calculated using multi-stage method

EFFICIENCY SUMMARY:

firm crste vrste scale

1 1.000 1.000 1.000 –

2 0.295 0.791 0.372 irs 3 1.000 1.000 1.000 –

4 0.445 0.938 0.475 irs 5 0.326 0.946 0.344 irs 6 1.000 1.000 1.000 –

7 0.207 0.836 0.247 irs 8 0.509 0.726 0.701 irs 9 0.530 0.854 0.620 irs 10 0.306 1.000 0.306 irs 11 0.952 0.996 0.956 irs 12 0.541 0.830 0.651 irs mean 0.593 0.910 0.639

Note: crste = technical efficiency from CRS DEA

vrste = technical efficiency from VRS DEA

scale = scale efficiency = crste/vrsteNote also that all subsequent tables refer to VRS results

Appendix C2 summary of input and output slacks

summary of output slacks:

firm output: 1 2

1 0.000 0.000

2 339.997 394.915

3 0.000 0.000

4 122.532 198.582

5 11.922 26.093

6 0.000 0.000

7 184.608 226.550

8 0.000 92.191

9 0.000 103.057

10 0.000 0.000

11 0.000 164.541

12 0.000 128.042

mean 54.922 111.164

summary of input slacks:

firm input: 1 2 3

1 0.000 0.000 0.000

2 0.000 3856809.369 0.000

3 0.000 0.000 0.000

4 10493.102 0.000 0.000

5 5178.917 0.000 0.000

6 0.000 0.000 0.000

7 0.000 2831716.511 0.000

8 4490.617 0.000 0.000

9 21085.980 0.000 0.000

10 0.000 0.000 0.000

11 48099.391 0.000 0.000

12 16958.833 0.000 0.000

mean 8858.903 557377.157 0.000

Appendix C3 summary of peers and peer counts

summary of peers:

firm peers:

1 1

2 1 10

3 3

4 3 10

5 10 3

6 6

7 1 10

8 1 10 3

9 1 3 10

10 10

11 3 1 10

12 1 3 10

summary of peer weights:

(in same order as above)

firm peer weights:

1 1.000

2 0.756 0.244

3 1.000

4 0.835 0.165

5 0.870 0.130

6 1.000

7 0.338 0.662

8 0.342 0.533 0.126

9 0.127 0.323 0.550

10 1.000

11 0.769 0.177 0.054

12 0.125 0.379 0.497

Appendix C4 peer count summary:

(i.e., no. times each firm is a peer for another)

firm peer count:

1 6

2 0

3 6

4 0

5 0

6 0

7 0

8 0

9 0

10 8

11 0

12 0

Appendix C5 summary of output and input targets

summary of output targets:

firm output: 1 2

1 760.480 838.690

2 606.197 656.985

3 306.320 395.750

4 277.142 346.222

5 152.362 134.413

6 694.780 510.330

7 342.448 346.360

8 367.300 387.181

9 266.540 286.637

10 129.420 95.470

11 377.050 457.821

12 275.010 301.742

SUMMARY OF INPUT TARGETS:

firm input: 1 2 3

1 4965.000 14117439.000 292.000

2 9321.197 12381130.873 274.642

3 13033.000 2554092.000 355.000

4 14641.167 3289954.146 332.898

5 21518.523 6436882.136 238.378

6 53120.000 11040000.000 398.000

7 16768.155 9412896.722 244.968

8 15468.941 8881842.104 262.100

9 17376.503 6475956.947 273.256

10 22783.000 7015479.000 221.000

11 12134.003 4840023.377 336.611

12 16872.034 6211275.805 280.575

Appendix D1 Firm by Firm Results

Results for firm: 1

Technical efficiency = 1.000

Scale efficiency = 1.000 (crs)

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 760.480 0.000 0.000 760.480

output 2 838.690 0.000 0.000 838.690

input 1 4965.000 0.000 0.000 4965.000

input 2 14117439.000 0.000 0.000 14117439.000

input 3 292.000 0.000 0.000 292.000

LISTING OF PEERS:

peer lambda weight

1 1.000

Results for firm: 2

Technical efficiency = 0.791

Scale efficiency = 0.372 (irs)

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 266.200 0.000 339.997 606.197

output 2 262.070 0.000 394.915 656.985

input 1 11777.000 -2455.803 0.000 9321.197

input 2 20516058.000 -4278117.759 -3856809.369 12381130.873

input 3 347.000 -72.358 0.000 274.642

LISTING OF PEERS:

peer lambda weight

1 0.756

10 0.244

Appendix D2 Firm by Firm Results

Results for firm: 3

Technical efficiency = 1.000

Scale efficiency = 1.000 (crs)

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 306.320 0.000 0.000 306.320

output 2 395.750 0.000 0.000 395.750

input 1 13033.000 0.000 0.000 13033.000

input 2 2554092.000 0.000 0.000 2554092.000

input 3 355.000 0.000 0.000 355.000

LISTING OF PEERS:

peer lambda weight

3 1.000

Results for firm: 4

Technical efficiency = 0.938

Scale efficiency = 0.475 (irs)

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 154.610 0.000 122.532 277.142

output 2 147.640 0.000 198.582 346.222

input 1 26803.000 -1668.731 -10493.102 14641.167

input 2 3508383.000 -218428.854 0.000 3289954.146

input 3 355.000 -22.102 0.000 332.898

LISTING OF PEERS:

peer lambda weight

3 0.835

10 0.165

Appendix D3 Firm by Firm Results

Results for firm: 5

Technical efficiency = 0.946

Scale efficiency = 0.344 (irs)

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 140.440 0.000 11.922 152.362

output 2 108.320 0.000 26.093 134.413

input 1 28223.000 -1525.560 -5178.917 21518.523

input 2 6804702.000 -367819.864 0.000 6436882.136

input 3 252.000 -13.622 0.000 238.378

LISTING OF PEERS:

peer lambda weight

10 0.870

3 0.130

Results for firm: 6

Technical efficiency = 1.000

Scale efficiency = 1.000 (crs)

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 694.780 0.000 0.000 694.780

output 2 510.330 0.000 0.000 510.330

input 1 53120.000 0.000 0.000 53120.000

input 2 11040000.000 0.000 0.000 11040000.000

input 3 398.000 0.000 0.000 398.000

LISTING OF PEERS:

peer lambda weight

6 1.000

Appendix D4 Firm by Firm Results

Results for firm: 7

Technical efficiency = 0.836

Scale efficiency = 0.247 (irs)

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 157.840 0.000 184.608 342.448

output 2 119.810 0.000 226.550 346.360

input 1 20056.000 -3287.845 0.000 16768.155

input 2 14645497.000 -2400883.768 -2831716.511 9412896.722

input 3 293.000 -48.032 0.000 244.968

LISTING OF PEERS:

peer lambda weight

1 0.338

10 0.662

Results for firm: 8

Technical efficiency = 0.726

Scale efficiency = 0.701 (irs)

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 367.300 0.000 0.000 367.300

output 2 294.990 0.000 92.191 387.181

input 1 27491.000 -7531.442 -4490.617 15468.941

input 2 12233273.000 -3351430.896 0.000 8881842.104

input 3 361.000 -98.900 0.000 262.100

LISTING OF PEERS:

peer lambda weight

1 0.342

10 0.533

3 0.126

Appendix D5 Firm by Firm Results

Results for firm: 9

Technical efficiency = 0.854

Scale efficiency = 0.620 (irs)

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 266.540 0.000 0.000 266.540

output 2 183.580 0.000 103.057 286.637

input 1 45042.000 -6579.516 -21085.980 17376.503

input 2 7583755.000 -1107798.053 0.000 6475956.947

input 3 320.000 -46.744 0.000 273.256

LISTING OF PEERS:

peer lambda weight

1 0.127

3 0.323

10 0.550

Results for firm: 10

Technical efficiency = 1.000

Scale efficiency = 0.306 (irs)

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 129.420 0.000 0.000 129.420

output 2 95.470 0.000 0.000 95.470

input 1 22783.000 0.000 0.000 22783.000

input 2 7015479.000 0.000 0.000 7015479.000

input 3 221.000 0.000 0.000 221.000

LISTING OF PEERS:

peer lambda weight

10 1.000

Appendix D6 Firm by Firm Results

Results for firm: 11

Technical efficiency = 0.996

Scale efficiency = 0.956 (irs)

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 377.050 0.000 0.000 377.050

output 2 293.280 0.000 164.541 457.821

input 1 60482.000 -248.606 -48099.391 12134.003

input 2 4860000.000 -19976.623 0.000 4840023.377

input 3 338.000 -1.389 0.000 336.611

LISTING OF PEERS:

peer lambda weight

3 0.769

1 0.177

10 0.054

Results for firm: 12

Technical efficiency = 0.830

Scale efficiency = 0.651 (irs)

PROJECTION SUMMARY:

variable original radial slack projected

value movement movement value

output 1 275.010 0.000 0.000 275.010

output 2 173.700 0.000 128.042 301.742

input 1 40755.000 -6924.133 -16958.833 16872.034

input 2 7482532.000 -1271256.195 0.000 6211275.805

input 3 338.000 -57.425 0.000 280.575

LISTING OF PEERS:

peer lambda weight

1 0.125

3 0.379

10 0.497

Appendix E1 CRS, VRS and Scale Efficiency results of fourteen DMUs

Results from DEAP Version 2.1

Instruction file = eg1-ins.txt

Data file = eg1-dta.txt

Input orientated DEA

Scale assumption: VRS

Slacks calculated using multi-stage method

EFFICIENCY SUMMARY:

firm crste vrste scale

1 0.823 1.000 0.823 irs 2 1.000 1.000 1.000 –

3 0.163 0.791 0.206 irs 4 0.652 1.000 0.652 irs 5 0.177 0.938 0.189 irs 6 0.081 0.946 0.086 irs 7 1.000 1.000 1.000 –

8 0.248 0.650 0.381 irs 9 0.084 0.836 0.101 irs 10 0.155 0.685 0.226 irs 11 0.138 0.802 0.173 irs 12 0.079 1.000 0.079 irs 13 0.306 0.915 0.334 irs 14 0.145 0.780 0.186 irs mean 0.361 0.882 0.388

Note: crste = technical efficiency from CRS DEA

vrste = technical efficiency from VRS DEA

scale = scale efficiency = crste/vrsteAppendix E2 EVIEWS results for Descriptive statistics for inputs, outputs and other variables of 12 Addis Ababa tax offices

Appendix E3 Comparison of the characteristics of the variables in each tax offices

Appendix E4 empirical data of input, output and other variables for the year 2015/16

Appendix F1 research activity and time schedules

No Activities 2017

Jan2-April14 2017 April 15- May 05

2017

May 16- July 20

2017

July 22-23

2017

Sept01-15 2017

Sept

16-Jan 01, 2018 2018

Jan. 15 2018

Feb 10-11

2018

Feb15-28,

1 Thesis title selection 2 Topic approval 3 Submit concept paper to advisor 4 Writing thesis proposal 5 Submission of first draft proposal 6 Submission of final proposal 7 Proposal defense 8 Field Work (secondary Data Gathering) 9 Data screening, encoding, entry, generating preliminary analysis & interpretation 10 Submission of 1st thesis draft 11 Submission of final draft Thesis 12 Thesis Defense 13 Final Compilation & submission Appendix F2 budgetsNo Activity/ Items Unit 0f Measurement Quantity Unit Price Total Price in Birr

1 Stationary Materials 1.1 Duplicating Paper Ream 4 120 480

1.2 Square Paper Ream 3 70 210

1.3 A4 paper Ream 4 120 480

1.4 Stapler Pieces 1 80 80

1.5 Writing Pad Pieces 5 20 100

1.6 Correction Fluid Pieces 3 18 54

1.7 Pen & Pencil Packet 1 250 250

1.8 Staples Packet 5 10 50

Sub Total 1704

2 Secretary Service 2.1 Typing Page 300 5 1500

2.2 Photocopy Page 1000 1 1800

2.3 Thesis 1st Draft Print Page 100*2 1 200

2.4 Thesis Final Print Page 100*4 1 400

Sub Total 3900

3 Auxiliary Material 3.1 Mobile Card Number 20 25 500

3.2 Field Bag Number 1 600 600

3.3 Scientific Calculator Number 1 180 180

Sub Total 1280

4 Other 4.1 Transport Cost Trip 20 times 150 3000

4.2 Entertainment cost Lump sum 30 100 3000

Sub Total 6000

5 Contingency Cost 15% 1932.6

Grand Total 14, 817 Birr

DATA CODING SHEETS

LETTERS OF INTRODUCTION

PERMISSIONS

SPECIAL DOCUMENTS