MEASURING OPERATIONAL EFFICIENCY OF THE INSURANCE INDUSTRY IN KENYA USING DATA ENVELOPMENT ANALYSIS

Transcription

1 MEASURING OPERATIONAL EFFICIENCY OF THE INSURANCE INDUSTRY IN KENYA USING DATA ENVELOPMENT ANALYSIS NOEL TOYA MWANGETI A RESEARCH PROJECT SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENT FOR THE AWARD OF MASTERS OF BUSINESS ADMINISTRATION DEGREE, SCHOOL OF BUSINESS, UNIVERSITY OF NAIROBI 2012

2 DECLARATION This research project is my original work and has not been presented for the award of a degree in any university Signed. Date.. Noel Toya Mwangeti Reg. no. D61/73693/2009 This research project has been submitted for examination with my approval as the University Supervisor Signed. Date... Stephen Odock Lecturer, Department of Management Science, School of Business, University of Nairobi i

3 ACKNOWLEDGEMENTS First I would like to thank the Almighty God for his constant strength throughout my life day after day. I would also like to appreciate my project supervisor Mr. Stephen Odock and moderator Mr. Lazarus Mulwa for the constructive criticism, effort, understanding, guidance and support throughout the research process. Finally, I wish to thank my Wife Pauline and lovely daughter Balbina for giving me conducive atmosphere and ample time to complete this project. Without forgetting my friends John, Raphael, Jesse, Stephen, Cousin Levin, my brother Sylvester and others who are not mentioned for their encouragement and support throughout my research. ii

4 DEDICATION I dedicate this paper to all my friends and my family members, who have encouraged, supported, inspired me to achieve my lifetime goals. iii

5 ABSTRACT There has been ever growing concern to measure relative operational efficiency of Decision Making Units (DMU). Stochastic frontier analysis and Regression have been the most popular methods used to measure the same. This study examines the use of Data Envelopment Analysis model to determine the relative operational efficiency of insurance firms in Kenya. The model used is a multi-variable usage in terms of inputs versus outputs. The data utilized was for the period between January to December The input and output data are analyzed using Data Envelopment Analysis computer program. The objective of this study was to obtain the following information of each insurance firm: relative efficiency score, peer for each inefficient insurance firm, objective output and inputs targets, slacks outputs units and surplus inputs. Finally, relationship between relative efficiency score and to its total assets admitted is noted. The study found out that 23 of the efficient insurance firm s average efficiency score was one (1) while 12 are inefficient with an efficiency score of Meanwhile, inefficiency in insurance firms in Kenya is caused by low net insurance premium earned and larger firms not utilizing assets effectively. The following recommendations are then made to the insurance industry: the 23 insurance firms should be investigated and their best practices adopted by its peers, products to be developed aimed at increasing net earned premium and finally, asset verification exercise should be done to ensure all assets are productively used and the obsolete ones disposed while excess deployed to needy insurance firms branches. iv

6 TABLE OF CONTENTS DECLARATION... i ACKNOWLEDGEMENTS... ii DEDICATION... iii ABSTRACT... iv LIST OF ABBREVIATIONS... viii CHAPTER ONE: INTRODUCTION Background of the Study Operational Efficiency Data Envelopment Analysis Insurance Industry in Kenya Research Problem Objectives of the Study Value of the Study CHAPTER TWO: LITERATURE REVIEW Introduction Operational Efficiency Value for Measuring Operational Efficiency Operational Efficiency Measurements Methods The Need for DEA Data Envelopment Analysis model Choice of Inputs and Outputs Variables Advantages and Limitations of Data Envelopment Analysis Efficiency Studies Summary CHAPTER THREE: RESEARCH METHODOLOGY Introduction Research Design Population Data Collection Data Analysis v

7 CHAPTER FOUR: DATA ANALYSIS, RESULTS AND DISCUSSIONS Introduction Correlation Analysis of Chosen Output and Input Variables Relative Operational Efficiency Scores Relationship between firm Relative Efficiency Score and its Total Assets in General Insurance CHAPTER FIVE: SUMMARY, RECOMMENDATIONS AND CONCLUSIONS Introduction Summary of findings Conclusion Recommendations Limitations of the study Suggestions for further studies REFERENCES...46 APPENDICES...54 Appendix 1: Secondary data Appendix 2: DEA computer program output file summary Appendix 3: SPSS Output for correlation analysis vi

8 LIST OF TABLES AND FIGURES TABLES Table 4.1: Input-output correlation matrix Table 4.2: P-Values for testing correlation significance at 95% confidence level Table 4.3: Relatively efficient firms within the insurance industry Table 4.4: Relatively inefficient firms within the industry Table 4.5 Best peers for each inefficient firm Table 4.6: Mean Score Vs Mean of Admitted Asset distribution Table 4.7: Paired Sample Mean Difference test FIGURE Figure 4.1: Efficiency Histogram vii

9 LIST OF ABBREVIATIONS AKI: Association of Kenya Insurers CCR: Charnes, Cooper and Rhodes CRS: Constant Return to Scale DEA: Data Envelopment Analysis DEAP: Data Envelopment Analysis Program DFA: Distribution Free Approach DMU: Decision Making Unit FDH: Free Disposal Hull GDP: Gross Domestic Product ICT: Information & Communications Technology IRA: Insurance Regulatory Authority IT: Information Technology KCB: Kenya Commercial Bank NGOs: Non Governmental Organisations SFA: Stochastic Frontier Approach SPSS: Statistical Package for the Social Sciences viii

10 TFA: Thick Frontier Approach TFP: Total Factor Productivity UNCTAD: United Nations Conference on Trade and Development VRS: Variable Return to Scale ix

11 CHAPTER ONE: INTRODUCTION 1.1 Background of the Study Continuous performance improvement, growth and survival of any organization operating in a business environment with stiff competition and technological advancement depend on its ability to measure and improve operational efficiency effectively. Bhagavath (2009) stated that operational efficiency concept has become of concern due to increased competition, business processes and new technology growth. Operational efficiency refers to the ability of a firm to minimize waste and maximize resource capabilities in order to deliver quality products and services to customers (Kalluru & Bhat, 2009). It involves identifying wasteful processes and resources that affects productivity and growth of organizations profits. Operational efficiency is concerned with redesigning new work processes that improves quality and productivity (Darrab & Khan, 2010). Charnes, Cooper and Rhodes (1978) defines operational efficiency as the maximum ratio of weighted outputs to weighted inputs. Improving operational efficiency has direct impact on the organizations profit margins. Sufian (2007) stated that efficient firms are more profitable. Therefore measurement of operational efficiency has become a very important issue for business managers and industries. Several methods have been employed to measure operational efficiency and to assess the business performance. These include parametric and non-parametric methods. The most commonly used parametric methods are Stochastic Frontier Approach (SFA), Distribution Free Approach (DFA), Thick Frontier Approach (TFA) and Multiple 1

12 Regression while non-parametric methods are Data Envelopment Analysis (DEA) and the Free Disposable Hull (FDH) (Cumming and Zi, 1998). Parametric methods use parameters called quantities which are mean and standard deviation. They assume data follows a normal distribution and the spread of the data is uniform either between the groups or across the range being studied. Non-parametric methods analyze data which do not meet normal distribution requirement of parametric methods. Skewed data are specifically analyzed using non-parametric methods (Cumming and Zi, 1998). However, organizations operational efficiency is a multidimensional which involves more than a single input versus output to meet its diverse business targets (Zhu, 2000). This study used DEA which is a multi-factor performance model able to identify how an organization is expected to increase its multiple outputs by improving its operational efficiency without the need to use any further input (Mohamad & Said, 2010) Operational Efficiency Operational efficiency is the ability of a business entity to deliver services or products in the most cost effective way while taking into consideration high quality products produced, services and support. Efficiency involves a firm being able to operate with minimum level of resources (inputs) to produce outputs and remain competitively over an extended period of time (Mayes, Harts & Lansbury, 1994). Efficiency is the success with which an organization uses its available resources to produce outputs of a given quality (Bhagavath, 2009). 2

13 Operational efficiency can either be absolute or relative operational efficient depending on the efficiency score measured. Absolute operational efficiency refers to the measure of a single business unit efficiency score independent of the other scores obtained from other business units whereas relative operational efficiency is the measure of scores of a certain firm obtained compared to other firms in the market or business units under study. Relative operational efficiency gives stronger results compared to accounting ratios which measures efficiency and can be used effectively by top management in benchmarking against the best performers in the industry (Reid and Sanders, 2007). Efficiency scores obtained for organization can be used to formulate operational strategy to enable a firm meet its business objectives and goals by enhancing efficiently allocation of available resources in order to maximize outputs of the firm, set achievable targets depending on its assets, compare different firms performance measured in the study, individual firms can make objective plan and can promote use of best practices within the firm with appropriate return to scale (Reid & Sanders, 2007). An organizational operational efficiency is dependent on several factors including skilled and experienced workers, proper technological advancement, proper procurement practices, return to scale of the firm, supply chain management among many others. Relatively more efficient firms tend to maintain more stability levels in terms of output and operating performance compared to their other industry peers (Mills and Schumann, 1985) and such kind of firms tend to have even more sustainable performance (Berger, Hunter & Timme, 1993). 3

14 1.1.2 Data Envelopment Analysis Data Envelopment Analysis (DEA) is an operational based research method which is a multi-factor productivity analysis model for measuring the relative operational efficiency of a homogeneous set of decision making units (DMU). These DMUs can be schools, banks and hospitals which they use a set of multiple homogeneous inputs to produce a set of multiple homogeneous outputs. Using linear programming formulations the model first gets the most efficient units or firms which forms the efficient frontier consisting series of points joining the most efficient units normally obtained from comparing outputs and inputs of all DMUs considered during calculations. These efficient frontiers act as benchmarks to measure the performance of all firms for the study (Lewis, 2000). DEA is a model that has the capability of combining all firms processes involving multiple inputs and outputs into a single measure of operational efficiency that lies between zero and one making it more appropriate choice and outweighs its statistical disadvantages (Saad & Idris, 2011). It normally gives detailed comparative performance information of the DMU in terms of efficiency score. An efficient DMU has an efficiency score of one while one that is inefficient has a score of less than one (Charnes et al., 1978). For inefficient DMU, DEA identifies its peers from the set of efficient units that it is compared with and improvements in outputs and inputs levels required by the firm or unit to be on the efficient frontier therefore simply DEA provides inefficient units or firms with a path to reach to the frontier (Mburu, 2011). 4

15 Since 1978, DEA has been used to compute comparative performances in organizations like academic institutions (Kao & Hung, 2008) hospitals and health centers (Kirigia, Emrouznejad & Sambo, 2002), financial institutions (Mostafa, 2007), manufacturing industries (Ali & Nakosteen, 2005) among others. The main underlying assumption for any DMU (For example an insurance firm) K is able to get Y(K) weighted output given X(K) amount of input then all firms (All insurance firms) should be able to get the same or obtain greater than the same when they are operating homogeneously and operate efficiently relative to one another. The major advantages of DEA compared to other statistical methods and accounting ratios are that it can accommodate multiple inputs and outputs when calculating relative operational efficiency. It provides potential role models for the firms that are not efficient by identifying peers for organizations with aim of improving its operations by setting benchmark. DEA can measure inputs and outputs with different measurements and firms can be compared with peers under similar business environments and capability. Leibenstien & Maital (1992) had identified DEA as superior method to all other available methods of measuring relative operational efficiency Insurance Industry in Kenya The insurance industry in Kenya employs significant number of employees who contributes 5.6% Gross Domestic Product (GDP) to the economy and raising of living standards of Kenyans (Insurance Regulatory Authority, 2010). The insurance industry is very important to the policy holders or clients and general growth of Kenyan economy. Clients benefits include transfer of risks, investments for future, builds savings, safe and 5

16 long term investment, annuities for retirement planning, facility of loans without affecting the policy benefits, tax benefits, mortgage redemption, peace of mind, stimulates country economy growth, makes businesses safer, eases business transactions, improves consumer and workers safety, alleviate poverty, reduction of losses among many others. As at 31 December 2010, Association of Kenya Insurers (AKI) insurance industry Report, the insurance sector comprised of 46 licensed insurance companies of which 22 companies were non-life insurance business, 9 life insurance while 14 were composite (both life and non life). There were also 163 licensed insurance brokers, 23 medical insurance providers and 4223 insurance agents. There were other licensed players in the insurance market which included 120 investigators, 80 motor assessors, 21 loss adjusters, 2 claim setting agents, 10 risk managers and 26 insurance surveyors. The insurance industry in Kenya is regulated by Insurance Regulatory Authority (IRA) under Insurance Act. Cap 487 which was revised in According to the United Nations Conference on Trade and Development (UNCTAD) survey, the penetration of insurance in Africa apart from South Africa is very small. In Kenya it was found to be 0.94% and 1.7% for life and General insurance respectively. However, there is evidence of potential growth and development in the Kenya insurance Industry from the 2010 AKI Statistics Report. The AKI Statistics report showed 22.7% growth in premium to 840 million dollars in 2010 from 640 million dollars in 2009 and penetration grew from 2.8% to 3% within the same period. The penetration of Kenyan insurance industry stands at 5.6% of the Gross Domestic Product (GDP) (Insurance Regulatory Authority, 2010). 6

17 In spite of these achievements the insurance industry has faced several challenges which it really needs to address. These challenges include; bad reputation due to slow or even failure to settle claims which is changing with the putting into place of the Financial Assets Bill 2011 which enforces 40% clearance of outstanding claims, lack of rigorous campaigns to improve their image to the public through marketing and sensitization to be more reliable in settling claims, lack of incentives to attract more to people take insurance policies and this can be spearheaded by AKI through mass media by educating public importance of taking insurance which will boost the industry, social hindrance which has made insurance not to grow like the spirit of harambee, which has demonstrated dependency attitude in terms of settling hospitals bills, school fees, funeral arrangements and others which can be discouraged by coming up with relevant insurance policies with several components like Death, Savings life, Disability, critical illnesses and others with affordable premiums (AKI Report, 2011). Finally, as in all developing nation economies, Kenya has never had an appropriate distribution channel since the target market is urban people in formal employment leaving out rural areas people. However, technological innovations has made it easier for dissemination of information through mobile phones and the internet. Anyone can now access several services wherever they are. The advancement of technology, knowledge and science has transformed the world into a global business environment and calls for business organizations like insurance companies to be ready to meet the challenges emerging if they are to survive and remain as key players in the market. The challenges they encounter including bad reputation, lack of rigorous campaigns, lack of incentives to clients, social hindrance, and lack of proper distribution channel, and frauds among 7

18 others in the homogeneous business environment with stiff competition calls for proper operational efficiency strategy (Saad et al., 2011). Therefore, measurement of a firm s operational efficiency is very important for its business and failure to sustain performance can damage the firm s reputation leading to customer s defections and breakdowns in relations leading to poor operational efficiency. Currently, insurance industry in Kenya uses accounting ratios to determine operational efficiency which is a simple ratio of single input versus single output. DEA being superior scientific method in measurement of relative operational efficiency and its multivariable ability usage in terms of inputs and outputs gives a more informed decision unlike accounting ratios which uses single variable for input versus output to make decisions (Liebenstein et al., 1992). 1.2 Research Problem Operational efficiency in business management is a very useful tool in re-allocation of available resources in business environment with homogeneous products and with multiple units of inputs and outputs (Oral & Yolalan, 1990). Efficiently operating firms are able to minimize inputs used and maximize outputs produced and therefore remain competitive in the market (Sherman & Gold, 1985). Operational efficiency gives the general view of the firm s performance and operational capability. This is mainly because it takes care of all operational related issues of such firms like supply chain management, quality management, employee motivation, product design, scheduling, layout, employee culture, inventory management, organizational structure and internal processes (Mburu, 2011). Therefore, the management and improvement of the operational efficiency of an industry has to take into consideration all the factors affecting the operations of the firms. 8

19 These factors need to be fully understood consolidated as they depend on each other and affect the operational performance of the firm. In spite of the insurance industry in Kenya contributing 5.6% GDP, employing significant number of employees to the economy and raising of living standards of Kenyans (Insurance Regulatory Authority, 2010), it uses traditional measurement methods of performance which provides unclear picture of the performance of an organization and can lead managers into missing important opportunities for performance improvement. Currently, insurance industry uses accounting ratios method for measurement of operational efficiency which has various disadvantages. These include; use of single input versus single output variable, they can t be used for benchmarking and resource allocation of the firm, they can t differentiate between dependent and independent variables affecting efficiency of a firm and they do not take into consideration the optimality of an output. Therefore, there is need to come up with a better alternative that will consolidate all operational factors to help the insurance companies make better informed decisions in terms of its operational strategy to be employed to make them more efficient. In view of all the above shortcomings of accounting ratios, this study employs use of DEA scientific method which consolidates all operational factors in form of inputs and outputs to measure relative operational efficiency of firms (Verma & Gavirneni, 2006). DEA has been applied by several developed countries to measure operational efficiency. For example, Rees & Keesner (2000) and Diacon, Starkey & O Brien (2002) conducted studies by comparing internationally, the efficiency of insurance companies in Europe. In Kenya a study on insurance operational efficiency was done by Waweru (2007) who 9

20 concluded that efficiency in insurance industry is achieved through using success factors of Information & Communication Technology (ICT). Michino (2011) focused on efficiency in terms of internal controls among Non Governmental Organizations (NGOs) while Gituto (2009) determined the relationship of operational efficiency between Commercial Banks in Kenya computed using accounting ratios. The most related study was done by Mburu (2011) who used DEA scientific method to measure operational efficiency in Kenya Commercial Bank. These studies were very important in Kenya but none has utilized a scientific method to compute relative operational efficiency of an insurance industry in Kenya. This study intends to fill this gap by showing that there are other variables in forms of inputs and outputs which can give a better measure of operational efficiency of an insurance firm. This then leads the researcher to pose the question; how can relative operational efficiency of firms in the insurance industry be measured using DEA method? 1.3 Objectives of the Study The overall objective of the study was to use DEA to measure the relative operational efficiency of insurance firms in Kenya. The specific objectives of the study upon completion were to: (i) Measure the relative operational efficiency of insurance firms in Kenya which were in operation by end of 31 st December (ii) Determine if there exists any relationship between relative operational efficiency score of an insurance firm and its total assets. 10

21 1.4. Value of the Study The study will be very important to the insurance industry as it will give a scientific operational based method of measuring individual insurance firm relative operational efficiency. The insurance firms in Kenya may use the results of the study to improve their operational management practices by identifying operational benchmarks. Insurance firms may use the outcome of the study in terms of re-allocation of available resources and come up with achievable performance target for an insurance firm. It can also be used to increase competitive advantage of the insurance industry by employing future operation strategy in decision making. This model can be applied in other industries and organizations for the same purposes. Academicians including researchers and students can use the information as a base for future studies since it encourages them to apply operational research models in their data analysis and management of business in the fields of operational performance and productivity. The study will also offer guidelines to businesses organizations and government regulation bodies like IRA in formulation of good policies related to operational efficiency. The government can also come up with policies across regions like East Africa for the efficiency levels to be measured with the same scientific method like DEA. 11

22 CHAPTER TWO: LITERATURE REVIEW 2.1 Introduction Research studies have been conducted and several theories developed on the concept of operational efficiency including measurement methods and various definitions. This chapter focuses on value of measuring operational efficiency, operational efficiency measurements methods, the need for DEA, DEA model, choice of inputs and outputs variables, advantages and limitations of DEA. Finally, literature explores efficiency studies. 2.2 Operational Efficiency Value for Measuring Operational Efficiency Operational efficiency of the insurance industry is very important since the industry is currently facing many challenges ranging from increased competition, solvency risks, consolidation and the dynamic regulatory business environment. Therefore, operational efficiency in the insurance industry is critical for revival (Berger et al., 1993). Operational efficiency enables individual firms in the market to survive business threats and gain distinctive competitive advantage (Sherman & Gold, 1985). According to Saad & Idris (2011) operational efficiency in the insurance industry in both emerging and developed economies has important implications for both the insurance operators, policy makers and regulators. For insurance operators it is important in improving competitive edge, and for policy makers and regulators, it is important in maintaining stability and effectiveness of the insurance industry. 12

23 Operational efficiency measurement is a useful tool for re-allocation of available resources in a firm. This process minimizes inputs and maximizes outputs produced and therefore a firm remains competitive in the market over an extended period of time (Sherman & Gold, 1985; Oral & Yolalan, 1990). It gives the general view of the operational performance of any business entity and takes care of all operational issues whether measured quantitatively or qualitatively. Sufian (2007) stated that efficient firms are more profitable and hence measurement of efficiency is important to business executives who wish to achieve certain set targets. Improving operational efficiency implies improving firm s profitability. Operational efficiency can be used by top management in benchmarking of firms against the best performers in the industry (Reid & Sanders, 2007). Using DEA method it provides a path on how the specific firms can reach the efficient frontiers. Efficiency scores obtained for organizations studied can be used in formulating operational strategy which enables firms to meet their business goals and objectives. It makes individual firms have objective plan and promote best organization practices. This enables the firm to have appropriate return to scale on investment (Reid & Sanders, 2007). 13

24 2.2.2 Operational Efficiency Measurements Methods Several methods have been employed for the measurement of operational efficiency, each of which has its advantages and disadvantages. Saad, Majid, Yusof, Duasa & Rahman (2006) pointed out that in the insurance industry, measurement of operational efficiency is mostly focused on two approaches, namely the parametric and non parametric methods. The most commonly used parametric approaches are the Stochastic Frontier Approach (SFA), Distribution Free approach (DFA), Thick Frontier Approach and Multiple regression. The most commonly used non parametric approaches are the Data Envelopment Analysis (DEA) and the Free Disposable Hull (FDH) (Cummins & Zi, 1998). Stochastic Frontier Approach (SFA) which is also referred as Economic Frontier Approach utilizes function form of cost, profit or production relationship on the inputs, outputs and environmental factors allowing for random error to follow a symmetric distribution, which is usually normal. It assumes composed error where inefficiencies follow an asymmetric distribution (Berger & Humphrey, 1997). Distribution Free Approach (DFA) utilizes the function form for the frontiers and separates the inefficiencies in the random error in a different way. DFA makes fewer assumptions and requires several years data. The stability of the firm is also assumed to be stable overtime. Multiple regression models expresses all outputs as a function of various inputs levels but does not produce overall measure of efficiency. Thick Frontier Approach (TFA) does not make any distributional assumptions for random error but specifies the functional form that assumes deviations from the predicted performance values in the highest and lowest quartiles. The observed deviation entails the random error. 14

25 Free Disposable Hull (FDH) is DEA case where the model does not assume possible substitution in terms of inputs combinations in an isoquant or efficient frontier which is a step function formed by intersection lines of combinations of inputs. An isoquant is a theoretical best smooth curve plotted with the minimum combination of two inputs and an organization producing at an isoquant point are deemed to be efficient (Bhagavath, 2009). The functions are used in estimating the firms distance from the optimizing envelope (Seale, 2000). DEA model assumes linear combination of inputs is possible in an isoquant. The approach obtains the frontiers of the inputs versus output ratios using linear programming methodology which combines all the outputs and inputs of the firm measured to single productive efficiency that lies between one for completely efficient firm and zero for completely inefficient firm (Saad & Idris, 2011). DEA model is capable of identifying a set of peers for organisations, determining relationships of data analysed, surplus or slacks variables of inputs and outputs to be used by firms and the relative efficiency score of firms. Therefore, DEA would be a better method to be used since it fulfils all the research objectives for the study to be achieved (Mburu, 2011; Mostafa, 2010) The Need for DEA Rao & Miller (2004) stated that efficiency is one of the most important performance measures of a business. Gandjour, Kleinschmit, Littmann, & Lauterbach (2002) concluded that many efficiency performance indicators used by business executives lacked in general validity. Therefore, there is need to come up with a recognized and valid measure of efficiency which is critical for managers and executives who are seeking 15

26 to increase the effectiveness of their organizations. Parametric methods like SFA, DFA and TFA have a disadvantage since they require prior assumption of functional relationship between the inputs and outputs to be used for any study unlike nonparametric method DEA which does not require such prior knowledge (Berger & Humprey, 1997). According to Oral & Yolalan (1990) DEA is complimentary to accounting ratios. The insurance industry in Kenya uses accounting ratios method to measure performance which has a limitation due to the multi-dimensional aspect of performance unlike frontier estimation methods like SFA and DEA which are more appropriate in single inputmultiple output and multiple input-output scenarios respectively (Tsolas & Giokas, 2012). Accounting ratio methods computes ratios by comparing single output to single input. Multiple inputs and outputs ratios can be computed by calculating multiples ratios of each DMU separately and then finally compares them to make conclusions. However, this makes it difficult to calculate overall efficiency of the firm. Therefore, an overall efficiency measure can be computed by consolidating all inputs and outputs by assigning weights to each input and output variable. DEA assigns weights to inputs and outputs using linear programming. No decision is normally made regarding the relative importance of each input and output because each operating units efficiency is compared to efficient operating unit rather than to average performance of the firms under study (Levary & Cesse, 2009). 16

27 2.3 Data Envelopment Analysis model The DEA technique was first developed by Charnes, Cooper, and Rhodes (CCR) in 1978 and the concept efficiency of frontier analysis was introduced by Farrell (1957). Charnes et al. s (1978) DEA model estimates efficiency under the assumption of return to scale while Banker, Charnes and Cooper (1984) assumed variable returns to scale. CCR model is an extension of the simple efficiency ratio of output(s) versus input(s) to scenarios with multiple inputs and outputs. Shim (2003) defined efficiency as the ratio of weighted sum of outputs to the weighted sum of inputs where the weights are calculated using mathematical programming which assigns weights to the inputs-outputs used in the study initially unknown by maximizing efficiency for each DMU. This approach uses linear programming to construct the frontiers of the input-output combinations. The Charnes, Cooper and Rhodes (CCR) model which was developed is as follows: Assuming that there are N number of DMU s to be studied, using m- inputs which will produce n-inputs. Therefore, for any given DMU B k, 1 k N inputs for B k = {X k1, X k2, X k2,.x km, }, the output matrix will be B k = {Y k1, Y k2, Y k2,.y kn, }, Therefore the relative efficiency of DMU B k, denoted by E k,would be Maximize E k = where k=1, 2, 3, 4..N Subject to, 1 k N 17

28 The weights, Vi, Uj, Ykj, Xki This model is commonly referred as CCR model. Since introduction of the basic CCR model, it has attracted attention of various researchers and different modifications have been developed from the first model. The first variation involved the standard constant return to scale (CRS) and variable return to Scale (VRS) models Grosskopf & Lovell that involved calculating technical and scale efficiencies (Fare,, 1994). The second models considered the extension of these models to account for allocative efficiencies and cost (Fare et al, 1994). Finally, the third option considered applying Malmquist DEA methods to calculate indices of total factor productivity (TFP) changes, technological changes, technical efficiencies changes and scale efficiency changes (Fare, Grosskopf, Norris & Zang, 1994) among many other methods Choice of Inputs and Outputs Variables Determining inputs and outputs is very important in service industries such as insurance. The choice of inputs and outputs affects the relative efficiency score obtained (Corton & Berg, 2009). Cummins and Weiss (2000) stated that the problem is very acute in service sector with many outputs which are intangible and their prices implicit because the results obtained can be misleading if they are not properly defined. Inappropriate choice 18

29 of variables can give inaccurate results and will not reflect the true relative operational efficiency. Great care must be taken to include all inputs and outputs that have impact on each insurance firm (Corton & Berg, 2009). Omitting of an input may disadvantage those insurance firms that are efficient in allocating them whereas omitting output may disadvantage those insurance firms that are efficient in producing them. Therefore, the quality and appropriateness of the data used are important as the method used this is because inefficiency of DMU can be caused by failure to include quality inputs and outputs (Coelli, Rao & Battesse, 2005). Insurance firms have many variables that can be used as inputs and outputs variables. The variables are both quantitative and qualitative. Some of the quantitative variables include investment income, net earned premiums, market share, commission expenses, asset admitted, incurred claims, other incomes, management expenses, interest payable to policy holders, other expenses and profit before taxation while some of qualitative variables are customer indexes and surveys. Alirezaee, Howland & Van De Panne, (1998) concluded that average DMU relative efficiency score is directly proportional to the number of input and outputs used for the study and it s inversely proportional to the number of DMU s. This implies that if there s positive correlation between the inputs and at least one output we can increase the efficiency score of each DMU by increasing the number of inputs and outputs. The more positive the correlation figure between the input and output the more accurate and reliable the efficiency score measured. 19

30 Murrey & Rowse (2006) stated that any input selected for study should not be negatively correlated with all the outputs to be used for study and every output selected should be positively related with at least one input to be used for the study. Therefore, all the variables which are negatively correlated in terms of input and output should be neglected when calculating efficiency scores of the insurance firms. Relative operational efficiency can give less reliable estimates of DMU s if the number of DMU s studied is not sufficiently. This problem is more pronounced when using stochastic approaches and the DEA model can also be affected by inadequate number of DMU s sample studied. However, a rule a thumb states that the number of DMU s studied should be at least four times the total number of inputs and outputs used (Rao, Kashani & Marie, 2010). According to Mehmet & Kale (2010) analysis of 49 prior studies the average number of inputs variables used was 3.9 and outputs was 4.7. Therefore, this study will comprise of 47 insurance companies in Kenya with 3 inputs and 4 outputs which satisfies both Rao et al. (2010) and Mehmet & Kale (2010) criteria Advantages and Limitations of Data Envelopment Analysis The main advantages of DEA are that it can incorporate multiple inputs and outputs, can identify a set of efficient frontiers (peers) that an organization can emulate, is capable of quantitatively analyzing different sources of data, it is useful in uncovering relationships difficult for other methods and there is no need to specify mathematical form for the production function to be used for measuring operational efficiency (Mostafa, 2010). However, like any other scientific method DEA has a number of disadvantages which need to be acknowledged when interpreting results of the study. Firstly, DEA is a 20

31 deterministic rather than statistical method and produces results that are sensitive to measurement error of variables. It measures efficiency relative to the best in the sample and it s not important to compare scores between two different studies. Secondly, DEA scores are sensitive to input and output used (Corton & Berg, 2009) and the size of the study sample (Bhagavath, 2009). Thirdly, the number of efficient firms tends to increase with the increase in number of inputs and outputs used for the study (Berg, 2010). Lastly, the linear program formulations obtained for each DMU under study makes manual solving very tedious and time consuming (Banker et al, 1984). 2.4 Efficiency Studies Diacon et al, (2002) and Rees & Kessner (2000) have conducted international studies in Europe involving comparison of the efficiency of firms in the insurance industry. Diacon et al. (2002) while studying insurance companies in the U.K., Spain, Denmark and Sweden found out that the U.K. insurers appeared to have low levels of scale and mix efficiency. Rees & Kessner (2000) in his study found out that the average efficiency level of German firms was about 48% compared to British firms whose level of efficiency was higher with mean of 57%. Dutta & Sengupta (2010) conducted a study to examine whether investment on Information Technology (IT) infrastructure which is a technological innovation has any positive impact on efficiency changes. 12 life insurances were used which covered the period to evaluate their efficiency scores using DEA and finally calculating scale efficiency. The study concluded that increasing investment in IT infrastructure has great positive impact on the technical and scale efficiency changes under constant and variable returns to scale assumptions. 21

32 Mansor & Radam (2000) studied the Malaysian Insurance Industry where they measured the productivity of life insurance in Malaysia using the non- parametric Malmquist index Approach. In measuring the efficiency performance they evaluated the Malmquist Index with a sample of 12 Malaysian insurance companies between the period 1987 to 1997 and found out that the overall productivity growth of insurance industry was attributed to both technical efficiency and technical progress. The variables which were used as inputs are claims, commissions, salaries, expenses and other costs while the outputs variables were new policy issued, premium and policy in force. The study recommended the use of information technology, human resource development, upgrading of distribution channels and rationalization efforts should be extended across countries to revitalize the insurance industry. In Kenya, Waweru (2007) did a study to establish the framework for the efficient and effective utilization of ICT in the insurance companies and brokers. The objective of the study was to evaluate the ICT in place to ascertain whether there was any form of IT governance in the insurance sector and to develop a framework model for efficient and effective use of ICT resources available. The study established that on average the sector has infrastructure but variances were noted between the two players with insurance companies ahead in software and hardware infrastructure as well a qualified staff in the ICT departments. The efficiency and effectiveness recommended could be leveraged by creating insurance community with similar or same ICT services. 22

33 Mburu (2011) conducted a similar study in Kenya measuring the operational efficiency of Kenya Commercial Bank using DEA model by using 168 branches of KCB for the period of January to December The main objectives of the study were to determine operational efficiency using Data envelopment analysis computer program and for each branch relative score, objective input and outputs targets, peers for each inefficient branch, slacks outputs units and excess inputs units and finally the relationship was determined between relative score and the location of the branch or asset size examined and relationship noted. Contrary to the management belief that the staff costs is the main cause of inefficiency, the study concluded that total asset input variable and interest receivable output variable are the main causes of inefficiency in KCB. The study recommended similar studies to be conducted to determine relative operational efficiency, determine banks benchmarks and why they do not apply scientific methods like DEA. 2.5 Summary From literature review it be noted that although numerous studies have attempted to assess the banks and insurance industry in Kenya, Europe and Malaysia among many other countries in the world, none of the studies has focused specifically on measuring operational efficiency of insurance firms using a scientific method like DEA in Kenya. In this study, the main aim is to fill this research gap by evaluating the operational efficiency of insurance industry in Kenya which is complimentary to accounting ratios (Oral and Yolalan, 1990). 23

34 CHAPTER THREE: RESEARCH METHODOLOGY 3.1 Introduction This study aimed at examining the relative operational efficiency of insurance industry in Kenya by means of DEA model. Prior studies have used different parametric and nonparametric methods to measure operational efficiency. This study used non-parametric DEA, a frontier analysis method to measure the relative operational efficiency because it provides an overall, objectively determined and efficiency numerical value. The most efficient firms are assigned a efficiency value of unity (100%) and lesser efficient firms a lesser numerical value of less than one and zero being the most inefficient firm(s). This chapter focuses on research design, population, data collection and data analysis. 3.2 Research Design A descriptive census survey research design was used to study the insurance industry in Kenya. Descriptive census research survey can use either quantitative or qualitative research methodology or both of them. This census survey used quantitative information from IRA report Descriptive census survey research involves gathering the data which describe events and organizes, tabulates, depicts and describes the data collection (Glass & Hopkins, 1984). Studies of this type are aimed at answering questions like what is, and survey methods are used to collect descriptive data (Borg & Gall, 1989) hence it can be able to describe the current operational efficiency state of the insurance industry in Kenya. Therefore, the design opted was able to sufficiently answer the research questions and meet the study objectives. 24

35 3.3 Population The study population consisted of all insurance companies operating in Kenya as at 31st December According to the Insurance Regulatory Authority (IRA) Report they are 47 in number. All the 47 insurance firms were studied because the population was considered small and data from the firms was manageable. The study utilized Insurance Regulatory Authority data at 31st December 2010 mainly because the Insurance Regulatory Authority (IRA) report data for the year 2011 had not been released during the time of study. 3.4 Data Collection The study used secondary data of each insurance firm performance in Kenya for the year The data was obtained from the IRA 2010 Report. The IRA obtains data from all insurance firms relating to their performance on yearly basis. The data is then analyzed and compiled to prepare the annual financial performance report for the whole insurance industry in Kenya. Therefore, there are minimal errors in terms of data collection because it s not manually collected and comes directly from the regulatory board of Insurance in Kenya. This study employed the following inputs: incurred claims, commission expenses, management expenses and total assets. The outputs used included; net income premiums, market share and investment income. The variables were obtained after intensive literature review of prior studies of Saad & Idris (2011) and Diacon (2001) that had used similar variables and showed that measurement of operational efficiency was successful. 25

36 3.5 Data Analysis DEA is incapable of handling negative data variables. Therefore, for consistency on all data variables the first step was to ensure that they are not null or negative by adding a constant to all the variables for all the DMU s (Mohamad & Said, 2010; Kristiaan & Ignace, 2009). The data collected was then analyzed using CCR model to calculate the relative operational efficiency for each DMU which computes the ratio of relative operational efficiency as the maximum weighted sum of outputs to inputs used for each the DMU under study. Using CCR model which was introduced earlier, assuming that there are N number of DMU s to be studied, using m-inputs which would produce n- inputs. Therefore, for any given DMU B k, 1 k N, inputs for the output B k ={X k1, X k2, Xk2,.X km, }, matrix would be B k = {Y k1, Y k2, Y k2,.y kn, }, Therefore the relative efficiency of DMU B k, denoted by E k would be Maximize E k = where k = 1, 2, 3, 4... N Subject to, 1 k N The weights Vi, Uj, Ykj, Xki 26

37 Where Vi represents the weight or value of contribution of one unit of output and Ui represent the weight or value of contribution of one unit of input. The constraints are added as follows, a) The efficiency of all DMU s should not exceed 100% i.e., E k this is mathematically expressed as, 1 k N, b) The weights Vi, Uj > 0, for all i =1, 2, 3, 4.m, j=1, 2, 3, 4 n c) The fractional programming above can be converted into a linear programming problem by add a constraint such that the denominator is equated to a unit i.e.. For each individual insurance firm k, a linear programming problem need to be formed separately and evaluated to obtain a relative operational score. However, solving the problems manually is very time wasting, tedious and one can get erroneous solutions. Therefore, the study used a DEA computer program DEAP version 2.1 (Coelli, 1996) to solve the linear programming problems. This computer program is very fast, accurate, efficient, has no maximum number of DMU s to be used, freely available and it has been used successfully in prior similar studies. The data obtained for the study was run with the specified format on the computer program and output results were able to answer the research questions posed. The results for each insurance firm (DMU) had a constant return to scale and variable return to scale 27

38 relative operational efficiency scores, peers for each insurance firm, slack variables measures and the maximum input-output target for each insurance firm. Finally, the relative efficiency score obtained for each insurance firm was correlated against its total admitted assets to check whether there was any relationship between them. 28

39 CHAPTER FOUR: DATA ANALYSIS, RESULTS AND DISCUSSIONS 4.1 Introduction This chapter presents and discusses the findings from data analysis undertaken in line with the research objectives. The data used was provided by Insurance Regulatory Authority which consisted of the performance data of Kenyan insurance companies for the year 2010 (Appendix 1). The research findings and discussions focused on the correlation analysis of chosen outputs and inputs variables, relative operational efficiency scores of each insurance firm and finally relationship between firm relative efficiency score to its total admitted assets. Out of 47 insurance firms in Kenya only 35 firms were considered for data analysis. This is because 2 (two) of the firms had extreme figures in terms of admitted assets and 10 (ten) of the firms did not have some of the variables to be used for this study limiting fair comparison. 4.2 Correlation Analysis of Chosen Output and Input Variables Measures of relative operational efficiency assume that there is a positive relationship between the chosen outputs and chosen inputs. Murrey & Rowse (2006) stated that there should be a positive correlation between each selected input and at least one selected output. Therefore, the analysis began by conducting an output-input correlation analysis between each output variable and each input variables. Table 4.1 summarises the findings of this analysis. 29