A Comparative Study on Efficiency Assessment between Public and Private Universities: Evidence from Tanzania

Our study builds on a twostage Data Envelopment Analysis to examine and compare the efficiency of public and private universities in Tanzania in 2008-2012. First, we use data envelopment analysis to measure the technical and scale efficiency of universities. Secondly, we examine factors that influence efficiency through Tobit regression model for both public and private Universities. The findings from the model affirm that the efficiency of the two University categories varies significantly. However, public University average efficiency is observed to be higher than that of private Universities. Whereas public Universities are inefficient in research and publications, private universities efficiency is determined by enrolment, academics staff and consultancy services. Given the contemporary significance of University education to social and individual development, we plead to the government University agency (TCU) to review the entire quality assurance process to improve Universities efficiency.

These sister policies, have accelerated the increase of Universities in Tanzania, from one University in 1970 to more than 40 Universities in 2014 (TCU, 2014). Eventually, the increase of Universities has also raised the enrolment. Between 2010 and 2011, for instance, there was an increase of enrolment by 49%. Despite this tremendous change, the efficiency and quality of Universities remain sceptical. Nevertheless, evidence from stakeholders' complaints about performance of graduates is vibrant. The incapability of University graduates has been reported, that they are unable to demonstrate the desired skills at the work place. In this scenario, organizations have undermined the role of Universities, and they are reluctant to employ University graduates (Bastos, 2011). To emphasize more, Msolla (2000) pointed out the prevailing complaint from the society about poor education quality. It was suggested to redress the situation through linking expansion of higher education with quality assurance of the education being offered for.
In line with this allegation, Makulilo (2013) also avows that proliferation of private Universities in Tanzania has led to the decline of education quality. However, this contradicts to Liu (1992) who states, ownership of an enterprise is empirically not a significant factor of performance. It is also unreasonably to relay on only one empirical study to reach conclusion, rather the same can be added to a body of evidence for further evaluation. Therefore, still more evidences are required to justify if private Universities are the major cause of education decline. Furthermore, Materu (2007) support Makulilo's assertion that entrance of private sectors and international providers of University education has decelerated both quality and efficiency of education. Additionally, financial constraints and human resource incapability are identified as the major causes of this tragedy.
On the other hand, Ferguson and Ferguson (1994), contradicts the hypothesis that public institutions are inherently poor performers than private enterprises. The reason provided to defend this stance is deficit of information driving to such conclusion. Moreover, most public enterprises do not have a specific benchmark for which their performances can be measured. Furthermore, their social activities are explicitly than that of private sectors. Thus, it is difficult to compare these two enterprises due to their varied nature of performance environment.
In a similar way, Thomas and Xiaoying (2009), highlights the impact of ownership on higher education institutions on performance and efficiency due to different institutional setting, priorities, perceptions, experience and others. Likewise, public and private Universities differ in various aspects such as resource capability, orientations in terms of education delivery if it is for service or business, management styles and the like (Alam, 2009). As result of these variations, efficiency variability could also be possible. This can only be realised through efficiency measurement of both.
Studies elsewhere have also revealed incongruity concerning higher education efficiency and quality. In Greece, Panagiotakopoulos (2012), while studying employability skills found that higher education institutions have failed to integrate desired skills in their curriculum, as result graduates are not esteemed in the world of work. Implicitly, Universities and their regulators do not play their role accordingly, and the future of University graduates is not determined in case the situation is not redressed. Even though, the study is not bias on whether public or private Universities are the source of the problem.
Notwithstanding scepticism about Universities, the argument that Universities should prepare employable graduates has no objection. As Fallows and Steven (2000), Hall et al 92009) and Harvey and Knight (1996) state, graduates j u n e 27, 2 0 1 4 employability skill competence, should be a paramount criteria which also need to be incorporated within a University curriculum.
Based on ongoing conflicting debate on whether public or private Universities are the source of the prevailing University education problem, this study examines the efficiency of both public and private University, and compares their efficiency. This paper is organized as following: In the second section, related literature review is provided, followed by methodological part in section three. Findings and discussions are provided in section four. Finally, section five presents conclusion and recommendations.

LITERATURE REVIEW
In economics term, efficiency is the ratio between the total output and total input. A firm is said to be efficient if it uses certain amount of input to produce more outputs, or spending less input resources to produce the same amount of outputs under given technology, as compared to other firms (Vincová, 2005 ). Therefore, University efficiency implies ability to use minimum education resources to produce desired outputs/outcomes such as research, consultancy services, and graduates, to name a few.
Measuring education efficiency has been very difficulty due to multiple inputs and outputs involved. In education the variable inputs are not directly related to outputs obtained. As result, efficiency assessment uses a proximal relationship between inputs and outputs. However, a non-parametric linear model, data envelopment analysis (DEA) has been in use for more than three decades in measuring the efficiency of various organizations, including education institutions. This is a linear programming paradigm which makes no prior assumption about variables (Aristovnik, 2013). Results from these studies reveal that there is variability of efficient among Universities. Furthermore, it is affirmed that an institution being efficiency in one aspect however, does not guarantee efficiency on another. Obviously, the efficiency or inefficiency will only base on involved inputs and output variables of a particular unit. Thus, an institution being efficient or inefficient is not a permanent behaviour; it is determined by particular environment in which an institution operates.
Moreover, Collier (2012) assessed the effect of institutional arrangement on student achievement in schools, using a two stage regression model and DEA model. The findings obtained suggest that there is a positive influence of institutional setting to student achievement. However, a slight relationship between educations' input like salaries and teachers on school efficient was observed. The use of DEA model is also emphasized by Abdulkareem and Oyeiran (2011) that Universities need to provide services to their clients while focusing on the market needs. They suggest that this can only be realised through DEA approach by analysing their efficiency.
In contrast, Ghana has reported a brain drain crisis focusing on Medical professional graduates (Dovlo and Nyonator, 2005). Findings from the study conducted in Sub-Saharan countries to detect the deficiency of medical Doctors reveal that most of professional Doctors seek jobs to abroad to earn more returns. Arguably, there variations of efficiency among University schools and departments in terms of graduate qualities. So, similar case should be expected when comparing public and private Universities efficiency.
Despite the overwhelming demand of graduates in the world of work, Universities are blamed for not preparing graduates who possess desired skills (Sutherland, 2008). Surprisingly, studies indicate policy successes while leaving graduates unemployed as they enter into the market without qualifications. To support the same, Ordonez (2012) states that Universities should train graduates who can be easily sucked by the global market. This will meet their dreams of working abroad and earning higher income. These arguments imply the challenges Universities face in terms of efficiency, as they cannot meet the social expectations. This has become a hurdle to most graduates who graduate without acquiring market desired skills.
In the same way, the current competition in higher education calls for constant measurement activities of these institutions to improve their efficiency (Nkirila, 2009). It is further stated that both public and private higher education institution needs to comply with government mandates, if they are to improve their efficiency. Therefore, Universities need to strike the balance between their increase in number and efficiency, while focusing on the three core objectives namely; knowledge transmission, knowledge creation and knowledge sharing to the society.
Many studies done in Tanzania focuses on other higher education such as quality, access, equity and financing of higher education, to name a few. The aspect of University efficiency is scantly discussed; for instance Kapinga and Dunrong (2010), their study determined reasons for higher education privatization and revealed that government fiscal incapacitation and delayed meeting social demand triggered privatization of higher education. Ishengoma (2008) assessed the trend of cost sharing higher education and realised that there were more challenges than solutions to arrest. Uneven distribution of funds and low capacity of the government to provide sufficient loans were some of the identified hurdles. A study by Makulilo (2012) on the proliferation of private universities in Tanzania focused on private Universities public Universities affirmed that private Universities were the major cause of low quality of University graduates. However, an interesting study by Kipesha (2013), done on public Universities efficiency, provide an impetus to the current, as it did not assess the efficiency of both University categories (public and private). None of the studies has compared the efficiency of public and private Universities for concrete inference. Therefore, we conduct this study to examine and compare the efficiency of both public and private Universities in Tanzania. j u n e 27, 2 0 1 4

METHODOLOGY
Data environment analysis (DEA) is powerful statistical tool which is broadly used to measure efficiency and performance of different decision making units (DMU). Usually, the investigated DMUs are characterized by multiple inputs and multiple outputs. Some of these variables are stochastic, making it difficult to execute other approaches than DEA model. Thus, with DEA model, DMUs transform multiple inputs into multiple outputs. The objective of DEA is to assess the efficiency of each comparable DMUs in relation to its peers. Additionally, DEA determines the sources of inefficiency and suggest solution for remedy.
The current study therefore, adopts an input oriented two stage DEA model. In stage one a technical efficiency and scale efficiency is obtained by performing two DEAs. The second stage uses Tobit regression to determine efficiency influencers. The first DEA, assumes that DMUs operates at constant return to scale (CRS), suggested by Charnes, Cooper and Rhodes (1978) CCR model. The second analysis is under assumption that the DMUs are operating at variable return to scale (VRS) as introduced by Barker et al (1984) namely; BCC model, modified from CCR model. Nevertheless, in real life optimal behaviour is unacceptable due to various factors such as, financial constraints, market competition, technological change and others (Cesaro, 2009). Thus, we prefer both CRS and VRS formulation due to the fact that public and private Universities relatively differ in terms of size, resources and orientations. This will facilitate to obtain the mean technical efficiency of both public and private Universities, and compare their means. Moreover, the decomposition of technical efficiency (TE) into scale efficiency (SE) is calculated given by:

SE
All CRS, VRS and SE can vary from zero and one. A score of one indicated by 100% implies a DMU is full efficient, whereas any score below that indicates inefficient of a DMUs.

CCR Model (CRS)
The original model was proposed by Charnes et al (1978). CCR introduced the following fractional programming problem to obtain values for input weights and output weights. Basic CCR formulation is; Where xij is the observed amount of input ith of the jth DMU (xij > 0, j = 1, 2 …n; i= 1, 2…m) and yij = observed amount of output of the rth type for the jth DMU (yij > 0, r = 1, 2…s; j = 1, 2…n)

The BCC Model (VRS)
As aforementioned, the CRS DEA model assumes that the DMUs are operating at an optimal scale. So, the implication of this model is that technical efficiency of DMUs does not vary. However, in real life, it is impractical. To remedy this j u n e 27, 2 0 1 4 Where xij is the observed amount of input ith of the jth DMU (xij > 0, j = 1, 2 …n, i= 1, 2…n) and yij = observed amount of output of the rth type for the jth DMU (yij > 0, r = 1, 2…3, j = 1, 2…n)

Tobit regression Model
Tobit regresssion model is a statistical non-linear model proposed by James Tobin to describe the relationship between a non-negative dependent variable Yi and an independent variable Xi. This model is used to examine the factors that influence efficiency for appropriate correction of inefficient DMUs (Foster, 2013). This is the second stage of efficiency analysis we implement a Tobit regression model. Therefore, Tobit regression equation created in this study has efficiency scores of each University as dependent variable and factors that are considered to influence the efficiency scores are independent variables. With Tobit regression results, a variable is considered significant if its P-value is (0.01  P  0.05 or t-value  |1.96|. We run two regressions to determine the factors influencing efficiency in public Universities as well as factors influencing efficiency in private Universities.

Data and variables
The Data employed in this study are both primary and secondary quantitative derived from multiple sources. Secondary data were derived from Tanzania Commission for Universities (TCU), National Bureau of Statistics (NBS) and Ministry of Education and Vocational Training (MoEVT), whereas primary data were obtained from 16 Universities in Tanzania those which were ready to provide data. The sample includes 8 public and 8 private Universities. However, we intended to use the entire population but unavailability of data is one of the limitations to this study.
We use a panel data for the period of 2008 to 2012. The selection of this period is based on two factors. First, this is the period when high enrolment was recorded in Tanzania Universities, resulting from massive increase of graduates in secondary schools and non-university colleges. The graduate increase in secondary schools was also accelerated by the secondary education development plan commenced in 2004(URT, 2004. Secondly, in 2010 MoEVT launched a higher education development plan (HEDP) to curb the increased enrolment in secondary schools, demanding for higher education (URT, 2010). Thus, it is worthwhile to assess the University efficiency to ascertain if the adapted plan is workable. So, we select inputs which are proximal to the selected output as shown in the table below due to complexity of educational variables. j u n e 27, 2 0 1 4

RESULTS AND ANALYSIS
This section, provide results from the two DEA modelling, for public Universities and private Universities. From the results, the technical efficiency and scale efficiency scores are examined, and their average means of the two University categories are compared. Furthermore, results from Tobit regression model are presented to ascertain the determinants of technical efficiency in public Universities as well as in private Universities.

Technical efficiency DEA results
After performing a two DEA analysis, we obtain mean technical efficiency under CRS and VRS assumptions. Scale efficiency (SE) is obtained from the ratio of CRS and VRS. Detailed output by institution is shown in Appendices A and B.
All score values for CRS, VRS and SE rows, varies from zero to one, where 100% implies technical efficient or scale efficient. Any efficiency scores less than 100%, indicates technical inefficient or scale inefficient of that DMU.

Average number of efficient Universities from 2008-2012
This is intended to observe the average number of efficient Universities, both public and private in each year of review, and detect if there is any variation between them. Table 3 below indicates the average number of efficiency Universities in the five years of review. Details of the results are in Appendix C, which is also extracted from Appendices A and B, indicating University efficiency scores modelled under CRS, VRS and SE.

Tobit regression for public Universities
We use Tobit regression model to examine the public University efficiency determining factors. The analysis is performed at 5% level of significance for both public and private Universities. In this case technical efficiencies modelled under VRS (Appendices A and B) are employed for both DEAs in public and private Universities.
Regression equation for public Universities is statically represented by; θPUi = β0PUi + β2ENPUi + β2ASPUi + β2NASPUi + β2GRPUi + β2RPPUi+ β2CSPUi + uPUi Where, θPUi is the efficiency of public Universities i, β0PUi is the constant term, ENPUi total enrolment in public Universities, NASPUi is non-academic staff public Universities, GRPUi number of graduates public Universities, RPPUi research publication public Universities, CSPUi consultancy services public Universities, β1PUi-β6PUi are coefficient of variable from 1 to 6 public Universities and uPUi is the error term in public Universities.  4 show that only research publication (RP) is significant at 5% level of significance and its relationship with efficiency is negative. This signifies that public Universities could reduce their efficiency by any increase of 0.03% of research and publication. Similarly, inefficient public Universities could increase their efficiency scores to 100% through reduction of research publications by the same quantity of 0.03%. So far, other variables have indicated no statistical significance to the efficiency of public Universities. The implication to this is that public Universities are operating at high average of efficiency. This is also supported by an average TE and SE in Table 2.

Tobit regression for private Universities
Regression equation for private Universities is statically represented by; δPRUi is the efficiency of private Universities i, β0PRUiis the constant term, ENPRUi total enrolment in private Universities, NASPRUi is non-academic staff private Universities, GRPRUi number of graduates public Universities, RPPRUi research publication private Universities, CSPUi consultancy services private Universities, β1PUi-β6PUi are coefficient of variable from 1 to 6 private Universities and uPUi is the error term in private Universities.  Table 5 above show that enrolment (EN), academic staff (AS) and consultancy services (CS) are statistical significant to private Universities efficiency. Whereas enrolment shows negative relationship with efficiency indicated by a negative sign coefficient, academic staff and consultancy services are positively related with the University efficiency. In this regard, inefficient private Universities could reduce the number of student enrolment by 0.01%, at the same time increase the number of academic staff by (0.20%) and number of consultancy services by (1.63%), if they are to become efficient. It is learned from the results that private Universities enrol more students over and above their capacity in other resources like academic staff. Thus, a balance between inputs and outputs could raise the efficiency.

DISCUSSIONS AND CONCLUSION
It is clearly stated earlier that assessing Universities' efficiency is a pervasive role of any government aspiring to attain sustainable development. This paper examines the public and private Universities efficiency, whereby a comparison between the two University categories is also done. The overall findings suggest that the there is statistical variation between mean technical and scale efficiency averages of public and private Universities in Tanzania. However, mean averages of public Universities are observed to be higher than that of private Universities, in all DEA modelling; CRS, VRS and SE. However, high variation is observed in CRS and a very low difference in SE. High variation in CRS results from the fact that CRS is not very practical due to variation factors hindering Universities not to operate under this assumption (Cesaro, 2009). Not only that but also, public Universities differs in terms of resources, experience, government support, management etc. All these, might have contributed to the observed results.
There is also an indication that public Universities have only one efficiency determinant to inefficient Universities, whereas there are three determinants for private Universities. Research publication in public Universities appear to be negatively related to efficiency. The implication to this is that inefficient Universities could reduce this factor by 0.03% to become efficient. Student enrolment is observed to be higher than the capability of the University to handle. Since tuition fee is the major source of running private Universities, it is clearly learned that raising fund through fees is paramount to private j u n e 27, 2 0 1 4 Universities. In so doing, causes efficiency to decline. Contrary to that, academic staff and consultancy services are found to be positively significant to University efficiency, hence requiring them to increase these variables.
Both hiring academic staff and conducting consultancy services are investments which consume University funds. Implicitly, private Universities do not capitalize on hiring more academic staff; rather they employ few, so as to incur less cost and maximize profit. To bridge the gap of academic staff they mainly depend on part time lecturers who are also cheaply paid. Likewise, private Universities do not invest much on consultancy, they mainly focus on teaching. This is in line with Hirshleifer (1980), argues, "According to the classical formulation, the aim of the firm as a decision-making agent is to maximize (economic) profit" (p. 265). In this case, there is no doubt that the orientation of private is for profit making while that of public Universities is for service delivery. Therefore, private Universities could increase these variables so as to elevate their efficiency.
Conclusively, we affirm that the existing efficiency variations between public and private Universities, is very significant and has a policy implication. The government through its agency (TCU) need to review its entire process of quality assurance to ensure that efficiency assessment is frequently and thoroughly done. All University stakeholders such as academic staff, entrepreneurs, parents, students, and alumna to name a few, could be involved to provide inputs on how efficiency can be attained and maintained. Nevertheless, this study precluded other exogenous factors which might have contributed to this finding. Consequently, we suggest another study of the same by increasing the sample of Universities and other qualitative variables which could come up with different findings.

ACKNOWLEDGEMENT
This study has been successful due to contributions of many people who assisted me in one-way or another. I wish to express my sincere appreciation to all. The fact that I can not mention each by name makes debt to them even bigger.