Efficiency of Micro and Small Enterprises in Ethiopia: Evidence from Data Envelope Analysis Model

This study measures the technical and scale efficiency of Micro and Small Enterprises (MSEs) and input slacks using Data Envelop Analysis (DEA) model and identifies the determinants of efficiencies of MSEs by employing ordinary least square ( OLS) econometrics model . A sample of 375 randomly selected MESs are included in the study. The study found that the average technical and scale efficiency of MSEs are relatively low; technical efficiency averaged at 30 percent and 38.4 percent under constant returns to scale (CRS) and variable returns to scale (VRS) assumptions, respectively. Besides, the overall average scale efficiency score of MSEs was estimated at 77.8 percent. The highest mean technical and scale efficiencies were registered in the construction (71.8 percent) and manufacturing (85.7 percent) sectors, respectively. Whereas, the lowest technical and scale efficiency goes to urban agriculture sector and service sector, with 38.9 percent and 67.2 percent, respectively. The level of inputs, enterprise age and sector, human capital, labor productivity variables significantly affect relative technical efficiency level of MSEs with different directions while variables such as start-up capital, gender of the enterprise manager and availability of support from the government identified statistically not significant in determining the MSEs’ technical efficiency.


Introduction
Micro and Small Enterprises (MSEs) has been recognized as a key component of many nations' economy. Development of MSEs is believed to be the major source of economic diversification, employment creation and income generation, particularly for urban youths who are engaged in informal sector of the economy. More importantly, MSEs facilitate the economic transition of countries without demanding skilled labor, large capital and advanced technological breakthrough (Minh et al., 2007). In addition, MSEs can innovate, adopt new technology and know-how, broaden the tax base, and diversify risk (Brixiova, 2009). Furthermore, the development of MSE permits greater decentralization, reduce income inequality, and mobilize latent entrepreneurs (Young, 1994).
The current government of Ethiopia, recognizing the significant contribution of this sector to overall development, showed its dedication to accelerate the development of MSEs through the issuance of National Micro and Small Enterprises Development Strategy in 1997 and the Establishment of the Federal Micro and Small Enterprises Development Agency (Mulu G., 2007).
Ethiopia's industrial development strategy issued in 2003 also singled out the promotion of MSE development as one of the important instruments to create productive and dynamic private sector (Ageba and Ameha, 2004;cited from Mulu G., 2007).Thus, measuring efficiency of the sector will be important to the government, policy makers, and MSEs.
Number of employees, capital investment and scale of operation can be used to define micro, small, medium and large scale enterprises. In the case of Ethiopia, there is lack of uniform definition at the national level to have a common understanding of the MSEs. While the definition by Ministry of Trade and Industry (MoTI, 1997) uses capital investment, the Central Statistical Authority (CSA, 2008) uses employment and favors capital intensive technologies as a yardstick (CLEP, 2006).
Despite their momentous contribution to economic development and poverty reduction, MSEs are also circled by various challenges, particularly in developing countries. The most often mentioned challenge is access to finance. There exists a general consensus among policy makers, scholar and business experts that MSEs faced with financial constraints (Akoten et al., 2006;Ishengoma & Kappel, 2008;Beck & Demirguc-Kunt, 2006). Hussain et al. (2012) has also accredited the human resource and technological capabilities as major determinants of MSEs' growth. Moreover, Sarwoko & Frisdiantara (2016) identified personal, managerial and environmental factors as determinants of MSEs' growth. Technical inefficiency has been also mentioned as significant contributor to low productivity, poor performance and slower growth of MSEs while improving the production efficiency of these business entities has been considered as the way-out mechanism. Moreover, the current state of world economy coupled with dynamism in technological progress forced enterprise to be more efficient to remain competitive and profitable in the market.
Productivity growth is a precondition for increasing people's living standards and maintaining competitiveness in the economy. Low total factor productivity is the key reason for persistent poverty in developing countries. Moreover, firms need to be productive enough in their operation. Firm level productivity in turn guides their survival and growth rate. Productivity and firm growth, on the other hand, determined by many variables; some are firm specific, some industry or sector specific while some of the determinant factors are economy wide. Some literature, (Gebreeyesus 2007;Page and Söderbom, 2012;Bigsten and Gebreeyesus 2007), indicate that MSEs grow faster than medium and large enterprises, though others found positive relationship between firm growth and size (Shiferaw 2012;Bewen et al. 2009). For instance, a study by Gebreeyesus (2007) based up on a survey of 972 MSEs in selected cities of Ethiopia found that firm's initial size and age are inversely related with growth providing evidence that smaller and younger firms grow faster than larger and older firms. Firm size in turn may relate with productivity. Bewen et al. (2009) has also found that three out five MSEs failed within the first few months of establishment, highlighted the positive and significant impact of age for firm growth.
Similarly, firm size explains MSEs' productivity and efficiency level. According to Page and Söderbom (2012), the average worker in a 160-worker firm produces as much value-added in 15 minutes as the average worker in a 5-worker enterprise does in an hour. The study also found the existence of strong positive relationship between value-added per employee and firm size. Firms with 30 employees have, on average, twice as much value-added per worker as firms with 5 employees; hence the size-productivity differential is very pronounced, even amongst small firms.
Financial constraint is the other major determinants of SME growth. Theoretically, it is true that firms can finance their business activities both from internal and external sources. However, many firms in developing countries (micro and small businesses in particular) cannot resort to capital markets the firms are not in a position to access such markets (Gebrehiwot & Wolday, 2006;Eshetu & Mammo, 2009;Anthony & Thomas, 2012). Formal financial institutions excluded MSEs because either the enterprises couldn't fulfill the bank's lending requirements or the banking sector considers these business firms as risky (Gebeyehu, 2002). Sometimes it is visible that formal financial institutions favor large and medium enterprises though their policies do not explicitly stated in such a way (Tybout, 1999). Access to credit for MSEs is limited due to unclear property rights and lack of assets that can be used as collateral (Lutz et al., 2011). This is empirically supported by Page and Söderbom (2012) in which out of an estimated 365-445 million, formal and informal MSEs in the developing world, approximately 70 per cent report that they do not use any external financing, although they would do so if financing were available. Besides, Brown & Earle (2004) indicated the existence of strong evidence that access to external credit increases the growth of both employment and sales.
On the other hand, some study results indicate that access to credit is not a significant determinant of small firm growth; instead, other observable and unobservable characteristics of firms appear to cause growth (Bewen et al., 2009;Bari 2005;Beck 2007). Business experience, education, location of the business to traditional market places, possession business license and male headedness relate positively to the growth of MSEs (Gebreeyesus, 2007). Competition among themselves and from large firms, cheap imports, insecurity and debt collection also hinder growth of MSEs (Bowen et al., 2009). On the other hand, Anderson & Tell A study by Le & Harvie (2010) examined the performance of manufacturing small and medium enterprises in Vietnam. Specifically, it evaluates firm level technical efficiency and identifies the determinants of technical efficiency using Stochastic Frontier Production (SFP) approach for 5,204 enterprises from three surveys. The results from the econometrics estimations reveal that manufacturing firm in Vietnam have relatively high average technical efficiency ranging from 84.2 percent to 92.5 percent. The study further examines the factors influencing efficiency and found that firm age, size, location, ownership, cooperation with a foreign partner, subcontracting, product innovation, competition, and government assistance are significantly related to technical efficiency, with varying degrees and directions.
Another study by Alvarez & Crespi (2001) tried to identify the basic determinants of efficiency among Chilean small manufacturing enterprises using plant survey data and non-parametric deterministic frontier methodology. They found that efficiency is positively associated with the experience of workers, modernization of physical capital and innovation in products. In contrast, they also found that variables such as outward orientation, owner education and participation in some public programs had no statistically significant impact on efficiency of firms.
Using a Data Envelope Analysis model, Heilbrunn et al. (2011) has also identified firm level efficiency of MSEs and level of input slacks which intern lead to technical inefficiency. The result revealed that 89 out of 248 SMEs were found to be on the frontier. Besides, financial management variables were the main causes of inefficiency amongst MSEs in the study. Leza et al. (2016), employing DEA model applied to 354 MSEs in Wolayita zone of Ethiopia, found a mean efficiency score of 0.61 which was disaggregated to 0.59 and .0.67 for micro and small scale enterprises, respectively. Meanwhile, promoter's age, social networking, initial capital, vocational training and investment in ICT were identified as significant and positive determinants of technical efficiency.
Since the seminal work of Farrel (1957), comparing the productivity and efficiency level of Decision Making Units (DMUs) has been recognized as one area of research. Consequently, significant bodies of empirical studies were conducted on technical efficiency and determinants of efficiency. Even though MSEs efficiency has been studied by different researchers in some sectors of the economy, to the best of the researchers' knowledge, there are scant of scholarly researches undertaken so far on efficiency of MSEs in Ethiopia, of which were Leza et al. (2016) and Gemechu & Fitwi (2016).
MSEs Efficiency, however, is a relative concept which needs area specific study. Therefore, this research aimed at measuring the efficiency and the determinants of efficiency among MSEs in East Gojjam zone, Ethiopia and emphasizing on firm specific determinants of efficiency. The efficiency of MSEs examined using DEA model and the determinants of efficiency by Ordinary Least Square (OLS) estimation. The model estimation results revealed that most MSEs in East Gojjam zone were found to be relatively inefficient, operated below the production frontier and exhibited huge potential of productivity through increasing their efficiency. Moreover, labor productivity, form of ownership, enterprise age and sector effects were found to be significant determinants of MSEs' technical efficiency, whereas, the effect of location is asymmetric.
The rest of this paper is organized as follows: Section II presents the data and research methodology adopted in the study. Section three of the paper deals with results and discussion.
Finally, the conclusion and policy implication part is presented in Section four.

Data and Methodology
The study was based on primary survey data collected from MSEs in East Gojjan Zone.
Structured questionnaires were distributed to owners/manager of MSEs by the researchers themselves and data enumerators. Secondary data were also obtained from Office of Trade and Industry and MSE Offices of the Zonal and Woreda 1 administrations.
The actual sample size of any study depends up on the total population, the level of precision to be achieved, the degree of homogeneity, the research budget and the available time to accomplish the study. Hence, using the formula set by Yamane (1967) The study employed three stage sampling technique. At the first stage, sample towns were selected from East Gojjam Zone purposively, namely; Debre Markos, Amanual, Debre Elais, Bichena, Dejen and Mota, based on their MSE size. In the second stage, using business types as strata, sample MSEs were selected proportionately from each town and sector. Finally, Sample MSEs were selected with in 10 th intervals as per the firm list obtained from the MSE development offices. MSEs which were operational for at least one year period and still being in operation at the time of the survey were considered. Enumerators were also selected from respective sampled town's MSE development offices and training was given regarding the questionnaire.

The Empirical Models The DEA Model
To estimate the efficiency level of DMUs the study employed a DEA model as developed by Charnes et al. (1978). This model is an input-oriented with constant returns to scale assumption, implying that all DMUs are deemed to operate at their optimal scale. Imperfect completion, financial constraints and others, however, forced business firms to operate suboptimally. The Banker et al. (1984) model on the other hand is an extension of the CRS-DEA model which accounts the VRS assumptions. This assumption of the DEA model enables us to isolate pure TE form the scale efficiency component (Coelli, 1996).
This method is a multi-factor, productivity analysis for measuring the relative efficiencies of DMUs (Ayaz and Alptekin, 2012). It is a non-parametric method of estimating production frontiers. DEA measures the efficiency of DMUs by making comparison with the best producer in the sample to derive relative efficiency scores. Based on the DEA model, technical and scale efficiency scores are measured relative to the production frontier; MSE is said to be efficient if it operating on the production frontier. Meanwhile MSEs are inefficient if it fails to achieve the maximum output from a given level of input(s) or operating below the production frontier.
Therefore, DMUs are ranked based on their distances from the efficient DMU(s) or the production frontier. The production frontier of the most efficient firm, however, is not known in practice, rather estimated on a sample DMUs (Coelli, 1996).
The input-oriented DEA model for multiple outputs and inputs case is expressed as: For DMUs, each with inputs and n outputs, the relative efficiency score for DMU is obtained by solving the following linear programming model as proposed by Banker et al. (1984): Where: ≥ 1 ∀ , ; k= 1,…, n; j= 1, …, m and i= 1, …, s = the amount of output k produced by = the amount of input j used by and = weights given to output k and input j, respectively.
For DEA model, the following output and input variables were used to measure relative efficiency of MSEs.

Output (Y):
Yearly values (in Birr) of total sales or revenue generated by MSEs were used as a proxy to measure the output of that particular enterprise. DEA has some advantages over the SFP efficiency estimation model. First, it is not necessary to assume a specific functional form for the production function. Second, it makes no priori distinction between the relative importance of outputs and inputs considered as relevant in firm decision-making process. Third, DEA is relatively insensitive to model specification because the efficiency measurement is similar, regardless of input oriented or output-oriented measurements (Minh, 2007).

The Econometric Model
Determinates of MSE's efficiency was estimated using OLS econometric model. As the value of the dependent variable, MSE's technical efficiency score, is continues, OLS estimation provide BLUE estimate of parameters amid the fulfillment of all the classical linear regression assumptions. Following Greene (2002) and Maddala (1983)

Descriptive Statistics
Description of enterprise in terms of their initial and current capital, age, and number of employees is indicated in the table below. It shows mean and standard devotion of the aforementioned variables across sectors and towns. The average start-up capital required to establish and run a new MSE in the study area was about Birr 12,000. The construction sector was found to be the most capital intensive sector while people engaged in service sector needed relatively low start-up capital. Standard deviation figures have also revealed important information; the capital requirement to start a small business varied significantly across sectors and administrative towns.
Moreover, MSEs' capital base has been exhibited dramatic increment during the average four years of their operation, jammed on average by 400% to about Birr 82, 000. Among economic sectors, the construction and manufacturing sectors showed remarkable growth in capital.
Likewise, the standard deviation figures of the current capital indicated gigantic variations among business types and towns. Moreover, the median age of MSEs was about 4 years. Sector wise, urban agriculture and manufacturing type of business incorporated older firms while firms in the construction sector have relatively young firms. One of the economic importance of MSEs is that they can create employment opportunities, particularly for marginalized and less privileged part of the society. In this regard, the mean employment status of MSEs was 2 people; the construction and manufacturing sectors create more job opportunities compared to other line of businesses. MSEs engaged in trading activities, however, need less workers and mostly run by the owner and/or family members.

Technical Efficiency of DMUs
One measure of efficiency of DMUs is their technical efficiency, which can be measured as the ratio of weighted outputs to inputs. Those enterprises which operate on the production frontier are considered as the best performers in terms of technical efficiency and have technical efficiency score of 100 percent (Coelli, 1996). Table 2 below presents a summary of technical efficiency results of the input oriented 2 DEA model for both CRS and VRS assumptions over different business types or economic sectors.
As indicated in the above table, majority of MSEs was found to be relatively technically inefficient; about 96 percent and 92.5 percent MSEs operated below the production frontier under CRS and VRS assumptions, respectively.  Overall, estimation results confirm that there is huge potential for MSEs to increase productivity through enhancing their technical efficiency. Besides, the technical efficiency scores are expressed in relative terms, indicating MSEs have mammoth potential to be relatively efficient as the better performers, those MSEs operated close to the production frontier, before even struggling for absolute efficiency.

Scale Efficiency
On the other hand, researchers estimate scale efficiency of DMUs to label them as efficient and inefficient in their production process. According to Coelli (1996), measuring scale efficiency is vital when we assume that MSEs were constrained by many factors (such as finance or policy) and unable to operate at their full capacity. Once we obtained technical efficiency using constant returns and variable returns assumptions, it is an easy task to drive scale efficiency of enterprises. Technically speaking, scale efficiency is the ratio of technical efficiency under CRS to technical efficiency under VRS. were identified as relatively scale inefficient. Comparisons among MSE sectors showed that the construction sector has relatively lowest proportion (54.5 percent) of inefficient enterprises while this proportion is relatively high for trade and manufacturing sectors, 94.5 percent in each. Moreover, the mean scale efficiency level of MSE estimated at 77.8 percent; the highest score registered in the manufacturing sector (85.7 percent) and the lowest in the trade sector.

Econometric Estimation Results
Before conducting the econometric estimation of determinants of efficiency among MSEs, cleaning of data to make it ready for estimation purpose was done. found to be significant determinants of efficiency, whereas gender of enterprise and availability of support from the government have no any statistical significant effect.

Conclusion
In most economies, particularly in developing countries, MSEs are recognized as important contributors to income generation and diversification, employment and poverty reduction.