The major bottlenecks of Micro and Small Scale Enterprises’ growth and alternative strategies in Ethiopia:

Econometric and CGE analysis

Final Report First Draft

Presented to

Partnership for Economic Policy (PEP)

By

Ermias Engida

&

Mekdim Dereje, Ibrahim Worku, Saba Yifredew and Feiruz Yimer

ETHIOPIA

April 21, 2016

1. Abstract

Given the fact that Micro and Small scale enterprises (MSEs) are put to be high

on the agenda of the Ethiopian government’s mid-term growth and

transformation plan (GTP), this study aims at investigating the major bottlenecks

and contributions of the sector. We particularly seek to answer two research

questions using two different data sets applying two different but interrelated

techniques. First, we aim to examine the factors that constrain enterprise growth

as given by either capital or employment growth. Second, we seek to assess the

role of MSEs to reduce unemployment and poverty. From our simulation results,

the government is somehow achieving success on both employment creation and

poverty reduction. Despite an increase in female unemployment, reduction in

male unemployment is registered from the current performance of the

development plan. Besides, poverty is found to be declining in urban areas.

However, if the government endures for the plan’s complete accomplishment,

then there will be a nationwide poverty reduction, especially in the urban

centers. Unemployment for both male and female workers and poverty in both

areas would decline significantly if the plan gets complete accomplishment at the

end of the day.

2. Introduction In developing countries, the development of MSEs is taken as a key strategy for job

creation, poverty alleviation and more generally support economic development.

According to some studies, for example, the contribution of MSEs along with medium

enterprises account for about 30% of employment and 17% of GDP in developing

countries (Beck & Demirguc-kunt, 2005). In developed countries, the share of the

enterprises is even larger accounting, on average about 50% to GDP and 60% to

employment. Thus, naturally, as economies grow, the share and contribution of these

enterprises in the economies of developing countries will improve. In these

economies, the expansion of these enterprises is doubly important as they are closely

associated to the relatively poor and especially so with disadvantaged groups of

women and youth (Robu M., 2013).

The growth of MSEs also serves as a link between financial development and poverty

reduction. There is a high correlation between the degree of poverty, unemployment,

economic wellbeing /standard of living of the citizens of countries and the degree of

vibrancy of the respective country’s MSEs. In country like Ethiopia MSEs face a host

of internal and external factors that act as a barrier to enter in to business and to

perform well thereafter.

Micro and small enterprises are playing significant role in the Ethiopian economy

also. According to the 2002 nationwide survey of the Central Statistics Authority

(CSA), in Ethiopia there were 974,676 cottage/handicraft manufacturing

establishments engaging more than 1.3 million people. The Small Scale

Manufacturing Survey (CSA 2003) also shows that there were 31,863 small-scale

manufacturing industries (of which, 62.8 per cent were in urban areas) engaging

97,782 persons (91.3 per cent male, and 8.7 per cent female). The informal sector, in

which most of the MSEs lay, is a large source of employment and livelihood for the

urban population particularly. About 25% of people who are working in Ethiopia are

engaged in the informal sector in which women account 33.6% and men only 18.7%.

Even if this sector is creating job opportunities for women, there is still gender bias in

the major cities of the country. CSA (2014) indicates that the unemployment rate of

women is 28% while that of male is 13.8%. This shows us the importance of the

informal sector in enrolling women in income generating activities. Despite this,

however, the Ethiopian MSE sector has not been adequately studied empirically.

The government of Ethiopia has placed considerable importance to the role of these

enterprises in the economy’s commendable performance as well as the potential of

the sub-sector to transform the economy. The 2010/11-2014/15 Mid-term plan of

the Ethiopian government which is called the Growth and Transformation Plan (GTP,

2011/12), for instance, envisages that, during this period, MSEs create employment

opportunities for about three million people and thereby enhance income and

domestic saving, so as to reduce unemployment and poverty; particularly to benefit

women from the sector (MoFED, 2014).

However, neither the growth of these enterprises nor their contribution to their

primary target has been considerable. According to the recent survey conducted by

Ministry of urban Development and Construction (MUDC), the majority of the existing

MSEs are owned by male (59.6% in Addis Ababa); contribute very little to employment

creation (more than 50% hiring two or fewer) and most (about 60%) have less than 1

year experience (MUDC, 2013). Figure 2.1 below also confirms, using growth in

capital, that the growth of enterprises is very stagnant, probably except for the

construction sector.

Figure 2.1: Growth in real capital by activity

Source: Authors' computation based on MUDC, 2012.

Literature identifies, inter alia, size of start-up capital, access to credit, access to

work premise, access to appropriate training, access to or linkages to markets,

and schooling and other demographic characteristics of operators; as main

factors determining the performance of the enterprises in the country (See for

instance, MUDC, 2012; Gebreeyesus, 2007; Gebremichael, 2014). But most of

these studies are either only descriptive, not exhaustive, focus only on single

location, methodologically faulty or a combination of these.

In this study, we primarily focus on exploring the determining factors of the

annual growth of these enterprises based on a very rich data set collected by

MUDC in 2012 employing pertinent econometric technique. Besides, a CGE

model is developed for Ethiopia to investigate the role of MSEs on unemployment

and poverty reduction at a country level.

0.1

.2.3

Aver

age r

eal c

apita

l gro

wth

Metal & wood work

Constructio

n

Agro-processing

Textile & garm

ent

Leather & fo

otwareRetail

Urban agriculture

Food processing

Others

This research work is significant both due to the unique data it employs and the

novelty of the econometric and CGE techniques it applies. The completeness of

the data in terms of covering key issues of MSEs and its spatial coverage offers

the opportunity to critically assess the proposed research questions both at

national and across the regions.

The technical contribution comes from additions and modifications provided by

both the econometric and the CGE model. With regard to the econometrics part,

we have tried to establish causal relationship between our explanatory variables

of initial capital and capital growth vis a vis a host of firm specific and location

factors. In addition results for important elasticities such as training to labor

productivity for the CGE have been derived using econometric techniques.

The technical contribution also comes from special features of and modifications

in the SAM and the CGE model. The SAM used in this study has some special

features. Even on top of that, the SAM has some modifications to explicitly

incorporate MSE activities. Besides, the labor force is disaggregated by gender

and skill level (skilled/unskilled). There are also additional equations developed

in the CGE model to incorporate this disaggregation in the labor market and

production of a single product from two activities using different technology.

This gives us the luxury to come up with very original and unique analysis that

meets the intended objectives of this research work. This enables us to show and

send very important and timely policy recommendation messages about the

gender and occupation based alternative strategies towards realizing the planned

role of MSE development in the Ethiopian economy.

As a result of all these modifications and use of new and recent data, we come up

with robust results that are relevant from policy perspective. This study is very

timely, because the government of Ethiopia is expecting a lot from the MSE sector

which is not yet developed. There are a lot of road blocks on the sector which this

study reveals. This study also recommends policy options which can help to

improve growth and efficiency of the sector so that it can live up to the envisioned

goal.

3. Significance and objectives

Micro and small enterprises have great potential to create employment and

incomes among the disadvantaged and vulnerable. In order not to achieve these

potential, they face several constraints one of which is the lack of access to start-

up capital. This problem is particularly severe in developing countries where

financial institutions are weaker; the competition for limited credit access is throat

cutting and MSEs lack the necessary skill and financial capability to prosper on

their own. The availability of this capital to entrepreneurs determines the capacity

of the country to create jobs, reduce unemployment, spur growth and reduce

poverty.

The second factor that limits the contribution of MSEs to employment and

economic growth is slow or stagnant growth in their capital. Growth of the firm as

given by growth of its capital somehow indicates the extent of the firm to create

employment, absorb risk, and contribute to national output. This, however, could

be affected by several factors including the business environment, entrepreneurial

capability, beginning capacity, education of the owner(s)/manager, experience of

the owner (s)/managers, etc. For instance, regardless of the importance of credit

for the success of MSEs, about 80 percent of MSEs state that access to credit from

formal financial institutions is their number one constraint. This forces a large

proportion (41%) of respondents to start business with their own limited saving.

Besides limiting their capacity, this forces them to resort to sources that require

extremely large interest rate (Bekele & Muchie, 2009). According to this study,

money obtained from relatives and friends, iqqub schemes (social capital) account

for 18 percent and 12 percent, respectively. The share of microfinance institutions,

banks and private money lenders account for 9.2 percent, 8 percent and 7.2

percent, respectively. The story is exactly the same once the enterprises are

operational. Figure 3.1 below shows the sources of MSEs’ finance and percentage

of MSEs reporting them as primary in meeting their working capital needs.

Figure 3.1: Sources of finance and percentage of MSEs reporting them as primary in meeting their working capital needs

Source: Gebrehiwot and Wolday, 2006

In Ethiopia, a thorough examination of the determinants of growth of MSEs is not

undertaken. We therefore, attempt to critically examine the determinants of MSEs’

growth employing appropriate statistical and econometric techniques.

The role of MSEs to employment creation and overall economic growth is strongly

emphasized in the literature (See for example Daniels, 1999; Beck et al 2005). For

instance, MSEs enhance competition and entrepreneurship and hence have

external benefits on economy-wide efficiency, innovation, and aggregate productivity

growth. Since they require relatively less financial and human capital, and are more

labor intensive, they especially appeal more to the poor and the vulnerable. Hence

the expansion of these enterprises could serve as a powerful tool to reduce poverty

and spur economic growth without worsening income inequality. In Ethiopia, there

are few studies that rigorously examined the role of MSEs on employment,

economic growth and poverty. In this study, we aim to address this issue using

CGE modeling.

0

10

20

30

40

50

60

70

80

90

saving/retainedearning

Formal borrowing Informalborrowing

Supplier credit Client's advance Remittance

Specifically, this study addresses the following two research questions:

A. What factors determine and/or constrain the growth of MSEs in Ethiopia? Is there any difference in the growth of MSEs operated by female owners vis-à-vis male owners, how about youth and matured owners? B. What is the role of MSEs - on reducing unemployment and poverty in Ethiopia, in particular to disadvantaged and vulnerable groups (women)? The MSEs Sub-sector has huge potential for economic growth, reducing the

high unemployment rate and poverty. Understanding this potential, the

government of Ethiopia has emphasized the development of MSEs as one of the

instruments to achieve the planned broad based growth in the country. The

mid-term economic plans; the currently concluded GTP I and the ongoing GTP II

are underlining the importance of this sub-sector for the realization of

government’s plan to enable the country become one of the middle income

countries in the coming few years. However, a study based on a recent survey

reveals that the sector is not growing as planned. Besides, the GTP annual

progress report also mentioned that expansion of the MSEs and growth in

productivity and competitiveness of the already existing enterprises are still

lower than the envisaged targets. Therefore, by pinpointing potential areas that

could maximize the contribution of the sub-sector, the finding of this study

would serve as input to foster the contribution of the sub-sector to the

envisaged long term goal. The study assesses the effect of public investment on

different areas like giving training and better access to working capital that can

improve MSEs’ productivity and competitiveness. Besides, teasing out the major

constraints and forwarding possible recommendations to rectify the problems,

the study would contribute to enhance the welfare of those that engage and

heavily depend on them; the poor and the disadvantaged.

4. Methodology

In order to address the aforementioned objectives, we use econometric and CGE

models. The econometric model helps us investigate the determinants of the growth

of MSEs in Ethiopia. It is more appropriate methodology to dig into what determine

growth of MSEs. As per our objective, we also want to see how MSEs operated by

women owners perform relative to their men counterparts. In order to see if there is

difference in the performance of MSEs based on age and gender, we include age and

gender of the owner as explanatory variables in the econometric analysis.

On the contrary, CGE model is found to better fit the second research question given

the availability of data. Therefore, CGE modelling approach is mainly used to

investigate the role of MSEs’ development on reducing unemployment and poverty in

Ethiopia, specifically for disadvantaged sections of the society: women. Besides, the

CGE part deals with the effect of public investment on MSEs to come up with the

planned productivity and competitiveness of the sector.

4.1 Econometric Model

We have made a detailed survey of the existing literature regarding the operation of

MSEs. Tadesse (2014) descriptively analyzed how access to finance affects the sector

development in one town. Zemenu and Mohammednur (2014) also used similar

analysis to understand determinants of the growth of MSEs operating in another

town. Both studies have methodological limitation that emanate from relying on

descriptive analysis as this does not take all the growth determining factors into

account and unable to make any casual inference.

The literature identifies a range of factors that play important role in shaping the

growth performance of a particular MSE, by influencing the opportunities available

to owners and employees and their capabilities to take advantage of such

opportunities. Nichter and Goldmark (2005) illustrates that there are host of

individual, firm, social and business environment factors that shape and determine

the growth of MSEs. This study focuses mainly on firm and individual factors that

can constrain MSEs’ growth as data on business environment and social factors is

hard to come by in the Ethiopia case.

Ageba and Ameha (2006) also descriptively analyzed how MSEs are financed in

Ethiopia. Although in-depth analysis is given on the issue and has a wide coverage,

it suffers from the same problem as the above study.

Tefera et.al. (2013) econometrically estimated the growth determinant factors of

MSEs in one town of Ethiopia. The authors used probit estimation of the likelihood

of MSEs to grow given a number of covariates. The study used growth in

employment as an indicator of growth; firms with negative and zero employment

growth rate are given zero value while the remaining are given one value. By doing so

dichotomous variable is generated to capture MSEs growth. Gender of the owner,

initial investment size, location of enterprise relative to road and the enterprise’s

sector of operation are considered to be control variables. Although there is

theoretical justification for relying on employment, creating clear dichotomy is an

imposition or creation of a data structure which might not be the case. We also

think that a number of growth explanatory variables are missing in the analysis.

Similar methodology is followed by Gebremichael (2014) to look at the impact of

subsidy on the growth of MSEs.

In order to analyze the growth of MSEs, we have adopted firms’ growth model

provided by Evans (1986) and later adopted by a number of studies; Gebreyesus

(2007), Garoma (2012), Hagos et.al. (2014) and others to analyze MSEs’ growth.

Depending on the data set they are working on, the authors have chosen either

employment, sales, profits or fixed asset, to measure firms’ growth.

In our study we used firms’ capital as an indicator to capture growth. The major

drawback that is highlighted in the studies using capital as a measure of growth is

inflation. In our study, we account for inflation by converting the nominal figures to

real using national deflator. Following the works of Evans (1987) as cited in

McPherson (Mcpherson, 1996) who used similar approach to measure growth

indicator, the growth in capital can be measured by taking the real (real meaning

after adjusting for inflation) difference in capital at the start of operation with

current capital and taking it as a ratio of year of operation.

Growth of capital = ln (current capita) − ln (initial capita)

Firm’s age… … … … … … … … … …. (1)

A number of factors come into interplay to determine firms’ growth. Nichter and

Goldmark (2009) identified education of owner- although education has a certain

threshold level, work experience, firms age, formality, location of the firm- in house

or around the market-, access to finance, value chains and intra-firm cooperation to

be some of the factors that determine the growth of firms in developing countries.

Mcpherson (1994), doing cross country analysis, identified firms’ age, firms’ size,

sector of employment, whether the firm is located in commercial district or

traditional market place, and human capital variable- including business training,

mode of establishment- sole proprietorship or partnership arrangement, and socio-

economic variables of the owner to be critical factors in determining firms growth in

micro and small enterprise setting (for detail also refer to Mcpherson, 1994).

Based on the above theoretical discussion, demographic factors, employment, and

amount of additional capital obtained from MFI and/or formal banks, number of

hired permanent employees, average number of temporary employees, trainings

received related to the business, number of years in the business, any policy change

introduce while the business is in operation- for example change in the lending cap.

In our analysis, unlike previous studies, we use employment growth as one of

growth explanatory variable. We argue that labor is one key input of the firm which

is similar to credit, training, land and resources that firms use to build up their

capital1.

Based on the arguments we have put, the methodological framework that we relied

on to see the growth of MSEs’ can be written as follows;

∆lnCt firm’s age

= lnC0 + βX … … … … … … … … … … … … … … … … … (2)

∆lnCt = lnCt - lnC0 represents the change in firm’s growth

Where: lnC t is firm’s log of real current capital, and

lnC0 is firm’s log of real initial capital

1 As a show case, we estimate our models using capital growth and employment growth, although our interpretation will solely be based on capital growth model.

While the vector Xs represents owner’s characteristic- demographic characteristics of

the head: age, gender, number of years in school, marital status, and other

constructed variables, like age square- to control for non-linearity.

We augment the model with two external policy environment factors that MSEs faced

in the country. When we look at the data set, the age of firms extends to 40 years

implying that there has been a regime shift and as well as introduction of new

policies with the current government. To capture these factors we introduce two

policy dummy variables.

∆lnCt

firm’s age= lnC0 + βX + δ1policy shift + δ2regime change … … . (3)

Where: δ1 policy shift dummy for the introduction of new policy related to MSE,

δ2 regime shift.

In our data set we have firms who have been operating since the previous regime (i.e.

prior to 1992). We want to control for such factors by incorporating a dummy for the

firms which are operational since 1992.

MSEs’ growth can be positively influenced by a number of factors. Receiving land or

working premise and training are the two main positive shocks which encourages

the growth of these firms, (see also Mead and Liedholm, (1998)). To account for these

factors we have included two variables and the expanded equation is given by

∆lnCtfirm’s age

= lnC0 + βX + δ1policy shift + δ2regime change

+ ρ premise + σ training … … … … … … …. (4) Where: ρ premise is positive shock dummy for firms that receive land or working premise as positive shock,

σ training, whether the owner has received training related to the business

Liedhom (2001) argued that sector of employment that MSEs are engaged in and the

location of the firm can be taken as two indicators to control for demand for firms’

product. In order to capture the contribution of each of these factors we have

included two variables in our regression framework: sector and location

∆lnCtfirm’s age

= lnC0 + βX + δ1policy shift + δ2regime change + ρ premise + σ training

+ γ location + α sector … … … … . … (5) Where: γ location is a dummy that controls for the location of the firm, α sector captures sector of employment The source of their initial capital have its own implication on the returns from MSEs’

investment and also how the return could be reinvested to prop-up the growth of the

firm. Thus, source dummy is incorporated in the model to account for it.

∆lnCtfirm’s age

= lnC0 + βX + δ1policy shift + δ2regime change + ρ premise + σ training

+ γ location + α sector + θ Source … . … … … … (6) Where: θ Source denotes source of start-up capital: formal banks, micro finance, friends, relatives, neighbors, NGO,

government, other

The β, δ1, δ2, γ, α, ρ, σ and θ are the parameters that we intend to estimate. Thus

equation 6 represents our estimable growth function.

In the analysis stage we also tried to control for any possible non-linear nature of the

variable by including squared terms. We also add interaction terms to account for

the existence of possible scale factors. In this regard, for instance, Garoma (2012)

relying on the works of Goedhuys, and Sleuwaegen (2009), used interaction between

firms and size to control for the role of reputation for firms’ growth. Hall (1987) also

underscored the importance of firm size for the growth of the firm.

Unlike previous studies that suffer from omitted variable bias, which in most cases

is introduced due to limited sample size as their sample is obtained from a specific

locality, in our analysis we account for such factors by utilizing unique nature of the

data set we obtained from MUDC (2012). The data set is representative of all thirteen

major towns of the country. This allows us to control for locational differences by

using regional dummies in our regression analysis.

In the estimation stage we also run a number of regressions to see the existence of

any possible difference in the growth of firms. Following the works of Belay (2012),

we examine any possible difference in the growth patterns of small and large firms

by running the regression into five capital quintile groups.

4.1.1. Test of Internal validly

To test the validity of our results, we use standard tests on each of our regression

framework. Specifically, we run mean difference test to check the statistical

significance difference across male and female MSE operators and youth and mature

operators. We also run joint and individual significance tests of our model.

4.1.2 Test of External Validity

In order to project our findings to the entire MSE sector, we mainly depend on the

results of the previous survey study that is reported by the ministry, MUDC (2012).

To check the validity of our result we make comparison with the result of this study.

Moreover, we also put our result in parallel with other studies.

4.2. CGE Model

A static PEP version CGE model is employed in the CGE part of this study. A CGE

model is a multipurpose and flexible model that has been widely applied for

macroeconomic and sector specific policy analysis in many developed and developing

countries (Lofgren et al, 2002).

The model is designed as a set of simultaneous linear and non-linear equations,

which define the behavior of all economic agents, as well as the economic

environment in which these agents operate. This environment is described by

market equilibrium conditions, and macroeconomic balances.

In the model a multi stage production function is used. Producers are assumed to

maximize profits subject to production technologies, taking prices (to output and

intermediate inputs) and factor wages as given. Producers in the model make

decisions in order to maximize profits subject to constant returns to scale, with the

choice between factors being governed by a constant elasticity of substitution (CES)

function. This specification allows producers to respond to changes in relative factor

returns by smoothly substituting between available factors so as to derive a final

value-added composite. Profit maximization implies that the factors receive income

where marginal revenue equals marginal cost based on endogenous relative prices

(Thurlow, J., 2008).

Our model uses labor and capital as factors of production. For our purpose, we

disaggregate labor by gender and skill level. The production in each activity starts

with a gender function where we used CES specification to combine gender

disaggregated labor of the same skill level in each activity. This makes the model

substitute male and female labor if and only if they are of the same skill level. Then

skilled and unskilled labor types are made to combine less substitutably. This

technically means, the manager first decides labor at what skill level to hire and

then decides whether male or female in that particular skill level. But not vis versa.

Then the composite labor force and (the only) capital combine at the top level of the

value addition. All these stages in the nested value addition function use CES

specification. Finally, at the top level, the sectoral output of each productive activity j

combines value added and total intermediate consumption in fixed shares (with

Leontief specification). As it is clearly seen in the chart below, the different types of

labor are first aggregated into the two skill levels before the final aggregation in the

labor value addition. This is the first value addition or contribution of this research

work into the original model.

Chart 1: Nested structure of production

Equation 7 and 8 below illustrates how men and women employed in the same

activity combine to form a composite labor force both skilled and unskilled at the

beginning of the production line. We opt for CES specification at this stage, to enable

the composition of male and female labor at each skill level to vary as needed. In

order to maximize its profit each activity uses a set of factors (including male and

female labor) up to the point where the marginal revenue product of each factor is

equal to its wage or rent. In unskilled labor market, assuming the work is more of

physical rather than technical, male and female workers are made to be less

substitutable. However, in the skilled labor market, since it is more of technical and

mental work we assume that female workers have better chance to get a job than in

unskilled labor market. Thus male – female substitutability is greater in skilled labor

market than unskilled.

𝐿𝐿𝐿𝐿𝐿𝐿_𝑈𝑈𝑗𝑗 = 𝐵𝐵_𝐿𝐿𝐿𝐿𝑈𝑈𝑗𝑗 ∗ [∑ 𝛽𝛽_𝐿𝐿𝐿𝐿𝑈𝑈𝑙𝑙2.𝑗𝑗𝑙𝑙2 ∗ 𝐿𝐿𝐿𝐿𝑙𝑙2,𝑗𝑗−𝜌𝜌_𝐿𝐿𝐿𝐿𝐿𝐿𝑗𝑗]

−1𝜌𝜌_𝐿𝐿𝐿𝐿𝐿𝐿𝑗𝑗…………………………………………. (7)

𝐿𝐿𝐿𝐿𝐿𝐿_𝑆𝑆𝑗𝑗 = 𝐵𝐵_𝐿𝐿𝐿𝐿𝑆𝑆𝑗𝑗 ∗ [∑ 𝛽𝛽_𝐿𝐿𝐿𝐿𝑆𝑆𝑙𝑙1.𝑗𝑗𝑙𝑙1 ∗ 𝐿𝐿𝐿𝐿𝑙𝑙1,𝑗𝑗−𝜌𝜌_𝐿𝐿𝐿𝐿𝐿𝐿𝑗𝑗]

−1𝜌𝜌_𝐿𝐿𝐿𝐿𝐿𝐿𝑗𝑗 ………………………………………….. (8)

Output (XSTj)

Value added (VAj)

Composite labor (LDCj) Capital (KDk,j)

Skilled labor (LDC_Sj)

Unskilled labor (LDC_Uj)

Skilled male

Skilled female

Unskilled female

Unskilled male

Product2 (DIij)

Product1 (DIij)

Aggregate intermediate consumption (CIj)

Where: 𝐿𝐿𝐿𝐿𝐿𝐿_𝑈𝑈𝑗𝑗 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 𝑢𝑢𝑢𝑢𝑐𝑐𝑢𝑢𝑐𝑐𝑢𝑢𝑢𝑢𝑐𝑐𝑢𝑢 𝑢𝑢𝑙𝑙𝑙𝑙𝑐𝑐𝑙𝑙 𝑓𝑓𝑐𝑐𝑙𝑙𝑐𝑐𝑐𝑐 𝑐𝑐𝑢𝑢 𝑙𝑙𝑐𝑐𝑐𝑐𝑐𝑐𝑎𝑎𝑐𝑐𝑐𝑐𝑎𝑎 𝑗𝑗 𝐿𝐿𝐿𝐿𝐿𝐿_𝑆𝑆𝑗𝑗 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 𝑐𝑐𝑢𝑢𝑐𝑐𝑢𝑢𝑢𝑢𝑐𝑐𝑢𝑢 𝑢𝑢𝑙𝑙𝑙𝑙𝑐𝑐𝑙𝑙 𝑓𝑓𝑐𝑐𝑙𝑙𝑐𝑐𝑐𝑐 𝑐𝑐𝑢𝑢 𝑙𝑙𝑐𝑐𝑐𝑐𝑐𝑐𝑎𝑎𝑐𝑐𝑐𝑐𝑎𝑎 𝑗𝑗 𝜌𝜌_𝐿𝐿𝐿𝐿𝑈𝑈𝑗𝑗is a substitution parameter, 𝜌𝜌_𝐿𝐿𝐿𝐿𝑆𝑆𝑗𝑗is a substitution parameter, 𝐿𝐿𝐿𝐿𝑙𝑙2,𝑗𝑗 𝑐𝑐𝑐𝑐 unskilled male and unskilled female labor in activity j, 𝐿𝐿𝐿𝐿𝑙𝑙1,𝑗𝑗 𝑐𝑐𝑐𝑐 skilled male and skilled female labor in activity j, 𝛽𝛽_𝐿𝐿𝐿𝐿𝑈𝑈𝑙𝑙2,𝑗𝑗 𝑐𝑐𝑐𝑐 CES activity function share parameter, 𝛽𝛽_𝐿𝐿𝐿𝐿𝑆𝑆𝑙𝑙1,𝑗𝑗 𝑐𝑐𝑐𝑐 CES activity function share parameter, 𝐵𝐵_𝐿𝐿𝐿𝐿𝑈𝑈𝑗𝑗 is efficiency parameter in the CES activity function, and 𝐵𝐵_𝐿𝐿𝐿𝐿𝑆𝑆𝑗𝑗 is efficiency parameter in the CES activity function

Equation 9 below shows the aggregation of skilled and unskilled labor into one

composite labor force of different gender and skill level. Assuming significant skill

level difference between skilled and unskilled workers, substitution between workers

in the two skill levels is set to be low.

𝐿𝐿𝐿𝐿𝐿𝐿𝑗𝑗 = 𝐵𝐵_𝐿𝐿𝐿𝐿𝑗𝑗 ∗ [𝛽𝛽_𝐿𝐿𝐿𝐿𝑗𝑗 ∗ 𝐿𝐿𝐿𝐿𝐿𝐿_𝑆𝑆𝑗𝑗−𝜌𝜌_𝐿𝐿𝐿𝐿𝑗𝑗 + (1 − 𝛽𝛽_𝐿𝐿𝐿𝐿𝑗𝑗) ∗ 𝐿𝐿𝐿𝐿𝐿𝐿_𝑈𝑈𝑗𝑗−𝜌𝜌_𝐿𝐿𝐿𝐿𝑗𝑗]

−1−𝜌𝜌_𝐿𝐿𝐿𝐿𝑗𝑗……………………(9)

Where: 𝐿𝐿𝐿𝐿𝐿𝐿𝑗𝑗 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 𝑢𝑢𝑙𝑙𝑙𝑙𝑐𝑐𝑙𝑙 𝑓𝑓𝑐𝑐𝑙𝑙𝑐𝑐𝑐𝑐 𝑐𝑐𝑢𝑢 𝑙𝑙𝑐𝑐𝑐𝑐𝑐𝑐𝑎𝑎𝑐𝑐𝑐𝑐𝑎𝑎 𝑗𝑗 𝜌𝜌_𝐿𝐿𝐿𝐿𝑗𝑗is a substitution parameter, 𝐿𝐿𝐿𝐿𝐿𝐿_𝑈𝑈𝑗𝑗 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 𝑢𝑢𝑢𝑢𝑐𝑐𝑢𝑢𝑐𝑐𝑢𝑢𝑢𝑢𝑐𝑐𝑢𝑢 𝑢𝑢𝑙𝑙𝑙𝑙𝑐𝑐𝑙𝑙 𝑓𝑓𝑐𝑐𝑙𝑙𝑐𝑐𝑐𝑐 𝑐𝑐𝑢𝑢 𝑙𝑙𝑐𝑐𝑐𝑐𝑐𝑐𝑎𝑎𝑐𝑐𝑐𝑐𝑎𝑎 𝑗𝑗 𝐿𝐿𝐿𝐿𝐿𝐿_𝑆𝑆𝑗𝑗 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 𝑐𝑐𝑢𝑢𝑐𝑐𝑢𝑢𝑢𝑢𝑐𝑐𝑢𝑢 𝑢𝑢𝑙𝑙𝑙𝑙𝑐𝑐𝑙𝑙 𝑓𝑓𝑐𝑐𝑙𝑙𝑐𝑐𝑐𝑐 𝑐𝑐𝑢𝑢 𝑙𝑙𝑐𝑐𝑐𝑐𝑐𝑐𝑎𝑎𝑐𝑐𝑐𝑐𝑎𝑎 𝑗𝑗 𝛽𝛽_𝐿𝐿𝐿𝐿𝑗𝑗 𝑐𝑐𝑐𝑐 CES activity function share parameter, and 𝐵𝐵_𝐿𝐿𝐿𝐿𝑗𝑗 is efficiency parameter in the CES activity function,

Equation 10, 11 and 12 generates the relative demand functions for male and female

labor in each skill level and between skilled and unskilled aggregated labor forces

respectively. The equations show relative demand for each labor (in the composition)

relies on a share parameter, the relative wage rate, and the sectoral elasticity of

substitution. The optimal mix of the different labor forces is a function of the relative

wage rates of each labor type.

𝐿𝐿𝐿𝐿𝑙𝑙2,𝑗𝑗 = [𝛽𝛽_𝐿𝐿𝐿𝐿𝑈𝑈𝑙𝑙2,𝑗𝑗 ∗ �𝑊𝑊𝐿𝐿_𝑈𝑈𝑗𝑗𝑊𝑊𝑊𝑊𝑊𝑊𝑙𝑙2,𝑗𝑗� �]𝜎𝜎_𝐿𝐿𝐿𝐿𝐿𝐿𝑗𝑗 ∗ 𝐵𝐵_𝐿𝐿𝐿𝐿𝑈𝑈𝑗𝑗

𝜎𝜎_𝐿𝐿𝐿𝐿𝐿𝐿𝑗𝑗−1 ∗ 𝐿𝐿𝐿𝐿𝐿𝐿_𝑈𝑈𝑗𝑗………………………. (10)

Where: 𝑊𝑊𝑊𝑊𝑊𝑊𝑙𝑙2,𝑗𝑗 wage rate paid by industry j for unskilled labor including tax, 𝑊𝑊𝐿𝐿_𝑈𝑈𝑗𝑗 wage rate of industry j composite unskilled labor 𝜎𝜎_𝐿𝐿𝐿𝐿𝑈𝑈𝑗𝑗 CES composite elasticity

𝐿𝐿𝐿𝐿𝑙𝑙1,𝑗𝑗 = [𝛽𝛽_𝐿𝐿𝐿𝐿𝑆𝑆𝑙𝑙1,𝑗𝑗 ∗ �𝑊𝑊𝐿𝐿_𝑆𝑆𝑗𝑗𝑊𝑊𝑊𝑊𝑊𝑊𝑙𝑙1,𝑗𝑗� �]𝜎𝜎_𝐿𝐿𝐿𝐿𝐿𝐿𝑗𝑗 ∗ 𝐵𝐵_𝐿𝐿𝐿𝐿𝑆𝑆𝑗𝑗

(𝜎𝜎_𝐿𝐿𝐿𝐿𝐿𝐿𝐽𝐽−1) ∗ 𝐿𝐿𝐿𝐿𝐿𝐿_𝑆𝑆𝑗𝑗………………………. (11)

𝐿𝐿𝐿𝐿𝐿𝐿_𝑆𝑆𝑗𝑗 = [�𝛽𝛽_𝐿𝐿𝐿𝐿𝑗𝑗(1 − 𝛽𝛽_𝐿𝐿𝐿𝐿𝑗𝑗)� � ∗ �𝑊𝑊𝐿𝐿_𝑈𝑈𝑗𝑗

𝑊𝑊𝐿𝐿_𝑆𝑆𝑗𝑗� �]𝜎𝜎_𝐿𝐿𝐿𝐿𝑗𝑗 ∗ 𝐿𝐿𝐿𝐿𝐿𝐿_𝑈𝑈𝑗𝑗………………….…………. (12)

Activity level wage rate for the composite labor is calculated as a weighted average of

wage for skilled and unskilled labors in that activity (equation 13). In detail, wage in

the non- MSE sector is assumed to be fixed since wage rate in the public sector and

big firms is not market driven and rather set and led by the government’s decision.

However, wage rate in the MSE sector, the sector in which unemployment is

assumed, is set to be efficiency wage rate calculated as shown in equation 14 below.

𝑊𝑊𝐿𝐿𝑗𝑗 ∗ 𝐿𝐿𝐿𝐿𝐿𝐿𝑗𝑗 = �𝐿𝐿𝐿𝐿𝐿𝐿_𝑆𝑆𝑗𝑗 ∗ 𝑊𝑊𝐿𝐿_𝑆𝑆𝑗𝑗� + �𝐿𝐿𝐿𝐿𝐿𝐿_𝑈𝑈𝑗𝑗 ∗ 𝑊𝑊𝐿𝐿_𝑈𝑈𝑗𝑗�………………………………….(13)

𝑊𝑊_𝑀𝑀𝑆𝑆𝑀𝑀𝑙𝑙 = 𝑐𝑐𝑐𝑐𝑙𝑙 + �𝑐𝑐𝑐𝑐𝑙𝑙 𝑞𝑞𝑞𝑞𝑙𝑙� � ∗ �𝑙𝑙𝑙𝑙𝑙𝑙 𝑢𝑢𝑢𝑢𝑙𝑙 � + 𝑙𝑙𝑙𝑙� …………………………………………..(14)

Where : 𝑊𝑊𝐿𝐿𝑗𝑗 wage rate of industry j for composite labor,

𝑊𝑊_𝑀𝑀𝑆𝑆𝑀𝑀𝑙𝑙 wage rate for labor in MSE sector

𝑐𝑐𝑐𝑐𝑙𝑙 effort desutility

𝑞𝑞𝑞𝑞𝑙𝑙 probability of detection of shirking

𝑙𝑙𝑙𝑙𝑙𝑙 exogenous probability to be fired

𝑢𝑢𝑢𝑢𝑙𝑙 unemployment rate

𝑙𝑙𝑙𝑙 discount rate

In this model unemployment is considered. The labor market from the supply side is

constructed as such the non-MSE sector satisfies its labor demand first (i.e every

worker prefer to work in the non-MSE sector). If a worker couldn’t get job in the

Non-MSE sector then she/he will search in the MSE sector. Some will be successful

and some will remain idle. Thus unemployment is considered in this sector. As it is

shown in equations 15 and 16 full employment is settled in the non-MSE sector

however the residual labor supply goes to the MSE sector. Thus whenever the MSE

sector is expanded it can call up some labor from the unemployed pool.

𝐿𝐿𝑆𝑆_𝑁𝑁𝑀𝑀𝑆𝑆𝑀𝑀𝑙𝑙 = ∑ 𝐿𝐿𝐿𝐿𝑙𝑙,𝑗𝑗2𝑗𝑗2 …………………………………………………. (15)

∑ 𝐿𝐿𝐿𝐿𝑙𝑙,𝑗𝑗1𝑗𝑗1

(1− 𝑢𝑢𝑢𝑢𝑙𝑙)� = 𝐿𝐿𝑆𝑆𝑙𝑙 − 𝐿𝐿𝑆𝑆_𝑁𝑁𝑀𝑀𝑆𝑆𝑀𝑀𝑙𝑙………………………………… (16)

Where: 𝐿𝐿𝑆𝑆_𝑁𝑁𝑀𝑀𝑆𝑆𝑀𝑀𝑙𝑙 Labor supply in the Non-MSE sector, 𝐿𝐿𝐿𝐿𝑙𝑙,𝑗𝑗2 Labor demand of the Non-MSE sector, 𝐿𝐿𝐿𝐿𝑙𝑙,𝑗𝑗1 Labor demand of the MSE sector, and 𝐿𝐿𝑆𝑆𝑙𝑙 Total labor supply in the economy

In each sector a CES production function takes care of the aggregation of the

composite labor and capital so as to come up with a final value-added composite.

The substitution between capital and labor at this level of the production is treated

differently for MSE and Non-MSE sectors. Since we assume technological change in

the MSE sector is not that easy and needs some time, the substitution between labor

and capital is set lower than in the non-MSE sector in which technological change is

more likely. The CES specification is shown in equations 17 and 18.

VAj = B_vaj. [β_vaj ∗ LDC𝑗𝑗−ρ_va𝑗𝑗 + �1 − β_va𝑗𝑗� .𝐾𝐾𝐿𝐿𝐿𝐿j

−ρ_va𝑗𝑗]( −1ρ_va𝑗𝑗

) ……………………….. (17)

LDCjKDCj

= [�RCjWCj

� ∗ �βvaj

1−βvaj�]σ_vaj ………………………………………………………………….…… (18)

Where: 𝑉𝑉𝑉𝑉𝑗𝑗 𝑐𝑐𝑐𝑐 𝑞𝑞𝑢𝑢𝑙𝑙𝑢𝑢𝑐𝑐𝑐𝑐𝑐𝑐𝑎𝑎 𝑐𝑐𝑓𝑓 𝑎𝑎𝑙𝑙𝑢𝑢𝑢𝑢𝑐𝑐 𝑙𝑙𝑢𝑢𝑢𝑢𝑐𝑐𝑢𝑢 𝑐𝑐𝑢𝑢 𝑐𝑐𝑢𝑢𝑢𝑢𝑢𝑢𝑐𝑐𝑐𝑐𝑙𝑙𝑎𝑎 𝑗𝑗 𝐵𝐵_𝑎𝑎𝑙𝑙𝑗𝑗 is efficiency parameter in the CES value-added function, 𝛽𝛽_𝑎𝑎𝑙𝑙𝑗𝑗 is CES value-added function share parameter for a factor in industry j, 𝐾𝐾𝐿𝐿𝐿𝐿𝑗𝑗 is quantity demanded of factor capital from industry j, RC𝑗𝑗 Rental rate of industry j composite capital 𝜎𝜎_𝑎𝑎𝑙𝑙𝑗𝑗 Elasticity (CES - value added) 𝜌𝜌_𝑎𝑎𝑙𝑙𝑗𝑗 is CES value-added function exponent,

Once determined, these value-added composites are combined with fixed-share

intermediates using a Leontief specification. The use of fixed-shares reflects the

belief that the required combination of intermediates per unit of output, and the

ratio of intermediates to value added, is determined by technology rather than by the

decision-making of producers (Thurlow, J., 2008).

In some industry and service sectors (which are classified into MSE and Non-MSE) a

product is produced from two lines of production which uses two different

technologies. In this study the disaggregation in the production part of the model

goes down to scale of operation level especially in the manufacturing and

construction activities from the industry sector and trade and hotel and restaurant

activities from service sector. We select these sectors because most of the MSEs in

Ethiopia are engaged in them2. The two lines of production follow two different

production technologies to produce the same output. As shown in table 1 below, the

MSE sectors are using labor intensive technology than the Non-MSE.

CES functional structure is used to supply a single product of the two production

lines. Thus a commodity is produced either from MSE or Non-MSE sector whichever

is found cheaper in its production cost. Assuming difference in the product quality

which is not visible for the consumer, substitutability between the two lines of

production is set to be imperfect.

Table 1: Production technologies in the MSE and Non-MSE sectors

Source: Data from the SAM documentation 2005/06 and own computation

Profit maximization drives producers to sell their products in domestic or foreign

markets based on the potential returns. It is assumed that marketed domestic

output is allocated to two alternative destinations: domestic sales and exports. The

model shows imperfect transformability between these two destinations, via the use

of constant elasticity of transformation (CET) functions (Lofgren, H. et al, 2002).

Domestic output net of exports from each production line is combined with a CES

2 Operations at micro (classified under handicraft and cottage according to a classification by CSA) and small level are grouped under the MSEs while operations at medium and large level are grouped under non-MSEs.

Labor share Capital share Labor share Capital shareAgriculture 75.4% 24.6%Dairy 61.0% 39.1% 46.5% 53.5%Food 57.1% 43.0% 38.7% 61.3%Beverage 61.0% 39.1% 46.5% 53.5%Textiule and leather

61.0% 39.1% 46.5% 53.5%

Wood 61.0% 39.1% 46.5% 53.5%Metal 61.0% 39.1% 46.5% 53.5%Construction 50.5% 49.5% 25.6% 74.4%Other industries 41.4% 58.6%Trade 48.0% 52.0% 20.6% 79.4%Hotel 62.9% 37.2% 50.3% 49.7%Administration 31.1% 68.9%Other service 54.1% 45.9%

Non_MSEIndustrial clasifications

MSE

function and supplied domestically as shown in equation 19. The aggregation is

made from the supply side not from the demand side. This technically means the

distributor knows whether from MSE or Non-MSE a product is but the consumer

doesn’t know. In turn this composite supply is combined with the imported products

in an imperfect substitutability (i.e. Armington assumption) and creates total supply

to satisfy domestic demand. This demand is the sum total of all demand by

economic agents: it constitutes final consumption demands by households, and

government and investment demand, intermediate consumption demands by

activities and transaction services’ demand.

𝐿𝐿𝐿𝐿𝑖𝑖 = 𝐵𝐵_𝐿𝐿𝑆𝑆𝑖𝑖 ∗ ∑ [𝛽𝛽_𝐿𝐿𝑆𝑆𝑗𝑗,𝑖𝑖 ∗ 𝐿𝐿𝑆𝑆𝑗𝑗,𝑖𝑖−𝜌𝜌_𝐿𝐿𝐿𝐿𝑖𝑖

𝑗𝑗 ]−1

𝜌𝜌_𝐿𝐿𝐿𝐿𝑖𝑖� …………………………………….. (19)

Where: 𝐿𝐿𝐿𝐿𝑖𝑖 domestic demand for commodity i produced locally 𝐿𝐿𝑆𝑆𝑗𝑗 ,𝑖𝑖 supply of commodity i by sector j to the domestic market 𝐵𝐵_𝐿𝐿𝑆𝑆𝑖𝑖 scale parameter 𝛽𝛽_𝐿𝐿𝑆𝑆𝑗𝑗,𝑖𝑖 share parameter 𝜌𝜌_𝐿𝐿𝑆𝑆𝑖𝑖 elasticity parameter Households have final consumption demands with the objective of utility

maximization subject to budget constraints. The model has one representative

consumer per household type, rendering identical preferences for all consumers in a

given category. In our model there are two types of households, urban and rural. As

shown in table 2 these households get their income from factor and non-factor

sources. The non-factor sources we already have in our SAM are government

transfer (social security for instance) and remittance. Representative household

groups maximize their incomes by allocating factors of production across activities.

Table2: Households’ source of income

Source: Data from the SAM documentation 2005/06

Agricultural labor

Non-agricultural

labor

Capital (capital +

Land)Government

transferRemittance

Rural household 33.2% 1.2% 34.5% 0.5% 4.2%

Urban household 10.9% 7.2% 0.7% 7.7%

Regarding poverty analysis, a separate consumption based micro-simulation module

is prepared. This links each respondent in the 2009/10 HICE (Household Income

Consumption and Expenditure) survey to their corresponding representative

household group in the model. Thus we employ a top-down approach in which

changes in commodity prices and households’ consumption spending are passed

down from the CGE model to the micro-simulation module, where per capita

consumption and standard poverty measures are recalculated. Poverty will be

modeled using the Foster-Greer-Thorbecke (FGT) measures (Foster et al, 1984). This

measure is noted as:

Pα = 1n∑ (z−yi

z)αq

i=1 ………………………………………………………………………….. (20)

Where: 𝛼𝛼 𝑐𝑐𝑐𝑐 𝑐𝑐ℎ𝑐𝑐 𝑐𝑐𝑐𝑐𝑎𝑎𝑐𝑐𝑙𝑙𝑐𝑐𝑎𝑎 𝑙𝑙𝑎𝑎𝑐𝑐𝑙𝑙𝑐𝑐𝑐𝑐𝑐𝑐𝑢𝑢 𝑐𝑐𝑙𝑙𝑙𝑙𝑙𝑙𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑙𝑙,

n is population size,

q is the number of people below the poverty line,

𝑎𝑎𝑖𝑖 𝑐𝑐𝑐𝑐 𝑐𝑐𝑢𝑢𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐,

𝑧𝑧 𝑐𝑐𝑐𝑐 𝑐𝑐ℎ𝑐𝑐 𝑐𝑐𝑐𝑐𝑎𝑎𝑐𝑐𝑙𝑙𝑐𝑐𝑎𝑎 𝑐𝑐ℎ𝑙𝑙𝑐𝑐𝑐𝑐ℎ𝑐𝑐𝑢𝑢𝑢𝑢.

The FGT Pα class of additive decomposable poverty measures allows us to measure

the proportion of poor in the population; poverty head count ratio if α = 0, poverty

depth if α = 1, and severity of poverty if α=2.

We model public investment for MSEs’ development in terms of provision of credit

through exogenous increase in MSEs’ capital and in terms of provision of training to

MSE operators through increasing production efficiency in the sector.

So the hypothesized flow of effects of government’s effort on capacity building of

operators and improved access to capital on MSEs is; provision of technical trainings

give the operators better knowledge of production and business management which

improve the efficiency of the sector. On the other line, better access to loan will make

the MSE sector augment its capital stock. Thus the increased production as a result

of improved productivity and more accumulated capital stock leads to better income

in the MSE sector which finally resulted in increased urban households’ income and

spending on different commodities. This also has positive effect on welfare of urban

households and reduces urban poverty.

5. Data As stated above, in this study, we aim to answer two key research questions. First,

we seek to identify the major determinants of the growth of firms in Ethiopia.

Second, we examine, the role of MSEs to reduce unemployment and poverty. For

the former question, we rely on a survey undertaken by Ministry of Urban

Development and Construction (MUDC) on about 3000 enterprises in 2012. The

survey covered 13 major cities in the country with a population of more than 100,

0003.

The objective of the survey was “….to generate adequate, up-to-date and reliable

information on growth oriented Micro and Small Enterprises (MSEs) ……”. To this

end, detailed information on the characteristics of the operators of the enterprises

(age, gender, education level, experience); profile of the enterprises (establishment

year, initial capita, current capital, number of employees at the time of

establishment, current number of employees);and major constraints facing the

enterprise, were collected, among others (MUDC, 2012).

For the second key research question we rely on CGE modelling. The CGE model

used in this study is calibrated on a Social Accounting Matrix (SAM) of Ethiopia.

This SAM was first developed by EDRI (Ethiopian Development Research

Institute) for 2005/06 Ethiopian economy. It was later updated for 2009/10 for a

research work on alternative financing of the GTP plan. The procedures taken to

update the SAM are discussed here: the first move was to simulate the growth of

the Ethiopian economy based on actual economic developments from 2005/06–

2009/10 using dynamic CGE model. A new and balanced SAM for 2009/10 came

out as a result. Then the projected 2009/10 SAM and the GDP were then

converted to current prices. Activities’ shares of actual value added and actual

aggregate demand components of 2009/10 (from national accounts) were then

used to adjust sectors’ value added in the projected 2009/10 SAM. This resulted

3 The complete list of the surveyed cities include Addis Ababa, Hawassa, Mekele, Gondar, Bahirdar, Dessie, Jimma, Shashemene, DireDawa, Bishoftu, Adama, Jijiga, and Harar.

in an unbalanced SAM, which was then balanced using a cross entropy program

(Ermias et al, 2011).

Further modification is made on the SAM in order to fit the objectives of this

study. The SAM in use in this study has 22 activities totally. 18 of them were 9

activities but split by scale of operation into MSE and Non-MSE. Disaggregation

of these activities by scale of operation is made maintaining the real composition

of value added and intermediate inputs for each activity and at each level of

operation. The data is found from CSA’s manufacturing sector surveys. Table 3

below shows the share of MSE and Non-MSEs from the entire activity’s value

addition and intermediate input consumption.

Despite the difference at scale of operation, every activity produces a single

output. Thus there are only 13 commodities in the SAM.

Table 3: Intermediate input and value added shares from an activity by scale of operation

Source: Data from CSA 2003 and 2004, and own computation

Intermediate input

Value addition

Intermediate input

Value addition

Agriculture 9.5% 90.5% 100.0%

Dairy 23.2% 8.8% 39.9% 28.1% 100.0%

Food 23.2% 8.8% 39.9% 28.1% 100.0%

Beverage 23.2% 8.8% 39.9% 28.1% 100.0%Textile and leather 15.3% 11.2% 61.4% 12.1% 100.0%

Wood 22.8% 24.0% 34.1% 19.2% 100.0%

Metal 8.2% 9.1% 65.6% 17.1% 100.0%

Construction 16.8% 11.1% 48.4% 23.7% 100.0%

Other industry 60.7% 39.3% 100.0%

Trade 16.1% 12.8% 49.9% 21.2% 100.0%

Hotel 23.2% 8.8% 39.9% 28.1% 100.0%

Administration 65.2% 34.8% 100.0%Other services 65.2% 34.8% 100.0%

MSE Non-MSETotal

There are 5 types of factors in the SAM; four types of labor disaggregated by

gender and skill level, and one type of capital. The real shares of labor and capital

are guided by the data shown in the previous section in table 1. The

disaggregation of labor by gender and skill level is governed by real shares based

on data from MUDC survey of 2013 for MSE sector and CSA’s labor force survey

data (CSA, 2006). Based on definition from SAM documentation 2005/06,

unskilled labor is worker engaged in elementary occupation which requires only

little skill or lowest level of education. Besides, in ILO website elementary

occupation is defined as occupation which requires only the 1st ISCO skill levels.

And skill level 1 requires only completion of primary level education or the 1st

level of education (ILO, 2012). Thus based on these definitions we classified the

labor force with primary level education and less as unskilled labor and higher

than primary level as skilled labor. (See table 4 below).

Having a male-female segmented labor market in the SAM gives chance to see

gender bias in terms of wages and employment opportunities in the Ethiopian

labor market, and also occupational differences.

Table 4: Share of labor by gender and skill level for MSE and Non-MSE sectors

Source: CSA 2006, MUDC 2012 and own computation

There are also 2 households that are disaggregated by location (urban/ rural).

Government, ‘saving-investment’, ‘rest of the world’ and different tax types are

also components of this SAM.

For the poverty analysis we employed the very recent, 2009/10 HICES data of

CSA. The households from this data are disaggregated into urban and rural. Thus

it enables us to look into urban poverty as a result of government’s effort to

develop the MSE sector, which is also very important for policy recommendation.

Female unskilled

Male unskilled

Female skilled

Male skilled

Female unskilled

Male unskilled

Female skilled

Male skilled

Agriculture 48.5% 48.5% 1.5% 1.5%Dairy 30.4% 30.1% 17.2% 22.3% 10.9% 26.1% 9.0% 54.1%Food 29.0% 14.0% 24.3% 32.7% 10.9% 26.1% 9.0% 54.1%Beverage 29.0% 14.0% 24.3% 32.7% 10.9% 26.1% 9.0% 54.1%Textile and Leather 20.6% 34.4% 16.5% 28.6% 20.3% 20.4% 16.8% 42.4%Wood 4.4% 22.6% 13.1% 60.0% 7.4% 28.1% 6.1% 58.4%Metal 4.4% 22.6% 13.1% 60.0% 8.0% 27.8% 6.6% 57.7%Construction 4.3% 36.1% 8.6% 51.1% 14.4% 24.0% 11.8% 49.8%Other industries 15.5% 23.3% 12.8% 48.4%Trade 25.9% 25.4% 23.7% 24.9% 12.4% 25.1% 10.2% 52.2%Hotel 29.0% 14.0% 24.3% 32.7% 10.9% 26.1% 9.0% 54.1%Administration 12.4% 25.1% 10.2% 52.2%Other services 12.4% 25.1% 10.2% 52.2%

Non-MSEMSE

6. Results and analysis 6.1 Analysis of econometric results

In the econometrics part of the analysis, we set out to address the 1stresearch

question: the major constraints of enterprises growth which we tried to look

through determinants of start-up capital and that of capital growth. We also set out

to examine if the size of firms and their pace of growth are different by age and

gender. Based on these key research questions, we started working our analysis

first by descriptive analysis and then running some regressions.

We have used descriptive analysis to see the association between age and gender

with our main dependent variable. In addition we have used standard regression

technique to look at the link between our main dependent variables in the model,

which are initial capital and capital growth, with key explanatory variables. We

have introduced a number of key explanatory variables in both models to look into

their importance in explaining each of them. First, we see that the size of initial capital is low and that the rate of growth of

capital is slow among Micro and Small Enterprises (MSEs) in Ethiopia. It is

particularly interesting to see that both the size of start-up capital and the rate of

capital growth differ notably by gender and age (see Figure 6.1 and Figure 6.2

below).

Figure 6.1: Size of initial capital by Gender

0.05

.1.15

.2

0 5 10 15 20log of initial capital (ETB)

Female Male

Figure 6.2: Real capital growth by gender

The literature identifies several reasons as to why the growth in capital of firms run

by women is slower than their male counter parts, particularly in developing

economies. One of these reasons is the difference in managerial capabilities

attributable to difference in educational background and lack of role-models. These

differences often reflect on the capacity of utilization of firms run by respective

group. In order to examine this issue, we calculated the rate of capacity utilization

by women and men and the result is presented in Table 6.1 below. The Table

shows that the average capacity utilization of men (54%) is statistically higher than

that of women (50%). In addition Women-owned firms face multiple challenges.

Although evidence shows they are as effective as male owner/managers, women

often use their firms as part of household survival strategy and opt not to grow.

Table 6.1: Capacity utilization rate of firms across gender

Group Average capacity

utilization rate Std. Err. Std. Dev. Female Owned 50.2 0.682 23.60 Male Owned 54.4 0.577 24.14 Combined 52.7 0.442 24.01 Difference b/n male-Female -4.2*** 0.897

Source: Econometrics result Our analysis also shows that there is considerable regional and sectoral variation

in capital growth indicating that there is a room to improve the return to

investment (capital formation) across all regions and sectors. Next, in this study we

0 .1 .2 .3Average real Capital growth

Male

Female

are working to identify optimal adjustment in investment portfolio so as to

maximize impact based on rate of growth of capital across regions and sectors.

One of the disadvantages of the above descriptive analysis is the fact that the

difference in size of initial capital, capital growth or even the rate of capacity

utilization between male and female owners or young and mature

owners/managers could be attributable to other factors since a lot of other changes

are happening in the country. To tease out the effect of gender and age, we run a

regression function, controlling for firm level as well as locational characteristics.

Table 6.2 and Table 6.4 below present results of this exercise respectively for initial

capital and capital growth.

Table 6.2: Determinants of Initial Capital (log) Variables Model 1 Model 2 Model 3 Model 4 log (average age of owners) 3.974** 4.341** 4.141** 3.974**

(2.227) (2.472) (2.420) (2.347)

log (average age squared of owners) -0.536** -0.593** -0.560** -0.535**

(-2.131) (-2.394) (-2.318) (-2.238)

Gender of owners (Male=1) 0.556*** 0.547*** 0.451*** 0.408***

(9.052) (8.617) (7.271) (6.624)

Education Level of Owners (base=No education) Read and Write 0.245 0.123 0.0441 0.0745

(1.593) (0.814) (0.292) (0.502)

Primary 0.462*** 0.362*** 0.208* 0.332***

(4.073) (3.338) (1.927) (3.052)

High school 0.896*** 0.780*** 0.564*** 0.680***

(8.063) (7.234) (5.232) (6.241)

Vocational 0.949*** 0.831*** 0.574*** 0.703***

(6.172) (5.484) (3.765) (4.700)

University 1.380*** 1.221*** 0.878*** 1.013***

(9.510) (8.665) (6.213) (7.115)

Sector of Participation (base=construction) Agro-processing

-0.0900 0.0153 0.0833

(-0.936) (0.165) (0.894)

Textile

-0.481*** -0.317*** -0.316***

(-4.359) (-2.926) (-2.921)

Leather

-1.381*** -1.126*** -1.121***

(-9.245) (-7.326) (-7.012)

Retail

-0.101 0.0682 -0.0180

(-1.196) (0.805) (-0.201)

Urban agriculture

-0.0688 0.103 0.156

(-0.446) (0.682) (1.041)

Food processing

-0.579*** -0.398** -0.371**

(-3.197) (-2.378) (-2.131)

Other

-0.444** -0.181 -0.152

(-2.119) (-0.966) (-0.787)

Formal Training on production (Yes=1)

-0.195** -0.175**

(-2.370) (-2.166)

Formal Training on management (Yes=1)

0.0678 0.0820

(0.789) (0.965)

Transaction recording (Yes=1)

0.507*** 0.424***

(8.373) (6.645)

Ownership of Bank Account (Yes=1)

0.675*** 0.676***

(11.19) (11.19)

log (total Number of Members)

0.0374 0.0833**

(1.026) (2.229)

Regional Fixed Effects

No No Yes

-0.907 -1.184 -1.352 -1.182

(-0.288) (-0.381) (-0.447) (-0.395)

Constant 2,895 2,895 2,823 2,823 R-Squared 0.085 0.116 0.181 0.207 Robust t-statistics in parentheses

*** p<0.01, ** p<0.05, * p<0.1

Table 6.2 presents the result of regression of initial capital on a number of

explanatory variables. To show robustness of the result and indicate the relative

effect of the variables included in the regression, we run regression models by

subsequently increasing the number of variables. In the most complete model,

model 4, we control for firm characteristics, owners/ managers characteristics, and

location fixed effects.

The result from this regression shows that age of owners (managers) is positively

associated with size of initial capital indicating that young owners of enterprises

face higher financial shortage for start-up capital compared to older enterprise

holders. However, the sign and magnitude of the squared age terms shows that

there is a limit after which this association changes. This result is consistent with

other studies conducted on the subject (see for instance Aubert, P. and Crépon, B.,

2006).

Education level and financial literacy of owners (in terms of keeping financial

record, saving attitude) are also key determinants of initial capital. The table shows

that an additional education above basic literacy level (reading and writing) is

consistently increases with the size of start-up capital. The size of start-up capital a

university graduate can access is, for instance, on average twice the size accessible

to an illiterate owner. Awareness of the owners about transaction recording and

ownership of bank accounts are also associated significantly with the size of initial

capital.

More interestingly, firms owned by women tend to mobilize less resource to start

up business compared to firms owned (run) by male counter parts. This also shows

that it is hard to mobilize bigger financial resource to start this business for women

than men. In the above paragraph we saw that education, training, and age are

associated with higher start-up capital. From our data we can confirm that women

tend to have lower education level, access training opportunities less and are

relatively younger (see table 6.3)

The result also shows that construction sector raises higher initial capital

compared to the other sectors. This is particularly notable when compared against

the leather, textile, and food processing sectors.

Table 6.3: Difference between male and female attributes

Variables Female Male Total Illiterate 15.9 4.9 9.4 Can read and write 5.9 5.3 5.5 Primary education 30.3 35.2 33.2 High school education 35.3 39.5 37.8 Vocational education 5.1 6.9 6.1 University education 7.5 8.2 7.9 Age 33.8 34.4 34.2 Production training 19.1 27.5 24.1 Management training 19.8 20.0 19.9

Table 6.4: Determinants of Capital Growth Variable Label Model 1 Model 2 Model 3 Model 4 Model 5

log (Initial Capital) -0.0305*** -0.0324*** -0.0355*** -0.0376*** -0.0392***

(-13.84) (-14.26) (-14.33) (-15.01) (-15.65)

log (average age of owners) 0.00661 0.0316 0.0486 0.0453 -0.0215

(0.0479) (0.231) (0.356) (0.335) (-0.160)

log (average age squared of owners) -0.00739 -0.0112 -0.0133 -0.0127 -0.00100

(-0.382) (-0.587) (-0.693) (-0.671) (-0.0532)

Gender of owners (Male=1) 0.0219*** 0.0168*** 0.0140** 0.0132** 0.0175***

(3.858) (2.866) (2.374) (2.197) (2.943)

Education Level of Owners (base=No education) Read and Write 0.0114 0.00852 0.00398 0.00253 0.00118

(0.776) (0.584) (0.267) (0.171) (0.0813)

Primary 0.0300*** 0.0263*** 0.0173* 0.0153 0.0129

(3.111) (2.788) (1.769) (1.508) (1.299)

High school 0.0360*** 0.0307*** 0.0190** 0.0188* 0.0175*

(3.789) (3.285) (1.971) (1.881) (1.786)

Vocational 0.0241 0.0165 0.00218 0.00137 -0.000226

(1.596) (1.110) (0.142) (0.0896) (-0.0151)

University 0.0628*** 0.0537*** 0.0375** 0.0366** 0.0382**

(4.348) (3.783) (2.541) (2.374) (2.507)

Sector of Participation (base=construction Sector) Agro-processing

-0.0223** -0.0174* -0.0107 -0.00962

(-2.511) (-1.927) (-1.180) (-1.079)

Textile

-0.0517*** -0.0461*** -0.0388*** -0.0359***

(-4.428) (-3.925) (-3.281) (-3.059)

Leather

-0.0832*** -0.0738*** -0.0564*** -0.0483***

(-7.262) (-6.130) (-4.587) (-4.059)

Retail

-0.0379*** -0.0278*** -0.0123 -0.0145

(-5.080) (-3.453) (-1.319) (-1.580)

Urban agriculture

-0.0191* -0.0139 -0.00430 -0.00192

(-1.765) (-1.259) (-0.386) (-0.176)

Food processing

-0.0594*** -0.0573*** -0.0504*** -0.0489***

(-3.773) (-3.608) (-3.199) (-3.178)

Other

-0.0379** -0.0235 -0.0107 -0.00750

(-2.156) (-1.381) (-0.647) (-0.460)

Formal Training on production (Yes=1)

0.00849 -0.000450 -0.000265

(1.112) (-0.0586) (-0.0351)

Formal Training on management (Yes=1)

-0.00172 -0.00408 -0.00471

(-0.221) (-0.533) (-0.625)

Transaction recording (Yes=1)

0.0222*** 0.0175*** 0.0180***

(3.533) (2.767) (2.886)

Ownership of Bank Account (Yes=1)

0.0299*** 0.0216*** 0.0219***

(5.059) (3.699) (3.796)

log (Average number of workers)

0.0298*** 0.0297***

(5.941) (5.973)

Regime change (After 1994=1)

0.0160*

(1.910)

Policy Change (After 2000=1)

0.0802***

(13.28)

Reginal Fixed Effects No No No Yes Yes Constant 1.425*** 1.441*** 1.411*** 1.396*** 1.400***

(5.797) (5.944) (5.810) (5.803) (5.879)

Observations 2,798 2,798 2,728 2,728 2,728 R-squared 0.118 0.136 0.151 0.174 0.197 Robust t-statistics in parentheses

*** p<0.01, ** p<0.05, * p<0.1

As argued before, in this paper, we proxy the growth of firms with their capital

growth. The caveat in the estimation we present below is the lack of relevant

variables indicating quality of institutions, degree of competition in the respective

sectors and external constraints (e.g. tax and credit system). More importantly, we

have only one round data. This hinders efficient estimation of our models especially

since it does not allow us to control for time invariant firm unobservable. Besides

running alternative models as a robustness check, we believe that part of this

problem could be addressed through the regional fixed effects we controlled for in

the last specification (Model 4).

The Key findings from Capital Growth regressions are the following: First, Initial

capital is negatively associated with capital growth indicating that smaller firms

tend to grow faster than larger firms. This is consistent with prior studies both in

agriculture and manufacturing sectors (see for instance, Coad and Tamvada

(2008); You, 1995). For economic policy, this offers case for improved support to

MSEs as a means of economic growth and poverty alleviation.

Capital growth of Male owners (managers) is higher than that of the female

counterpart. This shows that male owners are more successful both in mobilizing

resource to start up enterprises and grow faster what it is in operation. We believe

this can be explained by many factors. However, since we don’t have detailed

demographic characteristics in the data, we could not tell if this result endures

once other factors are effectively controlled for. Nevertheless, there are rich

literatures supporting this result (see for instance Coad and Tamvada, 2008).

Similar to the case of initial capital, education levels of owners (Managers),

particularly high school and University level educations, and financial literacy (in

terms of keeping financial record, saving attitude) play significantly important role

in firms' capital growth. However the magnitude of the estimates is far lower than

that of initial growth suggesting that there is probable convergence between

education levels.

The result also suggests that construction sector not only mobilizes higher initial

capital, but it also registers significantly higher capital growth rate compared to the

other sectors. This is not surprising given the level and pace of economic growth in

the country of which the growth in the construction sector is disproportionally

higher. To confirm this result, the variables we included in the regression for

regime change and policy change appeared statistically significant.

6.2. Analysis of CGE results

6.2.1 Simulations The study tests two sets of simulations. The first is exploring scenarios that

consider impact of prevailing government interventions on MSEs. The scenarios

in this simulation are:

1. Raising MSEs’ TFP with the prevailing access to and coverage of training.

Here we assume a trained labor can use all the other factors more efficiently.

Thus, training can increase productivity of not only labor but also the other

factors.

2. Capital growth in the MSE sector by the provision of loan so far.

These simulations measure impacts of public investment on the MSE sector on

unemployment and poverty.

The second set of simulation scenarios focuses on more public effort on trainings

and better access to loan for the MSE sector as planned in the mid-term plan of

the Ethiopian government. The scenarios in this simulation are:

1. Raising MSEs’ TFP assuming total coverage of training in the entire MSE sector,

2. Capital growth in the MSEs sector in line with doubling of the accomplishment

in provision of loan so far.

For both the simulations we also used some means of financing the intervention.

In the mid-term plan government of Ethiopia put the financial requirement to

accomplish this plan. It is explicitly stated that around 45% of the finance is

expected from domestic sources and the remaining 55% from external sources

(MoFED, 2010). Thus, financing from domestic source is simulated as dissaving

from government account and financing from external sources is simulated as

increased transfer to the government.

The effects of the intervention on productivity and capital growth cannot be found

for free. The cost should be considered. Thus, increase in government

expenditure is introduced as a shock which is calculated to be equivalent with

government’s expenditure for the accomplishment of the plan.

The outcomes of the second set of simulations show us how large the role of

MSEs would have been in terms of reduction of unemployment and poverty if the

interventions went as planned. Thus, based on this difference on the outcomes

we recommend some alternative strategies for MSEs’ development.

6.2.2 Analysis of the results From the CGE side, as we can get economy wide effects of the interventions, we

would first go for the direct or first hand effects.

6.2.2.1. Effects on production and prices

A shock with increased productivity together with increased capital stock in the

MSE sector only definitely, will come up with increased output and decreased

commodity price. Production increases in the MSE sector significantly with a

slight decrease in the non- MSE’s output as result of competition between the two

sub-sectors. Since outputs from MSE and non-MSE in the same sector are made

substitutable, increased supply of the products from the MSE sub-sector as a

result of the interventions which favor this sub-sector, makes the product from

MSE activities cheaper than those produced from the Non-MSEs. Following the

reduction in demand, price in the non-MSE sub-sector also decreases after

wards. (See table 6.5)

Table 6.5: Changes in production and price

Source: Simulation results Administration activity expands significantly as a result of the increased

government expenditure. The highest share of government’s expenditure on goods

and services is spent on public administration services. Thus the biggest effect of

the increase in government expenditure is exhibited on administration service.

On the other hand, the positive effect from the government side is curved by a

negative effect from the rural households’ side in the case of other services. As a

result of income effect, rural households are spending less on services from

activities in the other services. The effect of this only reduces the magnitude of

the change but not the direction.

In the 2nd simulation, accomplishment of government’s plan in MSE sub-sector

can come up with larger effects. The efficiency and capital growth interventions in

the MSE gives bigger hit to the sub-sector which results in big increase in supply

and highest drop in price. In the other sectors, the 2nd gives bigger effects than in

the 1st simulation in the same dimension.

Base Sim1 % Sim2 % Base Sim1 % Sim2 %

MSE 74 81 9.1% 85 15.2% 1 0.93 -6.7% 0.90 -9.8%

Non-MSE 182 180 -1.1% 179 -1.3% 1 0.99 -1.0% 0.99 -1.2%

Agriculture 190 190 0.1% 192 0.9% 1 1.00 -0.1% 1.00 0.0%

Other industry 34 34 1.6% 35 2.9% 1 1.00 -0.3% 1.00 -0.1%

Administration 21 22 5.0% 23 13.1% 1 1.01 1.4% 1.05 4.9%

Other service 99 101 2.3% 104 5.0% 1 0.99 -0.7% 0.99 -0.7%

Production Price

6.2.2.2 Effects on factors of production and returns

As our major intervention is increasing efficiency of the factors in the MSE

activities, it means we need less number of factors to produce equal amount of

the output from the sector. Thus, at this point what matters most is the demand

for that product. If the economy creates higher demand for that product for some

reason, then the negative impact of the efficiency shock on labor demand may be

netted out by the positive effect on labor demand from increased demand for that

particular output.

In our case, the positive effect couldn’t net out the negative effect. Thus the

efficiency shock resulted in significant reduction in labor demand for all the skill

levels in the MSE sub-sector. Regarding gender difference, female workers face

lower lay off rate than their male counter parts at any skill level. Specially, in

skilled labor market male workers’ demand fall bigger and significantly than the

demand for the female. (See table 6.6)

On the contrary, rise in labor demand is exhibited in the Non-MSE sub-sector.

This is mostly due to the expansion in the administration and other services

sectors as a result of increased government’s spending. Since the proportion of

skilled workers is higher than the unskilled in the Non-MSE sub-sector, the

positive effect is stronger on the demand for skilled workers.

Table 6.6: Simulation results on labor demand in MSE and Non-MSE sub-sectors

Source: Simulation results In the specification of the model, as already discussed in the methodology

Base Sim1 % Sim2 % Base Sim1 % Sim2 %

Unskilled female 2.8 2.7 -5.1% 2.3 -19.6% 68.6 68.7 0.3% 69.6 1.5%

Unskilled male 4.0 3.8 -5.5% 3.0 -25.3% 74.1 74.3 0.4% 75.4 1.8%

Skilled female 2.7 2.6 -3.4% 2.5 -5.6% 6.7 6.8 1.3% 7.0 4.2%Skilled male 5.3 5.0 -6.1% 4.3 -18.5% 25.3 25.7 1.6% 26.6 5.1%

Non-MSEMSE

section, any worker whether male or female and skilled or unskilled first goes to

the non-MSE sub-sector searching for job. If he is not successful then he will go

to the MSE sub-sector. If he is not still successful then he will be unemployed.

In the 1st simulation, the accomplishment of the plan so far can be somehow

considered as ineffective as we measure it from employment creation (reduction

of unemployment) goal. Demand for the entire labor force in the MSE sub-sector

reduces by around 5% due to increased factor productivity. However the effects of

the other interventions in the simulation make labor demand in the Non-MSE

sub-sector grow slightly by 0.5%. As a result of Non-MSE sub-sector’s higher

weight in the labor market, the overall effect in the entire economy becomes a

slightly increasing labor demand. Consequently, the overall unemployment

reduces slightly. However, unemployment of male workers decreases at a

magnitude of from 1.5% to 2.2% for unskilled and skilled respectively. On the

contrary, the vulnerable groups are not still beneficiary of the intervention to

develop the MSE sub-sector. Female unemployment is rising for both skill levels.

In fact, unemployment in unskilled female increases slower than unemployment

in skilled female.

However, if MSEs’ development plan was achieved completely, unemployment

would drop by maximum of 34% in the case of male unskilled workers. Unskilled

female workers get the 2nd biggest reduction in unemployment which is 29%.

Skilled workers also enjoy higher demand in the labor market which drops the

number of unemployed significantly. Thus full accomplishment of the

development plan is a must to come up with a meaningful reduction in

unemployment.

Table 6.7: Simulation effects on unemployment and wage rate

Source: Simulation results

As it is already stated, wage rate in the Non-MSE sub-sector is made exogenous

which cannot be changed by market forces rather by policy decisions. Thus the

only change we can discuss on is the MSE wage rate. As we can see from table

6.7 above, female wage rate in the MSE sub-sector for both skill levels decreases

and that of male increases. The magnitude of the change is a bit stronger for

skilled workers than unskilled counterparts for both male and female.

On the other hand, the economy wide return to capital also decreases on average

in both simulations by 11% and 17% for sim 1 and sim 2 respectively. The

increased productivity which is coupled with growth in capital stock in the MSE

sub-sector resulted in biggest (i.e 27%) fall in capital return. Following

government’s dissaving to finance its increased expenditure, investment in the

country shrinks as government’s borrowing from domestic banks crowd out the

private sector. As a result, there is a negative effect on capital demand and

returns to capital from those industries that supply their output for investment.

On the other hand, larger share of the outputs from these industries is consumed

by the other industries as intermediate inputs or consumed by households. Thus,

the increased demand for their output from households and other activities

increases their demand for capital and returns to capital. Since this positive

effect is stronger, then it wiped out the negative effect. Therefore there is a slight,

a 0.7%, increase in capital return from the Non-MSE sub-sector side.

Base Sim1 % Sim2 % Base Sim1 % Sim2 %

Unskilled female 0.27 0.27 1.5% 0.19 -28.9% 1 0.99 -0.7% 1.21 20.8%

Unskilled male 0.14 0.14 -1.5% 0.09 -33.6% 1 1.01 0.9% 1.34 34.0%

Skilled female 0.27 0.28 3.0% 0.25 -6.3% 1 0.99 -1.5% 1.03 3.4%

Skilled male 0.14 0.13 -2.2% 0.11 -21.2% 1 1.02 1.6% 1.18 18.4%

Unemployment MSE wage rate

6.2.2.3 Effects on household income, consumption and poverty

Household income decreases for both household types. As shown in figure 6.3

below, income from almost all sources decreases in the 1st simulation. Rural

households exhibit a higher fall in income as compared to their urban

counterparts. Urban households got a slight increase in income from their labor

force which is the highest contributor to their income. That’s why the higher drop

in income from capital and government transfer is minimized by the smaller

increase in income from the labor side. If we look at the 2nd simulation, complete

accomplishment of the plan can make urban households better off. Rural

households still face a reduction in their income but at a lower rate than in the

1st simulation.

Regarding consumption, rural households remain intact whereas urban

households enjoy rise in consumption. Households’ better position in terms of

consumption despite the decrease in income is completely the effect of a bigger

drop in price as compared to the decrease in income. This technically means the

households are now richer in real terms, though their nominal income is falling.

Table 6.4: Simulation effects on household income and household consumption

Source: Simulation results As it can be clearly seen in figure 6.3, households’ income from labor force

increases in the 2nd simulation bigger than the other sources. It can compensate

the fall in income from the other sources and come up with a positive effect on

household income in the case of urban households but it cannot compensate the

fall in income in the case of rural households.

BASE SIM1 % SIM2 % BASE SIM1 % SIM2 %

Rural households 243.3 240.6 -1.1% 241.8 -0.6% 193.9 193.9 0.0% 195.4 0.8%Urban households 90.5 90.1 -0.4% 91.6 1.2% 60.5 60.9 0.8% 62.1 2.6%

Household income Household consumption

Figure 6.3: Sources of household income

Source: Simulation results

Implications of the interventions on price and households’ income make the

households consume more or the same amount (from the BASE) as we discussed

above. This intern has its own implication on households’ poverty. As shown in

figure 6.4, the current accomplishment of the plan has brought no poverty impact

nationwide. Rural poverty also remains unchanged. The only success registered in

terms of poverty reduction with the current accomplishment of the plan is

observed in the urban centers, which is even a slight change, not more than 0.6

percentage points.

However assuming complete accomplishment of the plan, national poverty reduces

slightly by around 1 percentage point. Rural households benefit very slightly, with

0.7 percentage points reduction in poverty while urban households enjoy the

maximum poverty reduction, around 2 percentage points. Looking at the poverty

results, it seems that price effects come out to stronger than the income effects.

Figure 6.4 Poverty implications of the plan

Source: Simulation results 6.2.2.4 Effects on investment and Gross Domestic Product

The plan to develop the MSE sub-sector in Ethiopia is planned to be financed from

both domestic and external sources. Government borrows some portion of the

spending to finance the plan from domestic banks which directly crowds out the

private investors. This effect is clearly seen on the simulation results. Total

investment expenditure drops by 2% to 3.5% in the 1st simulation and the 2nd

respectively. (See figure 6.4)

Shrinking investment has a negative impact on the country’s GDP. However, the

expansion in the MSE sub-sector and the other Non-MSE activities as result of the

interventions come up with a positive impact on the GDP. This positive effect is

coupled with a significant drop is price. The consumers’ price index (CPI) falls by

1.2% in sim1 and 1.5% in sim2. These give the country’s GDP grow in real terms

by 1% to 2.5% in the 1st and 2nd simulations respectively. This is the other

significant positive effect of government’s effort to develop the MSE sub-sector.

(See figure 6.4)

Figure 6.4: Simulation effects on macro variables

Source: Simulation results

7. Conclusion and Recommendation The major objectives of this study are to examine the factors that constrain enterprise

growth as given by either capital or employment growth and to assess the role of MSEs

to reduce unemployment and poverty. Having these objectives this study is very timely,

because the government of Ethiopia is expecting a lot from the MSE sector which is not

yet developed. Econometrics model is used to reveal the roadblocks towards the sector’s

growth. We used relatively new data set which has 3000 observation from around the

country. Besides we used CGE modelling to answer the 2nd research question.

From the econometrics results we can conclude that gender, start-up capital and

educational background of owners can significantly determine growth of the firm which

is measured by growth of capital. Firms with lower startup capital or smaller firms are

found to be fast growing. Besides, firms lead by men or educated person are found to be

more successful.

From the CGE modelling results, the government is somehow achieving success on both

employment creation and poverty reduction. Despite an increase in female

unemployment, reduction in male unemployment is registered from the current

performance of the development plan. Besides, as a result of significant decrease in

consumer price index, now cost of living becomes cheaper. That has also implications on

poverty in the second simulation. Specially, urban poverty reduces by around 2

percentage points. However, the current accomplishment of the plan brings about

almost no effect on poverty reduction. Unemployment for both male and female workers

and poverty in both areas would decline significantly if the plan gets complete

accomplishment at the end of the day.

Based on these results, the policy recommendations of this study are; 1st Government

has to focus on small scale firms since they are found to grow fast than the bigger firms.

Since these firms are easier to establish and fast to grow, then they can be used as

actual solution for poverty eradication policy of the government. Thus small scale firms

need due consideration during policy formulation. 2nd Training the operators in the MSE

sector is found to be critical. Since most of the operators are having lower level of

educational background, some sort of training can come up with bigger effect on the

firms’ growth. From the recent survey around 65% of the operators reported that they

haven’t got training opportunity. Thus, this is one critical point on which the

government must work a lot. Giving due attention to female entrepreneurs is the other

important point to reduce unemployment and poverty. 3rd The construction sector and

firms in this sector are found to be the most successful among all types of firms. Thus

more and more number of firms should be established to work in this sector. 4th The

government should strive to live up to the MSE development plan since it is proven that

the complete accomplishment of the plan can come up with a very significant effect on

unemployment and poverty.

8. Policy influence (or research communication strategy)

The following table constitutes a list of institutes that we targeted for effective

implementation of the policy prescriptions. We are approaching each of the following

institutions to gather information about the actual operation of the sector. We also

have continuous consultation with each of them about the challenges and constrains

of the sector.

Institution Contact Target

Micro and Small Enterprises Development Agency Yes

MSE Development Offices Yes

Ministry of Urban Development and Construction Yes

Addis Ababa City Administration MSE section Yes

Ministry of Women's, and Children Affairs Yes

We have already done discussion with few government officials at these

institutions and with some researchers who have been engaged in MSE and

microfinance related studies currently or in the past.

During these sessions the team introduced this research work and undertook

discussions on simulation areas and variables, structures and design of the

models (i.e both the econometrics and the CGE) and possible implications of the

output. Besides, we closely follow government policies and strategies to adjust our

models and simulations. We strongly believe that this allows policy makers and

other concerned bodies to follow the progress of the study.

As part of our dissemination strategy, we will have meetings with experts from

MSEs Development Agency, and experts on MSE related works from Ministry of

Urban Development and Construction, Addis Ababa City Administration MSE

section, and Ministry of Women's and Children Affairs. During these meetings, the

team will be presenting the findings of the research work and the policy

recommendations of the study. We expect insights from these experts how to use

the results of the study as an input to develop strategies and policy formulations

even in future.

During our discussion with few of government officials, we learnt that they are

keen to review this kind of studies and use it as input in developing alternative

strategy and even to supplement their strategies for the sector’s development. We

strongly believe that the findings of this study will be asset for policy makers in

allocation of resources for development of MSEs and to have clear understanding

of the bottlenecks of the sector, design to solve factors that hinder MSEs’ growth;

and maximize the contribution of the sector to reduce unemployment and poverty.

The team has also plan to organize a workshop to present these findings to policy

makers and academicians, to get their feedback and comments which we believe is

invaluable to improve the quality and relevance of the work. We have also plan to

present the paper during the coming EEA (Ethiopian Economics Association)

annual international conference. From past experiences, this conference is well

known for creating a platform where policy makers, academicians and

practitioners are all gathering. As a result of this conference, we anticipate that our

findings can be more widely disseminated and we will benefit from constructive

comments and feedbacks from participants with different insights which will

strengthen the paper. Besides as a strategy to reach the wider audience, we plan to

publish the final paper on one of peer reviewed local or international journals.

9. List of team members

Name Age Sex (M,F) Training and experience Ermias Engida 31 M Holds M.Sc. in economics (Applied Trade

Policy Analysis). He took his B.A in

economics. He has taken several

professional trainings on CGE modeling,

GAMS software, and applied Micro

econometrics by renowned professors with

several years’ experience in teaching and

research. He has extensive experience in

macro-economic modeling. He has played

vital role in the design of Ethiopia’s

Growth and transformation plan

employing CGE modeling and examined

different aspects of the plan overtime

including alternative financing options. He

has also applied CGE modeling to examine

agricultural productivity, public

investment, role of livestock, and public

services. He also applied CGE to examine

alternative policy schemes and their

implication to welfare. His areas of

research are multi sectorial and goes

beyond national boundaries as could

easily be seen from attached CV in detail.

He is working as a research officer at

International Food Policy Research

Institute (Ethiopian Strategic Support

Program) where he has produced several

policy relevant papers independently and

jointly with varied international

researchers. As part of Ethiopia strategic

Support program’s capacity building

project he is advising several graduate and

post graduate students with their thesis

on CGE modeling. Before he joined IFPRI

he used to teach at Arba Minch University

where he lectured core departmental

courses. He is fluent in English and

Amharic languages.

Ibrahim Worku 33 M Holds M.Sc. (Economic Policy Analysis)

and B.Sc. in economics. He has also

successfully completed training by the

World Bank on “Impact Evaluation in

Agriculture and Community Driven

Development Program”. Economic policy

analysis using GAMS and CGE Software.

He works for International Food Policy

Research Institute (Ethiopian Strategic

Support Program and Ethiopian

Development Research Institute) as

research officer. In this capacity he has

written and published several papers

relevant for both the academia and to

inform policy. He has expert level skills in

several statistical and econometric soft

wares including STATA, SPSS, and GAMS.

He has extensive experience in

coordinating large surveys. He also has

proven experience in cleaning, analyzing,

and publishing papers based on these

massive datasets. Before he joined IFPRI,

he was teaching as a lecturer and working

on different projects as a researcher at

Addis Ababa University, department of

Economics. He also briefly took a research

position at Ethiopian Development

Research Institute (EDRI) where he

collaborated with experienced researchers

in several studies. He is fluent in English

and Amharic languages.

Mekdim Dereje

33 M Holds M.Sc. (Finance and economic

development) and B.Sc. in economics. He

has successfully completed training in

survey management and CSpro Data entry

and processing application. He works for

International Food Policy Research

Institute (IFPRI). At IFPRI, he has,

independently and jointly with colleagues,

published in working papers and peer

reviewed journals. He has also taken part

in report writing, survey coordination,

data management, analysis and

presentation at conferences and seminars

organized by government offices, research

institutes and the academia. He has

expert level skill in several statistical and

econometric soft wares including STATA,

SPSS, and GAMS. He has extensive

experience in coordinating large surveys.

He also has proven experience in cleaning,

analysing, and publishing papers based

on these massive datasets. Before he

joined IFPRI he used to be a lecturer at

Haramaya University, department of

Economics where he participated both in

research and teaching. He also worked as

a research assistant in Copenhagen

University, Denmark and chief economist

in Wabekon Consult. He is fluent in

English, Afan Oromo and Amharic

languages.

Feiruz Yimer 31 F Holds M.Sc. (Economic Policy Analysis)

and B.Sc. in economics. She has also

successfully completed training

Economics policy analysis using CGE and

GAMS software. She has also taken

training on “Impact evaluation on

agricultural and community driven

development programs”; Training on

leadership skills and training on

GAMS/CGE –dynamic version. She works

for International Food Policy Research

Institute (Ethiopian Strategic Support

Program and Ethiopian Development

Research Institute) as research officer. In

this capacity she has written and

published several papers relevant for both

the academia and to inform policy. She

has expert level skill in several statistical

and econometric soft wares including

STATA, SPSS, and GAMS. She also has

proven experience in cleaning, analyzing,

and publishing papers based on massive

datasets. Before she joined IFPRI, she was

teaching as a lecturer and working on

different projects as a researcher at Addis

Ababa University, department of

Economics. She served as a coordinator of

Gender office of Faculty of Business and

Economics at Addis Ababa University.

Addis Ababa, Ethiopia. She has also

participated in several community

development programs such as creating

awareness to youth regarding HIV-AIDS

and importance of women empowerment.

She is fluent in English and Amharic

languages.

Saba Yifredew 31 F Holds M.A. (Economics of International

Trade), M.A (Economics) and B.Sc. in

economics. She has also successfully

completed training on Economics policy

analysis using CGE and GAMS software;

Training on leadership skills and training

on GAMS/CGE –dynamic version. She is

currently the head of the department of

economics at Addis Ababa University. She

is also Academic Programs Unit Head of

Addis Ababa University. She has

independently and jointly with colleagues

at Addis Ababa University and elsewhere

written several research papers. She has

extensive experience in report writing and

communication with affiliate

organizations.

10. Capacity building

The research team is composed of individuals with different specializations. For

some of the team members this is the first CGE type research (Mekdim and

Ibrahim) while the experiences of the three remaining researchers ranges from

more than 6 years (Ermias) to roughly 2 years (Saba and Feiruz). This, therefore,

creates an excellent platform, particularly, for two new members to grasp the

basics of CGE modeling. The team leader devotes some time introducing the

different steps of CGE modeling including data compilation, model implementation,

simulation and debugging to Mekdim and Ibrahim. On the other hand, Ibrahim and

Mekdim are well acquainted with econometric techniques pertinent to establish the

type of relationship we seek in this study including dichotomous data model

estimation and instrumental methods (IV method). This gives opportunity for

Ermias and possibly Saba and Feiruz to learn rigorous econometric techniques.

The team members also hugely benefit from proved report writing experiences of

Saba and data management skills of Feiruz. This varied, yet vital, skills help the

team to produce quality research while simultaneously equip members in their

future research endeavor.

Below we present specific tasks each team member would carry out in executing

the project:

Name Task/contributions Ermias Engida He is the team leader. Overall responsible in

coordination of activities. He serves as a focal person for meetings with government officials and MSEs leaders, and communicating proposal and research outputs of the team to all stake holders. He also takes active part in CGE modeling and mentoring of team members to develop capacity for future research.

Ibrahim Worku Responsible to coordinate the sub-team that do econometric modeling. He focuses on running and testing the binary and IV regressions and establishes the results with different sensitivity tests.

Feiruz Yimer Coordinates the econometric work to estimate elasticity for the CGE part. She was also responsible to clean and make ready the data that is used in this estimation. She also helps in organizing and synthesizing pertinent literatures and to cross check the consistency of the result with other studies. She was also responsible for writing the section on conceptual framework of the basic models.

Mekdim Dereje Mainly worked on the descriptive analysis of the study. He produced tables and graphs on key variables to complement the econometric modeling. Together with Feiruz, he also took active part in literature review.

Saba Yifredew While taking active part both in CGE and econometric modeling, she also coordinates the report writing. She takes care of the coherence between the different sections of the report. Her extensive experience in this regard was essential for other team members to share her experience.

11. List of past, current or pending projects in related areas involving team members

Name of funding

institution Title of project Team members involved

ReSAKSS-ECA

Role of livestock in the Kenyan Economy: Policy Analysis using Dynamic CGE model for Kenya

Ermias Engida

PEP

Alternative Policy Strategy to ADLI for Ethiopia: A Dynamic CGE Framework Analysis

Ermias Engida

IFPRI

Ethiopia’s Growth and Transformation Plan: A CGE Analysis of Alternative Financing Options

Ermias Engida

Ethio-Telecom

Three rounds of Customer satisfaction survey

Saba Yifredew

World Bank

Responsiveness of Rural Households to cereal Price Changes

Saba Yifredew

DFID

Exploring Demand and supply factors behind recent cereal prices

Saba Yifredew

Ethiopian Development Research Institute

Inflation-Growth nexus- Estimation of Inflation threshold for Ethiopia

Saba yifredew

Ethiopian Development Research Institute

Cereal Consumption and Demand Patterns in Urban Ethiopia

Saba Yifredew

Ethiopian Development Research Institute

Road Sector Development and Economic Growth in Ethiopia

Ibrahim Worku

12. References

Ageba, Gebrehiwot. and Wolday Amha. (2006). Micro and Small Enterprises (MSEs) Finance in Ethiopia: An Empirical Evidence. Eastern Africa Social Science Research Review, Volume 22, Number1, pp. 63-86, Michigan State University Press, DOI: 10.1353/eas.2006.0002

Annabi N., Cockburn, J., Decaluwe, B. (2004). A sequential Dynamic CGE model for poverty analysis.

Arndt C., Benfica, R., Thurlow, J. (2011). Gender Implications of Biofuels Expansion in Africa: The Case of Mozambique. World Development Vol. 39, No. 9, pp. 1649–1662

A.R. Quisumbing, Lynn R. B., Hillary S. F., Haddad L., Pena C. (1995). Women: The key to food security. Food policy report. The international food policy research institute. Washington D.C.

Aynadis, Zemenu and Mohammed Mohammednur (2014). Determinants of Growth of Micro and Small Enterprises in ETHIOPIA: (A Case of MSEs in Mekelle City, Tigray). International Journal of Advance Research in Computer Science and Management Studies, Volume 2, Issue 6.

Aubert, P. and Crépon, B., (2006): Age, Wage and Productivity: Firm-Level Evidence

Beck, T., & Demirguc-kunt, A. (2005). Small and Medium-Size Enterprises : Overcoming Growth Constraints. Journal of Banking & Finance, 30(February), 1–31.

Bekele E. & Muchie M, 2009: Promoting micro, small and medium Enterprises (MSMEs) for sustainable rural Livelihood Development, Innovation and International Political Economy Research (DIIPER); Aalborg University, Denmark; DIIPER Research Series; Working Paper No. 11

Cameron, A. C. (2007). Panel data methods for microeconometrics using Stata, 1–39.

Central Statistical Authority (CSA). 2003. Survey of urban informal sector operations. Addis Ababa.

_____. (CSA) 2003. Report on small scale manufacturing industries survey. Addis Ababa.

Coad and Tamvada (2008): The Growth and Decline of Small firms in Developing Countries; working paper; Max Planck Institute of Economics Evolutionary Economics Group

CSA (2014) Central Statistics Authority “Statistical Report on the 2014 Urban Employment Unemployment Survey” Statistical Bulletin. Addis Ababa, Ethiopia

Evans, David. S. (1987). The relationship between firm growth, size and age: Estimates for 100 Manufacturing Industries. Journal of Industrial economics, Vol. 35, No. 4, pp.567-581

Evans, David S. (1987). Tests of Alternative Theories of Firm Growth. Journal of Political Economy, Vol. 95, No.4, pp. 657-674, The University of Chicago Press

Ermias E., S. Tamru, E. Tsehaye, Dario D., Paul D., and Sherman R. (2011). Ethiopia’s Growth and Transformation Plan: A Computable General Equilibrium Analysis of Alternative Financing Options. ESSP II Working Paper 30. Addis Ababa, Ethiopia. Ethiopia Strategy Support Program II

Foster, James, J. Greer, and Eric Thorbecke. 1984. “A Class of Decomposable Poverty Measures.” Econometrica 52 (3): 761–65.

Garoma, Belay File (2012). Determinants of Microenterprise Success in the Urban Informal Sector of Addis Ababa: A Multidimensional Analysis. Rotterdam, International institute of social studies (ISS) of Erasmus University

Gebreeyesus, Mulu (2007). Growth of Micro-Enterprises: Empirical evidence from Ethiopia. Ethiopian Development Research Institute (EDRI), First Draft – February, 2007

Gebremichael, Biru Ashenafi (2014) The Impact of Subsidy on the Growth of Small and Medium Enterprises (SMEs). Journal of Economics and Sustainable Development, Vol.5, No.3.

Goedhuys, Micheline, and Sleuwaegen Leo (2009) High-Growth Entrepreneurial Firms in Africa: A Quantile Regression Approach. Research Paper No. 2009/11, UNU-WIDER

Greene, W. H. (2003). Econometric Analysis. (P. Education, Ed.)Journal of the

American Statistical Association (Vol. 97, p. 1026). Prentice Hall. doi:10.1198/jasa.2002.s458

Hagos, Hailay Aregawi., Aregawi Gebremichael, Assmamaw Getie (2014) Determinants of Micro and Small Enterprises Growth in Rural Area: Evidence from Feresmay Town. Journal of Economics and Sustainable Development, Vol.5, No.19

Hall, B. H. (1987). The relationship between firm size and firm growth in the US manufacturing sector”, Journal of Industrial economics, Vol. 35, No. 4, pp.583-606

Kinyondo G. and Margaret M. (2009). The general equilibrium effects of a productivity increase on the economy and gender in South Africa. Department of Economics, University of Pretoria

Liedholm Carl (2001) Small Firm Dynamics: Evidence from Africa and Latin America. The International Bank for Reconstruction and Development, Stock No. 37189, The World Bank, 1818 H Street, N.W., Washington, D.C., 20433, U.S.A.

Lofgren H., R. L. Harris, and Robinson, S. (2002).A Standard Computable General Equilibrium (CGE) Model in GAMS.Trade and Macroeconomics Division Discussion Paper 75. Washington, D.C: International Food Policy Research Institute

Mcpherson, M. A. (1996). Growth of micro and small enterprises in southern Africa, 48, 253–277.

Mead, Donald C. and Carl Liedholm (1998) The Dynamics of Micro and Small Enterprises in Developing Countries. World Development, Vol. 26, No. 1, pp. 6 I-74, Michigan State University, East Lansing, U.S.A.

Ministry of Urban Development and Construction. (2013). Survey on Micro and Small Enterprises (MSEs) in Selected Major Cities of Ethiopia. Addis Ababa, Ethiopia.

MoFED. (2014). Ethiopia’s Gowth and Transformation Plan (Vol. I). Addis Ababa, Ethiopia.

Nichter, Simeon and Lara Goldmark, 2009, Small Firm Growth in Developing Countries, World Development, Vol. 37, No. 9, pp. 1453–1464, 2009

Paul Dorosh and James Thurlow (2009). Implications of Accelerated Agricultural Growth on Household Incomes and Poverty in Ethiopia: A General Equilibrium Analysis. Discussion Paper 002. Addis Ababa, Ethiopia. Ethiopia Strategy Support Program II

Robu M., 2013: The dynamic and importance of SMEs in economy

Springer, K. (2010) Climate Policy and Trade: Modeling dynamics in a multi-regional framework. Kiel Institute of World Economics

Tadesse, Brhane (2014). Access to Finance for Micro and Small Enterprises in Debre Markos Town Ethiopia. Global Journal of Current Research Vol. 2 No.2., Pp.36-46

Tefera, Habtamu., Gebremichael, Aregawi. and Abera., Nigus (2013). Growth Determinants of Micro and Small Enterprises: Evidence from Northern Ethiopia. Journal of Economics and Sustainable, Development, Vol.4, No.9

Thurlow, J. (2008). A recursive dynamic CGE model and microsimulation poverty module for South Africa. International Food Policy Research Institute, Washington, D.C.

Wasihun, Rahel and Issac Paul, (2010). Growth Determinants of Women-Operated Micro and Small Enterprises in Addis Ababa. Journal of Sustainable Development in Africa (Volume 12, No.6,), Clarion University of Pennsylvania, Clarion, Pennsylvania

Wooldridge, J. M. (2002). Econometric Analysis of Cross Section and Panel Data. (M. Cambridge, Ed.)booksgooglecom (Vol. 58, p. 752). MIT Press. doi:10.1515/humr.2003.021

You, J.-I. (1995). Small firms in economic theory. Cambridge Journal of Economics, 19:441–462.