Enterprise dynamics and employment elasticities, 1986-2006

Enterprises and Employment Growth in Bangladesh

Propoor growth, and the role of spurring the growth of  SMEs in it, are the defining characteristics of a paradigm that has already emerged in the practice of international development.  Bangladesh is no exception to this general rule. Employment elasticity of output is among the focal metrics in evaluating the pattern of growth, and this short note makes a modest contribution of that genre.  A number  of major efforts in Bangladesh contributing in this context range from technical assistance for fostering demand-driven enablers affecting the diffusion of better business practices, to understanding the manner how investment-climate impediments drag firm performance in SMEs sector (IFC/SEDF, 2006), tackling the policy issues surrounding SMEs’ access to finance (Chowdhury and Miah, 2006), launching new nation-wide supply- and demand-side surveys tasked to distill from primary data guidelines for implementing Bangladesh government’s SME Policy Strategies-2005 within the framework of the PRSP (SMESDP, RFP, 2006).  Bangladesh has also been liberalising its economy since 1993, and has also gone along some distance towards substantial privatisation.  How manufacturing and services subsectors have fared in the period 1986 – 2006 in terms of the growth especially of the employment generated by enterprises would perhaps be useful.   

Before proceeding any further, we must quickly define “small”, “medium” and “large” enterprises used in this paper.  BBS informally defines “small” firms as those with between 10 and 49 employees; and medium firms, with between 50 and 99 workers.  Anything larger in employment is deemed large.  Units with between 1 and 9 employees are considered “micro”.  

The objective of this short paper is to process data from three economic censuses carried out over the last twenty years by Bangladesh Bureau of Statistics (BBS), and throw some new light, especially relating to SMEs.  These censuses were carried out in 1986, 2001/2003, and 2005/2006.[1]  The first two were full-blown enterprise censuses, and have been reported on by the BBS (BBS, 1992; BBS, 2004).  The scope of the last was limited to that of a Business Registry (BR) of all enterprises in Bangladesh with an employment of at least 10 workers.  The latest updating as of 2005/2006 did not cover establishments with between 1 and 9 workers, which, it must be said, account for an overwhelming proportion of all establishments in the country. While decomposing changing employment among its various contributory components is our main concern, we offer derivative observations about the employment elasticity of growth in Bangladesh.

The following scheme is used for classifying 2-digit entities under Bangladesh Standard Industrial Classification (BSIC): (a) manufacturing---10 through 45; (b) trade and eateries---50, 51, 52, 55; (c) finance and business services---60 through 74; and (d) other services---75 through 93.

Decomposition of jobs growth in economic development

The economy of Bangladesh grew at 5% or so between 1986 and 2006.  Per-capita income growth during that  same period of some 3.4% or so was certainly not tepid.  It is certainly worth testing whether any resulting demand stimulus did perceptibly alter the industrial composition of either SMEs or large enterprises. Organic growth by secularly increasing average enterprises size.  It would be nice to know if a strong size effect is discernible in the data, as a source of employment growth. The more is the increase in average size of enterprise from one period to the next, relative to the change in the number of enterprises alive, the greater would be the percentage contribution of the pure enterprise-size effect as a growth source.  The “dynamic” is the “industrial-composition” setting, the larger is likely to be the pure “composition” effect in percentage terms. 

Decomposition methodology, and the data

Tables-1 and 2 present the underlying data required to implement the MV decomposition.

Table-1

The underlying data relating to SMEs

Industry class	1986			2006			% change between 1986 and 2006
	No. of SMEs	‘Industry’ mix	Size Per unit	No. of SMEs	‘Industry’ mix	Size Per unit
							SME Number	% share	Size Per Unit
Manufac- ing	17033	41.3	22.9	29273	37.8	23.7	71.86	-3.5	3.49
Trade	7765	12.97	15.7	7509	7.05	17.2	-3.3	-5.92	9.55
Financial Business srvcs	6494	12.4	18.0	9557	11.35	21.8	47.1	-.95	21.1
Other srvcs	18045	33.2	17.4	38281	43.7	20.96	112.1	10.5	20.46
Total	49337	100.0	19.1	84620	100.0	21.7	71.5	-	13.6

Source: BBS, 1992; BBS, 2006

Table-2

The underlying data relating to large enterprises

Industry class	1986			2006			% change between 1986 and 2006
	No. of large units	‘Industry’ mix	Size Per unit	No. of large units	‘Industry’ mix	Size Per unit
							Large unit Number	% share	Size Per unit
Manufac- ing	1854	80.6	449.6	4579	71.4	429.1	147	-9.2	-4.59
Trade	63	2.7	244.0	128	2.0	207.1	103	-.7	-15.17
Financial Business srvcs	150	6.5	277.3	413	6.4	319.1	175	-.1	15.16
Other srvcs	232	10.1	252.8	1291	20.1	298.5	456	10.1	18.07
Total	2299	100	412.8	6411	100	385.9	179	-	-6.5

Source: BBS, 1992; BBS, 2006

The decomposition methodology

In obtaining the sources of growth in terms of total employment outside agriculture we customise the Minhas-Vaidyanathan (MV) decomposition scheme.  The MV scheme decomposes the employment growth into (i) an enterprise-change effect; (ii) an industrial-composition effect; (iii) an enterprise-size effect; and (iv) a couple of interaction effects.

E_t – E₀ = ( N_t - N₀) åⁿ_i=1 S_i0C_i0 + N₀ åⁿ_i=1 (S_it - S_i0) C_i0 + N₀ åⁿ_i=1 S_i0(C_it - C_i0) + N₀ åⁿ_i=1 (S_it - S_i0)(C_it-C_i0) +( N_t - N₀) åⁿ_i=1 S_i0(C_it-C_i0) + ( N_t - N₀) åⁿ_i=1(S_it - S_i0)(C_it-C_i0)                (1)

where

E=   Total employment by enterprises;

N = Number of enterprises;

S =  Average size per establishment (or enterprise);

C = percentage share of an industry-group in total employment;

i   = 1, 2, 3, 4 (and relate to manufacturing, trade, finance and business services and other services);

t, o  = time subscripts

Before going into the specifics of the results relating to the decomposition itself, we should perhaps take note of the principal findings from Tables 1 and 2.  Between 1986 and 2006, annual growth in number of enterprises and the employment by them grew at 1.24% and 1.8%, respectively.  In contrast, real output outside agriculture has grown  at about 6% per year.  What do this pair of estimates tell about the employment elasticity of non-agricultural output in Bangladesh during these two decades?  They say that employment elasticity of output in non-agriculture is less than a third, or so.  

Table-3

Growth rates of the number of enterprises and employment in manufacturing and services, Bangladesh, 1986-2006

(Percentages)

Type of output	Compound annual growth rates, 1986-2006, in
	No. of enterprises	Employment	Output
Manufacturing	1.27	1.7	6.0
Services	1.22	1.98
Both	1.24	1.8

Note: Between 1986 and 2006, the number of SMEs in manufacturing registers an annual compound growth rate of 1.18%. For large enterprises in manufacturing, the compound annual growth rate over the same period is 1.98%. The corresponding SMEs growth rate for all the other three kinds of services taken together is 1.176.    For large-enterprise services of all kinds, the corresponding growth rate is  3.12%. 

Explaining the decomposition of enterprise employment generation

It is the net enterprise creation that predominates any other source of growth in employment in Bangladesh’s enterprise: net job creation is driven by net growth in the number of enterprises in the economy. This effect is quite pronounced when we treat SMEs and large enterprises separately from each other. That said, there is a twist in the data.  When we lump the data together, the dominance, just witnessed, of net enterprise creation is seen to wilt somewhat.  Thus, whereas the contribution of net enterprise creation contributed 85% of the growth in the total SME employment between 1986 and 2006, and whereas large enterprises contributed 108% of the growth  in the total large-enterprise employment, when we treat the enterprise sector as one lump, that percentage falls to 71% alone.   

Secondly, changes in average size of enterprise is also found to be a much more substantial contributor to the employment creation by the enterprise sector than when SMEs are separated from large enterprises. Roughly, a quarter of employment change is due to changing average size of the typical enterprise. 

Table-4

Relative contributions of net enterprise growth, average enterprise size, and others

(All numbers are percentages)

Particulars	Employment change in SMEs alone (%)	Employment change in large enterprises alone (%)	Employment change in all types of enterprises(%)
Pure growth in the number of  enterprises	85.4	108.2	71.0
Size-of-enterprise effect	13.4	-1.5	24.8
‘Industrial-composition’ effect	-0.6	-2.7	1.0
Interaction between size and composition	1.3	1.0	1.4
Interaction between number and composition	-0.4	-4.7	.8
Interaction between number, size and composition	1.0	-0.3	1.1
Total effects	100.0	100.0	100.0

Source: BBS data and author’s calculations

Factors determining subsector employment growth

Data from 49 BSIC 2-digit categories were further consolidated into 22 “super-categories”.  They exhausted the entire range of industries and services.  The following equation was fitted to the data:

∆lnN = lnα+ %_share +  ∆lnS + Y_ +                                                           (2)

where N = Employment in a sub-sector;

          %_share = subsector relative share in employment;

          S  = Average size of enterprise in a sub-sector;

          Y_{  =} Income elasticity of demand in a sub-sector;

              = An error term;

          lnα = A constant term;

          ∆    = Proportionate change between 2001/2003 and 2006.

Data relating to N and S are obtained from BBS (2004) and (2006).  The first source covers data from 2001/2003 Economic Census.  The second  source covers the data from the updating of the Business  Registry of SMEs and large enterprises as of 2005/2006.  The estimates relating to income elasticity of demand for the 22 sub-categories are obtained from Dorosh (1994), Ahmed and Shams (1993), Goletti (1993), Bakht et al., (1992), and other assorted  sources. 

The estimated coefficients are as follows:

∆lnN=-0.3629^*+0.5838 (%_share)+0.9194(∆ln S)^*+0.1991(Y_)^*      adj. r-squared = 0.62

          ( - 4.7 )     (1.0)                           (2.52)                  (4.2)

The symbol ^* denotes that the coefficient in question is significant at 5% error probability level.

The results suggest that subsector income elasticity of demand and proportionate change in the average size of enterprise are statistically significant determinants of proportionate changes in subsector employment.  A ten-percent increase in the typical size of enterprise leads to a nine-percent percentage increase in the subsector employment level.  Each doubling of the income elasticity of demand contributes almost a twenty basis point of growth in the subsector employment level.  Results suggest that growth in goods and services demand that is driven by the growth of incomes is crucially important driver of employment generation, as is the achievement of organic growth by enterprises already in existence, causing mean enterprise size to grow.

The responsiveness of employment generation to demand growth is nothing new.  It is however a long-run source. In Bangladesh case, moreover, income inequality is growing secularly (Sen et al. 2004).  Increasing incomes inequalities are likely to pare back income elasticities across the board.  The imperative is therefore for creating room for equalising, pro-poor growth.  We shall return momentarily to what kinds of public investment and “sectoral policies” are likely to spur such beneficial patterns of growth in Bangladesh.

The second important result points at “organic enterprise growth” as a driver of labour-intensive economic growth.  Achieving such growth typically stems from a subsector achieving higher total factor productivity (TFP) levels compared with its competitors.[2]   To quote from David Dollar, of the World Bank, and his colleagues: “We expect that part of inter-firm differences in total factor productivity reflect plant-level heterogeneity in technology and firm capability, but that part of it may reflect regional differences in investment climate as well” (Dollar, 2002).  Following this characterisation, we shall argue that enterprise strengths in terms of product positioning, technology platform choice, firm branding and design capability, firm commitment to product and process innovation are critically-important determinants of TFP growth within enterprises.  The most important supra-firm variables that may likely influence TFP growth include location within agile and learning eco-systems which might be in the form of creative clusters, and a small number of conducive investment-climate variables. These eight target strengths highlight the centrality of creating, managing and adapting knowledge, skills, organisation and infrastructural delivery as a well-orchestrated package.  A basis of high-quality information and insights illuminating the characteristics and the laws-of-motion of enterprises in general and SMEs in particular, cross-fertilised with intimate knowledge of where government and market failures create strong head-winds for TFP growth, is a pre-condition. There is an undeniable need for professionally competent policy-oriented applied enterprise research, generating primary data diligently, ferreting out market failures rigorously, embedding pro-poor institutional interventions in the bedrock of solid empirical knowledge, and fostering inclusive forms of public-private partnership.   There is a great need for technical assistance aimed at each of these eight areas of policy change and well-aimed capacity building.  There is a great need to harness information and communications technologies in building virtual clusters of entrepreneurs and other enablers steeped in widening virtuous cycles of creating, managing and adapting business-friendly knowledge and practice, in pursuit of efficiency and inclusiveness in production and distribution. 

What public investment and “sectoral policies” are likely to spur pro-poor patterns of growth?  Research by International Food Policy Research Institute (IFPRI) has shown that the highest payoff in terms of poverty reduction belong to rural roads, followed by agricultural research.  Empowerment of women and augmenting their production and marketing skills is also a powerful avenue for pro-poor growth.  Expanding the outreach of primary and secondary education, especially in rural areas also has high returns in terms of poverty reduction. 

---

[1] Small & Medium Enterprise Sector Development Programme (SMESDP) being implemented by the Ministry of Industries, Government of Bangladesh has partially funded an updating of a list-frame covering SMEs and large enterprises in Bangladesh as of 2005/2006.

[2] TFP is a residual, and calculated as the excess in production over and above what can be attributed to factor accumulation.  It is typically ascribed to the existence of technical progress.