Alternative Dataset: The World Bank Enterprise Survey (WBES) of 2016

One of the critical issues that arise as a result of the underlying assumptions of the Hsieh &

Klenow (2009) method is that the variation of the marginal products of inputs may depict (survey) measurement error rather than misallocation. In this section, we use the World Bank Enterprise Survey (WBES) as an alternative dataset to measure misallocation using the same exercise as in the above sections. One reason to use the WBES is that sampling is enormously difficult in Zimbabwe given the poor firm register. Consequently, the WBES data provides us with an alternative sample of firms to test the sensitivity of our results.

The WBES for Zimbabwe shares several similarities with our Matched Employer-Employee data, especially on the sampling process and coverage. The WBES is a World Bank ongoing project that seeks to collect enterprise-level datasets from several middle and low-income countries since 2002. In Zimbabwe, two rounds of survey have so far been collected, in 2011 and 2016. The 2011 WBES only included firms with more than five employees, but the 2016 Zimbabwe WBES includes micro establishments (those with five or fewer employees).

The major difference between our dataset and the WBES is that in our data we have comprehensive information for both the formal and informal manufacturing sector firms. The WBES only includes detailed information for formal sector firms. For informal firms, the WBES lacks crucial information, such as the value of capital, that is vital for the calculation of TFPR. The WBES also includes the service sector while in our survey we only concentrated on the manufacturing sector. For these reasons, we only use the WBES for the analysis of the large, small and micro firms in the formal manufacturing sector.

We, therefore, compute the indicators of misallocation using the WBES and determine if they pose similar results to those we find above. Figure 2.8 shows the distribution of TFPQ and TFPR for micro, small and large firms. In this figure, micro firms are those with five and fewer employees while small and large firms are those with greater than five workers. Consistent with our results above, the WBES shows the presence of many less productive firms in the manufacturing sector in Zimbabwe as shown in Panel (A). Results in panel (B) also show a dispersion in TFPR which according to Hsieh & Klenow (2009) confirms resource misallocation. However, the dispersion is quite small as compared to the results we found in Figure 2.3 above.

44

Figure 2. 8. Distribution of TFPQ and TFPR using ES Data 2016

Notes: Distribution constructed using Word Bank Enterprise Survey 2016

Figure 2.9 provides the relationship between misallocation and productivity. The positive correlation between our misallocation measure (TFPR) and productivity in Figure 2.9 provides robust arguments to the conclusion we find in Figure 2.4 above. These results provide evidence of resource misallocation which stifle the growth of more productive firms.

Figure 2. 9. Distortions vs Productivity using the WBES Data 2016

Notes: Distribution constructed using Word Bank Enterprise Survey 2016 2.4.6.3 An alternative measure of misallocation: The OP Covariance

Olley and Pakes (1996) formulated an aggregate productivity decomposition technique to quantitatively measure resource misallocation. They derived the covariance term (OP

0

.05 .1.15 .2

-20 -15 -10 -5 0 5

Log(TFPQ)

Small and Large Micro

Panel (A) TFPQ Dispersion

0.1.2.3.4

-10 -5 0 5 10

Log(TFPR)

Small and Large Micro

Panel (B) TFPR Dispersion

-5 0510

log_S_TFPRs

-15 -10 -5 0 5

log_S_TFPQ

lpoly smooth Large Micro

kernel = epanechnikov, degree = 0, bandwidth = .97

Panel (A): Enterprise Surveys TFPR vs Productivity

-5 05

log(TFPR)

-15 -10 -5 0 5

Physical Productivity: Log TFPQ

Large Small

Micro

Panel (B): Enterprise Surveys TFPR and Productivity

45

covariance) that depicts the presence of allocative inefficiency. The covariance term depicts whether high productive firms have more than average market shares as compared to less productive firms. The context of their model is that, within an industry, more productive firms demand more resources and hence grow faster and produce more output. By contrast, less productive firms demand less factors of production and should shrink in size as compared to high productivity firms. If this relationship between productivity and size is not realized, then resource misallocation is confessed.

𝐴_𝑡 = ∑ 𝜃_𝑖𝑡

𝐾

𝑘=1

𝐴_𝑖𝑡 = 𝐴̅_𝑡+ ∑(𝜃_𝑖𝑡− 𝜃̅_𝑡)

𝐾

𝑘=1

(𝐴_𝑖𝑡− 𝐴̅_𝑡) = 𝐴̅_𝑡+ ∑ 𝜃̃_𝑖𝑡

𝐾

𝑘=1

𝐴̃_𝑖𝑡

where 𝐴 is a measure of productivity such as labour productivity or TFP and 𝜃 is a measure of firm size, 𝑘 is used to index firms, a bar over a variable is used to indicate the arithmetic mean of that particular variable and a tilde over a variable represent deviations from the mean value.

In the model above, aggregate productivity (which is equal to the weighted average of firm- level productivity, with the firm size used as weights) is decomposed into two components:

unweighted mean of firm-level productivities and sample covariance between firm size and productivity. The second component (covariance term) is key as it exhibits the market selection mechanism. In an efficient market system selection only depends on a market-oriented mechanism such as demand and productivity shocks and input cost. In this regard, less productive firms shrink their activities while more productive firms expand. The covariance is, therefore, an ideal tool to measure allocative inefficiencies. Markets and policy distortions constrain the role of market mechanisms and productivity on market selection and hence influence the magnitude and sign of the OP covariance term.

The OP covariance term is zero if all firms have the same relative size or productivity and/or firm size and performance are uncorrelated. The covariance term is positive (negative) if firms with higher-than-average productivity have also a larger (smaller) size than average firm size and firms with below-average productivity have smaller (larger) than average firm size. A positive (negative) OP covariance term, therefore, signifies a positive (negative) relationship between productivity and firm size. A larger covariance term shows that a higher proportion of economic resources goes towards more productive firms, hence higher aggregate productivity.

Although the underlying assumptions of the OP model are different from the HK model, the intuition of the two models are the same; more productive firms within an industry should

46

demand more production resources, produce more output and grow faster. By contrast, less productive firms shrink in size as they demand less resources and produce less output.

Table 2.5 presents the results for sectorial and the industry OP covariance. The results show positive but small values of the OP covariance term. This suggests that economic resources are not flowing to more productive firms, thus indicating misallocation across sectors and industries. These results corroborate with our previous findings using the Hsieh & Klenow (2009) framework. These findings are generally in line with other results in the literature. For example, Masiyandima & Edwards (2018) found small overall OP covariance of -0.047 in Zimbabwe using the 2011 RPED. They also concluded that the misallocation of resources in Zimbabwe has worsened since 1995. In their study, they showed that the OP covariance ranges from -0.047 to 0.237 points in Zimbabwe, Kenya and Ghana. Although Ghana and Kenya have improved in allocative efficiency of resources between 1993 and 2013, Zimbabwe’s efficiency is on a declining path.

Table 2. 5. Sectorial and Industry OP Covariance

Sector OP Covariance

Formal Sector 0.238

Informal Sector 0.219

Aggregate 0.224

Industry OP Covariance

Metal 0.294

Textile 0.140

Wood 0.233

Notes: Results for the covariance between firm size and firm productivity. We use labour productivity (value-added per worker) as a measure for firm productivity. Firm size is measured by the number of employees. For the formal sector, we consider only small-sized firms (with less than 20 employees).