1. Firm Dynamics, Growth and Survival in South Africa
J. Paul Dunne and Rethabile Masenyetse
School of Economics, University of Cape Town
May 2014
TIPS International Conference on Manufacturing Led Growth for Employment and Equality

2. Introduction
In dealing with economic problems in RSA a major concern the changing character of the corporate sector
One important requirement with a healthy industrial structure where a range of sizes of firms provide competition, innovation, employment and sustainable growth.
A tendency towards increased concentration has implications for competitiveness, while a lack of small firms has implications for innovation and employment creation.
So studying the changing size distribution of firms over time is important.

3. Analysing Firm Growth and Size
There is an extensive literature on the theory of firm growth and market structure, reviewed in: Hart (2000), Sutton (1997) and Trau (1996), ...
A useful for analysing the growth of companies is an empirical analysis within the framework of testing Gibrat’s law or Law of Proportionate Effects(LPE) (Dunne and Hughes, 1994 and 1993; Sutton, 1997; Caves, 1998).
This approach assumes that the determinants of firm growth rates are complex and determined by a range of factors and behaviour that make treating growth as a random shock on initial firm size.
It was used in the 1970s to analyse the reasons for an observed inexorable rise in concentration of manufacturing industry, which led to concern that this would continue and lead to increasing monopoly power (Hannah and Kay, 1977).

4. Data
The study uses dataset constructed on South African companies listed on the Johannesburg Stock Exchange (JSE) for the period
The data was collected from Datastream system, Profiles stock exchange handbooks and Macgregor handbooks and online database.
The dataset comprise of company accounts for each year and with the use supporting information documents the reasons for exit from the JSE.
The study uses net sales and employment as measures of size. Sales data has a higher coverage compared to employment giving information on firms each year.

5. Firm Growth and Survival in South Africa
Producing a transition matrix showed:
Out of the 400 companies that were alive in 2005, 288 (72 per cent) survived to 2010.
As expected, the highest survival rate is observed in higher size groups with survival rate of over 90 per cent compared to 54.7 per cent in the lowest size group. The smaller companies are generally more vulnerable.
Of the surviving companies 121 (42 per cent) remained in their size groups and a sizable number moved up to the next size group. A smaller number of companies downgraded to lower size groups.

6. Companies alive in 2005 by Sales Size
Survivors
Sales2010(billions)
<0.1b b
0.5-1b
1-2b
2-3b
3-4b
4-5b
5-10b
>10b
Rbn
Number %
128 70
54.7 39 24 4 2 1
101 72
71.3 5 25 16 33 27
81.8 13 3 18
75.0 6
81.3 17 15
88.2 12 8 23
95.8 14 41 38
92.7
Total
400
288
72.0 46 50 35 9 59

7. Firm Growth and Survival in South Africa
Considering what happened to the 28 per cent that died:
Takeover was main cause of death (13.5 per cent) and varied across the size classes, with the highest proportions in the R1-2billion and R4-5billion groups, at 25 per cent each.
Proportionately takeovers are pronounced in the middle sized firms and not the small as it would be expected. Though in abs no.s
This may indicate that future generation of large firms never reach that stage and possible new competitors are being eliminated.

8. Companies alive in 2005 by Sales Size
Non-Survivors
Type of Death
Takeover
Liquidated
Delisting
Other(1)
Missing(2)
Rbn
Number %
<0.1b
128 35
27.3 14
10.9 2
1.6 10
7.8 7 b
101 29
28.7 18
17.8 5
5.0 1
1.0
0.0
0.5-1b 33 6
18.2
1-2b 24
25.0
2-3b 16 3
18.8
12.5
3-4b 17
11.8
5.9
4-5b 4
5-10b
4.2
>10b 41
7.3
4.9
2.4
Total
400 89
22.3 54
13.5 8
2.0
12.0
3.0
0.5 13

9. Firm Size and Growth in South Africa
A useful way of analysing the changing size distribution is to consider law of proportionate effect (LPE) which implies:
Taking logs gives the estimation equation:
and testing if β=1 tests the LPE.
If β>1 larger firms are growing faster than smaller firms
if β<1 small firms grow faster than larger firms.

10. Firm Size and Growth in South Africa
The results for the aggregate show that:
for the period and the coefficients are significantly less than one thereby rejecting Gibrat’s law.
This means smaller firms are growing relatively faster than large firms.
There is a noticeable difference in magnitude between the two periods.
Consider whether the overall results hold generally
Across size groups, sectors and industry
Whether aggregation bias

11. * reject the null that the coefficient is 1
OLS
ls2005
ls2000
ALL
0.805*
0.906*
SMALL
0.468*
0.582*
MEDIUM
0.832*
1.153*
LARGE
0.958
1.121
PRIMARY SECTOR
0.740*
SECONDARY SECTOR
0.809*
1.045
SERVICES SECTOR
0.759*
0.902*
MANUFACTURING SECTOR
0.811*
0.923
BASIC MATERIAL
0.956
0.726*
INDUSTRIALS
0.753*
1.101
CONSUMER GOODS
0.921*
0.973
CONSUMER SERVICES
0.786*
1.036
TECHNOLOGY
0.788*
0.700*
FINANCIALS
0.721*
0.874*
* reject the null that the coefficient is 1
It is possible that the above result may be subject to aggregation bias arising from size groups, sectors or industries.

12. Disaggregation The results indicate that:
small and medium sized firms to reject the LPE restriction, but not the large firms.
This implies that small firms tend to grow faster, and the smallest firms in the small and medium categories grow faster than the larger ones in those categories.
At the sectoral level, the law is rejected in all sectors except the primary sector.
The law is also rejected for all industries except the basic material industry.
Note the change of the results across the two periods

13. Robustness
There are a number of problems that could reduce any confidence in the results
Heteroscedasticity:
Introduce age of a firm is important in the regression that could lead to heteroscedasticity.
In general, adding age did not change most of our results.
Persistence:
Regressing growth for on growth showed past growth to be significant in all equations except primary sector, consumer goods and consumer services.
For the surviving companies there is evidence of persistence which suggest that our estimates may be biased upwards
But if LPE is rejected this will just make a stronger case for rejection of unitary restriction.
Sample selection:
Heckman sample selection model is one way of dealing with the problem.

14. Employment
That smaller firms tend to grow faster leads to an emphasis on their role in job creation. But
small firms may not be good in creating sustainable employment due to their high failure rate.
A closure of one plant in a major company will require significant growth from sizable number of small firms.
Considering employment reduce coverage as more missing data, but worthwhile.

15. Employment Changes by Size
(Thousands)
(Thousands)
Size Class
Growth in total employment
Total Employment in 2005
Employment in
Surviving firms
Total Employment in 2000
Surviving firms N
Mean
2005
2010
Change
2000
<500 36
0.70 80
0.2 12 14 2 29
0.68 63
0.15 10 17
-0.11 23
0.7 19
0.14 35 24 -8 46
0.26 64
2.5
162
165 3 48
-0.03 79
2.25
178
-16 16
0.17
6.8
157
141 21
0.09 34
7.27
247
-90
>10000 41
0.07 44
33.2
1460
1475 15
-0.16
35.57
1459 -1

16. Firm Growth and Employment
Looking at :
Growth in employment largest for small firms, employing less than 500
But number jobs created higher for med and large
Consider just firms that survived small created 2,000 and largest 15,000
So again small have high growth in employment but lower absolute increase than large.
Note the much poorer performance of all in earlier period
Estimating LPE relation gives rejection of LPE overall, suggesting employment in smaller firms grows faster

17. Firm Growth and Employment
OLS
le2005
le2000
ALL
0.902*
0.842*
SMALL
0.879
0.464*
MEDIUM
1.112
0.983
LARGE
0.908
1.068
PRIMARY SECTOR
0.813*
0.776*
SECONDARY SECTOR
0.937
SERVICES SECTOR
0.966
0.879*
MANUFACTURING SECTOR
0.775*
0.933
* reject the null that the coefficient is 1

18. Firm Growth and Employment
While overall rejects LPE:
For size breakdown smaller reject and larger do not
For sector primary reject but secondary and services do not
For manufacturing reject
So smaller firms grow faster in terms of employment overall, in the primary sector and within the secondary sector only within manufacturing
These results are robust across heteroscedasticity and sample selection

19. Survival Further analysis of survival probit analysis
On size and age
On previous period growth
Evidence that survival to 2010 is positively related to size in 2005
Not apparent if break down by size group?
Only for large firms is there any evidence eof growth in affected survival in

20. Conclusions
The study has analysed the changing size distribution of firms in RSA
Focus on listed companies so relatively larges companies
Within this group smaller firms were less likely to survive as expected
Takeover was found to be the main cause of death for and varied across size groups
The policy issue is whether firms are being taken over to improve efficiency or reduce potential competition.

21. Conclusions
Smaller firms were found to growing relatively faster than larger firms
in the small and medium category it was the very smallest that were growing fastest
In primary not rejected but in secondary and services smaller growing faster
In manufacturing smaller growing faster
In industries only in basic materials is LPE rejected
Robust results –heteroscedasticity and sample selection
Suggests a relatively healthy industrial structure –limited tendency towards concentration generally

22. Conclusions As regards employment 2005-10
smaller firms have higher growth but larger firms have a bigger impact on employment generation.
Smallest firms have lowest coefficient values, but high ses
Primary LPE rejected, but not secondary and tertiary
But in manufacturing LPE rejected
So no evidence of concentration in these
Rather different to sales results and require further investigation (reduced observations).

23. Policy Implications
Small firms play an important role in providing innovation, competition and healthy industrial structure.
But results suggest larger firms that are important for employment generation.
Policies that encourage and support smaller firms need to be introduced, as they are a basis for entrepreneurship and innovation, this should not be to the detriment of large companies.
The high death rate among small firms is also a major concern that needs some policy attention.