Market Reaction to Industrial Actions in South Africa

Literature › Works

Ngidi, Nondumiso (2011). Market Reaction to Industrial Actions in South Africa. Master of Management in Finance and Investment dissertation, Wits Business School. University of the Witwatersrand.

Type	thesis
Scope	Post-1994 South African economy
Method	econometric
Period studied	2003–2009
Regions	South Africa, Johannesburg Stock Exchange
Themes	unions strikes, labour markets wages, methods data sources, macro fiscal finance debt, transition post apartheid
Citation key	ngidi2011

Argument

The thesis uses an event study of 49 JSE-listed firms hit by strikes in 2003-2009 to test how share prices react to industrial action. It finds negative cumulative abnormal returns beginning before the strike and continuing for several days after it. From this it concludes that the JSE is not fully efficient and that investor sentiment moves prices. A second, probabilistic regression tries to predict strike occurrence from firm size, profitability and the repo rate, and finds little explanatory power. The author attributes strikes to the legacy of apartheid wage discrimination, union leaders' objectives, information asymmetry and the economic climate, but these claims are asserted from the literature rather than tested.

Evidence and method

Department of Labour strike statistics and annual reports (2003-2009) to identify strikes. Share prices and financial data from McGregor BFA, and repo rate data from the SA Reserve Bank. Of 495 firms with strikes, 414 were unlisted and 32 lacked data, leaving 49 firms. Daily abnormal returns are measured against an industry index over a window of -14 to +14 days. Regression outputs are in appendices C-E. Background comes from secondary literature.

Findings

• The sample is small after filtering a large population of strike-affected firms. (p. 50)

  Companies experiencing industrial action/strike between 2003 and 2009 Exclude unlisted companies Exclude companies with incomplete information, for example earnings’ report, strike date and so forth 495 414 32 Total sample 49

• The pre-event CAR turns negative about 11-12 days before the strike and cumulates to roughly -6 to -7% by the end of the window. (p. 66)

  Pre-event CAR changes from 0.0068 in day 12 to -0.0082 in day 11. It can be inferred that the market merely reacts to the news for the ‘looming strike’ prior commencement.

• CARs reach -0.0721 on day 13 and -0.0646 on day 14. These are worked from the day-by-day table; the text also notes some recovery on day 14. (p. 65)

  13 -0.0721 -3.7004 14 -0.0646 4.8223

• The average abnormal return before the event is higher than after, which the author reads as investors adjusting only after the strike announcement. (p. 71)

  Average ARs pre-event are -0.0012 and post-event returns are -0.0022. The average ARs prior to the event are higher than the average returns post the event.

• The mining portfolio dominates the overall result, and its CAR falls much further than the retailers' (about -12% to -14% by days 11-13). (p. 68)

  The mining industry portfolio is bigger in size than the other portfolio; therefore, overall CARs for the sample in Figure 4.1 are substantially influenced by the trajectory of the mining industry portfolio.

• Strikes in SA are mostly short and increasingly about wages. (p. 23)

  in that year 50.8 per cent of the working days lost were related to wages, bonuses and other compensation. The data is on an upward trajectory, as noted in 2009, with 91.2 per cent of stoppages related to wages.

• The probabilistic model finds firm performance, firm size and repo rate individually insignificant in explaining strike occurrence. (p. 74)

  The p-value for company performance, firm size and economic indicator is 0.644933, 0.260886 and 0.97899 respectively at 95% confidence level.

• The author concludes the market reacts about five days before a strike and that the JSE is inefficient, tempered by caveats about sample size and period. (p. 82)

  The findings of the study indicate that the market reacts to the news of a strike action approximately five days prior to commencement, which has been labelled as day 0.

• The author admits the limits of the study: small sample, short period and an unreliable probabilistic model. (p. 83)

  The probabilistic model results are not conclusive as they exhibit the effect of heteroskedasticity. In the study lack of data for one omitted variable caused the model to reveal inconclusive results.

• Background claim that apartheid labour-market control produced a large black-white wage gap. (p. 10)

  Prior to 1994, African and white workers received average wages that differed by a factor of 5, although part of that gap, roughly half according to one assessment