At 27.2% in the second quarter of 2018 the official unemployment rate in South Africa ranks as one of the highest in the world. However, depending on whether one uses the official or broad definition of unemployed, since 2008 there are on average between 2 and 3.3 times as many unemployed people as there are people in the informal sector.
This article seeks to explore empirically, using time-series data, the extent to which an increase in the number of unemployed leads to increased entry of workers into the informal sector.
We use a Markov-switching vector error correction model.
We find that such entrance is very limited, lending credence to the notion that significant entry barriers exist into the informal sector.
From a policy point of view these results suggest the need to consider measures that will ease entrance into the informal sector.
At 27.2% in the second quarter of 2018 the official unemployment rate in South Africa ranks as one of the highest in the world. And this rate does not even include those who gave up looking for work, the so-called discouraged work-seekers. Including them yields an unemployment rate of 37.2%. Skills and education levels play a key role in explaining why the unemployed fail to find jobs. For instance, according to the CDE (
However, given that the formal sector typically is more skills intensive, one would expect that those who fail to find a job in the formal sector, would turn to the informal sector. The informal sector might not be a sector of preference, paying much lower remuneration, but presumably it requires significantly lower skills than the formal sector. Nevertheless, at about 17% of total employment, the South African informal sector remains quite small compared to the informal sector in peer-group countries. Depending on whether one uses the official or broad definitions of unemployed, since 2008 there are on average between 2 and 3.3 times as many unemployed people as there are people in the informal sector (Kingdon & Knight
Using the official unemployment rate, there are twice as many people unemployed as are working in the informal sector in South Africa. For the broad definition of unemployment, which includes discouraged work-seekers, this factor rises to 3.3 (StatsSA
According to Kingdon and Knight (
Whereas most of the work on the informal sector and its role with regard to employment or unemployment has been done by labour economists or development economists (using survey data), almost no macroeconomic work has been done in this regard (see Fourie
Addressing this knowledge gap, Burger and Fourie (
Standard macroeconomic theory ascribes longer-run unemployment mostly to product and labour market imperfections. Efficiency wages, labour unions and hysteresis all feature as explanations (see typical textbook explanations in Cahuc and Zylberberg
Augmenting these standard new-Keynesian theories, Bulow and Summers (
As a corollary, in these models people who are still unemployed once they have considered the options in
These models go a long way to consider the presence of both formal and informal sectors, with the formal sector characterised by new-Keynesian features and the informal sector not.
Based on these suggestions, Burger and Fourie (
Unemployment in the three-segment model with involuntary unemployment.
The entire
Note that the informal sector labour demand schedule is horizontal, thus assuming that labour has a constant marginal product. This is not an altogether unrealistic assumption. For instance, according to Berry (
Looking at the whole model, in equilibrium the distance
Suppose a shock to profitability occurs that reduces the effective mark-up; in
In shifting their supply of labour from the primary to the secondary sector, these former formal sector workers have to reveal their reservation wage (given the new-Keynesian characteristics of the primary sector, they were paid more than their reservation wage when they were employed in the primary – formal – sector). The difference between the formal sector wage that they received before becoming unemployed and the wage they are willing to work for in the informal sector is the difference between WP1 and WS1.
Barring entry barriers all those entering the informal sector will get employed. Thus, the number of unemployed people will not be higher than before the shock – the distance between EP1 and ES1 (=
The presence of barriers to entry into the informal sector significantly changes this model outcome. Given such barriers, not all the people who lose their jobs in the primary sector will be able to get employment in the secondary sector – there will only be partial absorption or mop-up. Thus, instead of the labour supply in the informal sector moving from
So how might these barriers to entry work? Grimm et al. (
An alternative explanation for higher unemployment after the shock would be that the reservation wage of some of the workers who became unemployed is higher than the prevailing wage in the secondary sector.
Unemployment in the three-segment model with only voluntary unemployment.
Note that in
Lloyd and Leibbrandt (
In the previous section it is argued that if unemployment increases in the face of an economic shock, barriers to entry into the informal sector would cause the informal sector to fail as a full shock absorber that employs all the unemployed. An increase in the unemployed will therefore, at best, only result in a partial absorption into the informal sector. This section seeks to establish empirically whether an increase in unemployment is followed by an equal or a smaller number of people flowing into the informal sector in South Africa. If only a fraction enter the informal sector, it could be taken as an indication of the presence of entry barriers into the informal sector.
To model the relationship between formal and informal sector employment, as well as unemployment, this section uses a Markov-switching vector error correction (MS-VEC) model. The time-series data used covers the full period for which quarterly employment data is available – 2008Q1 to 2017Q1; these data originate from Statistics South Africa’s Quarterly Labour Force Survey (
Formal employment (’000).
Informal employment (’000).
The official unemployment rate.
Except for the period of the global financial crisis and the accompanying recession, the formal and informal sector employment series both display an upward trend over time. Generally, as the population and labour force grows, more people will be employed. Thus, the formal and informal sector employment series are not expected to be stationary time series. However, the unemployment rate, being a rate of change, is a series one would expect to be stationary over time. The evident non-stationarity of formal and informal sector employment requires the use of a VEC model to ensure that long-run information about the relationship between these two variables is not lost. Thus, the formal and informal sector employment series will enter the long-run component of the model, while the unemployment rate will enter the short-run component of the model. In other words, the long-run component captures the long-run relationship between the levels of formal and informal sector employment, while the unemployment rate influences the short-run changes in formal and informal sector employment.
The long-run relationship is normalised on the informal sector employment variable. We therefore postulate that informal sector employment will adjust to shocks in the relationship between the formal and informal sector employment. (The extent of this adjustment is discussed below.) Furthermore, as the labour force grows and more people are employed over time, one would expect this to reflect in increases in both formal and informal sector employment. Thus, in the long run one would expect a positive relationship between the two. (If employment in the formal and informal sectors expands at the same rate, that would constitute a one-to-one relationship between formal and informal sector employment.)
Usually, assuming responsive wages, if the unemployment rate increases, one would expect it to put downward pressure on the real wage rate, which in turn is expected to lead to higher employment. Thus, in terms of the model below, a higher unemployment rate might be expected to lead to more people entering formal and informal sector employment in later periods. Therefore, we expect a positive relationship between the unemployment rate in period t-1 and the change in both formal and informal sector employment in period t.
Furthermore, although the formal sector is a sector of preference, workers who are unable to find employment in the formal sector would want to enter the informal sector. However, in terms of the theoretical model above, the presence of entry barriers would prevent them from doing so. Moreover, the presence of entry barriers in the informal sector means that the sum of workers subsequently entering the formal and informal sectors would fall short of the initial increase in unemployment (number of unemployed workers). Thus, although we still expect a positive relationship between the unemployment rate in period t-1 and the change in both formal and informal sector employment in period t, the presence of entry barriers in the informal sector implies that the size of its parameter will be too small to ensure that all those becoming unemployed in period t-1 are employed in the informal sector in period t. This is the hypothesis that we will test empirically.
Because the behaviour of economic agents may not remain constant over time, economists use a number of techniques to model such changing behaviour. The Markov-switching technique, developed by Hamilton (
Markov-switching vector error correction model for employment growth.
Similar to the Engle-Granger co-integration method, the model below is estimated in two steps. Using a simple ordinary least squares (OLS) regression, the first step entails estimating the long-run component, as in
Lastly, to allow for the possibility that behaviour might change over up- and downswings of the business cycle, the short-run component is estimated as a Markov-switching model, allowing the constants to take different values depending on whether the model is in Regime 0 or 1. (The short-run component of the model also was estimated with three seasonal dummies to cater for possible seasonal effects.)
Thus, the long-run component of the MS-VEC model:
while the short-run component is:
where:
The data for 2008Q1 to 2017Q1 yield 37 observations. Ideally one would prefer a longer sample, but Juselius and Toro (
The formal and informal sector unemployment series are non-stationary series, while unemployment, after some further investigation, turns out as stationary.
The residual series of the long-run component is also stationary, as can be seen from the KPSS result reported in
The second panel contains the results for the short-run component of the model. It shows what proportion of a shock to the long-run relationship between formal and informal sector employment is corrected (reversed) in one time period (i.e. one quarter). This component was estimated in the second step of the model estimation process. First all the variables specified in
Of interest is the impact of the lagged value of unemployment in both equations. The parameter of 1.294 in the equation for the change in informal sector employment and 0.693 for formal sector employment (second line of second panel), shows that a one percentage point increase in the unemployment rate leads to a 1.294% and 0.693% increase in informal and formal sector employment in the next period.
To answer the question whether or not the unemployed are subsequently absorbed into formal and informal sector employment once they lose their jobs, consider the following example using labour market data for the first quarter of 2017. Given the total number of unemployed of 6.214 million, a one percentage point increase in the unemployment rate would render 224 332 additional people unemployed. Multiplying this number by 1.294% and 0.693% gives us an idea as to how many unemployed people would be absorbed by the informal and formal sectors in the following quarter. In the case of the formal sector it is 78 558, while for the informal sector it is 34 700. On average, about a third of the additional unemployed find work in the formal sector in the following quarter, and about half of that find work in the informal sector. Thus, about half of the number of additional unemployed do not find employment at all in the following quarter. In addition, the impact of the unemployment rate on formal and informal sector employment is limited to the first lag (one quarter).
The results of the empirical analysis above indicate that all the unemployed are not absorbed into the informal sector.
Lastly,
Regimes for the formal and informal sector employment: (a) change in formal sector employment, (b) change in informal sector employment, (c) smoothed probabilities.
South Africa has been suffering from an inordinately high unemployment rate for quite some time. The discussion above highlighted that the number of unemployed people in the country is between 2 and 3.3 times as many as the number of people working in the informal sector. To provide an explanation for this high unemployment rate and the relatively small informal sector this article puts forward a theoretical framework linking employment in both the formal and informal sectors of the economy to involuntary unemployment. The model ascribes the failure of the informal sector to absorb the large number of unemployed to entry barriers in the informal sector.
The empirical part of the article provides corroborating evidence for the existence of entry barriers into the informal sector. It shows that, indeed, once the unemployment rate increases, the ability of both the formal and informal sector to reabsorb the unemployed in the next period is very limited. From a policy point of view these results suggest the need to consider measures that will ease entrance into the informal sector. Access to capital, credit and other financial services, as well as opportunities for owner-operators and workers to improve their skills (notably accounting skills) might serve as top agenda points in this regard. However, there are also other factors that could serve as entry barriers, and therefore also need specific attention. Aspects such as location, premises, facilities and business services (including internet), transport cost, crime, as well as the physical distance between informal firms, their suppliers, and even in some cases their clients, are likely factors. This might require a broader rethink of the role of the informal sector. Typically, policy tends to focus on formalising informal enterprises. Viewed in this way the informal sector is seen as a problem sector, a sector from which enterprises must be assisted (or made) to exit. However, it is also possible to see the informal sector as a sector that can do what the formal sector is unable to do: provide work to low-skilled workers. In such a view, a resilient, buoyant informal sector that can, in the long term, absorb those workers who are unable to find employment in the formal sector will become part of the solution to the unemployment problem. But then the barriers to entry into the informal sector need to come down.
The author declares that they have no financial or personal relationships that may have inappropriately influenced them in writing this article.
Both authors contributed equally to the article.
The Kwiatkowski–Phillips–Schmidt–Shin test reported in
The residual of the long-run component.
Kwiatkowski–Phillips–Schmidt–Shin test (2008Q1–2017Q1).
Level | First Difference | |
---|---|---|
Formal sector employment (log) | 0.617 | 0.175 |
Informal sector employment (log) | 0.532 | 0.348 |
Unemployment rate I | 0.675 | 0.191 |
Unemployment rate II (1999Q3–2017Q1 – biannual observations) | 0.136 | - |
Residual of long-run relationship | 0.127 | - |
Note: Critical value (5%): 0.463 (Null hypothesis: Variable is stationary)
In this article the primary sector is synonymous with the formal sector, while the secondary sector is synonymous with the informal sector. This, of course, need not always be the case, as one could also have two formal sectors, one with new-Keynesian characteristics and the other not. However, in a country such as South Africa with a rather concentrated formal sector and an informal sector defined as comprising small firms not registered for tax purposes, the small-firm nature of the informal sector and the concentrated nature of the formal sector render it possible to portray the primary/secondary sector nature of the model as synonymous with the formal/informal sector nature of the economy.
In Burger and Fourie (
If the labour market was free and not subject to new-Keynesian features, the equilibrium wage would have been equal to the wage in the secondary sector (i.e. equals to WCS) – in fact, the primary and secondary sector distinction would fall away, given the absence of the new-Keynesian features.
It is also used to explore the behaviour of inflation, employment and unemployment (cf. Simon [
All four lags were not included simultaneously to save degrees of freedom and because the unemployment rate series displays a high degree of serial correlation. High serial correlation would lead to multicollinearity and hence, cause the lags in the unemployment rate to turn up as statistically insignificant even if they are not.
The KPSS stationarity test shows that all three variables (formal and informal sector employment and the unemployment rate) are I(1) variables, that is, they are non-stationary (see
For purposes of brevity the first round results are not reported here, but are available from the author on request.
Negative in the equation for the change in informal sector employment and positive (given the positive relationship between formal and informal sector employment in the long run) in the equation for the change in formal sector employment.
As mentioned earlier, the models estimated with earlier lags in the unemployment rate (i.e. second, third and fourth period lags) showed these lags to be statistically insignificant.
The model cannot show whether the unemployed left jobs in the formal sector (the data does not allow for this). Nevertheless, the long-run relationship of the model has been set up with formal sector employment driving informal sector employment (the latter is the dependent variable in the long-run component). Thus, if formal sector employment drops one percentage point, it correlates almost one-to-one with a one percentage point drop in informal sector employment.