Turbulence or Orderly Change? Teacher Supply and Demand in South Africa  Current Status, Future Needs and the Impact of HIV/AIDS (CIE, 2000, 36 p.) 
(introduction...) 
MultiSite Teacher Education Research Project (MUSTER) 
List of Acronyms and Abbreviations 
1. Overview 
2. Background and Introduction 
3. The Demographic Characteristics of Teachers 
4. Incomes of Teachers and Nonteachers 
5. Teacher Turnover  The Dynamics of Teaching Employment 
6. Forecasting Basic Numbers 
7. Regionality and MicroRegionality of the HIV/AIDS Epidemic 
8. Concluding Remarks 
Annex 
The PERSAL data allow a much more finegrained examination of the dynamics of the teaching force. Unfortunately, unlike the OHS data, they do not allow for direct comparisons between the (public) teaching force and the rest of the (public and private) working labour force.
Perhaps the system has indeed stabilised as of 1998, and there was much more turbulence between, say, 1996 and 1998 than between 1998 and 1999, the only period we can ascertain with the PERSAL data at hand. In any case, the picture that emerges from a comparison of the numbers and characteristics of those disappearing from the database between 1998 and 1999 and those appearing in it, hardly suggests a system in turmoil, at least at the macro, national level.
Table 2 shows the leaving and joining rates, plus the sum of the two as an index of turnover (e.g., 2.6% + 0.7% = 3.3%) by province. Note that the national level is not very high, but there is enormous variance between provinces.^{9}
^{9} South Africans may find these rates high. Indeed, some early readers of this document commented on how high these rates seemed. But, as the following quotes from the USA and Canada show, South Africa’s rates of turnover, at least during this period, are low, or at worst on par with some other countries: (1) “Nationally, it is estimated that 30 percent of new teachers leave during their first two years, and more than 40 percent depart during their first four years. Studies also show that teachers who leave the profession reported a lower mean income than those who stayed, challenging the belief that teachers quit to earn more money in other careers.” (Mentoring and Leadership Resource Network, mentors.net/LibraryFiles/OutaHere.html). Notesimilarity of these rates for young teachers between these U.S.wide numbers and those of South Africa. (2) “On a Canadawide basis, 21.7% of teaching staff had left their jobs in the previous 12 months. Of those who left, 38.1% quit voluntarily, 13.3% were fired for poor performance, 11.5% were laid off for reasons such as decreased enrollment or their timelimited contract period ended, and 11.0% took a leave of absence. The remaining 26.1% of staff who left did so for a variety of unstated reasons.” (Child and Family Canada Website, http://www.cfcefc.ca/docs/00001054.htm).
The wealthiest and traditionally bestendowed provinces (Western Cape, Northern Cape, and Gauteng) underwent considerable turnover, whereas in the poorest provinces the turnover was lower. There is a net loss from the sector (e.g. 5.3%  1.9% = 3.4%). The numbers for net loss follow no clearly discernible patterns.
The PERSAL data confirm the OHS results that the teaching force is not just femaledominated, but increasingly so.^{10} More men than women left, and more women than men joined. And, the proportion of women joining is larger than the proportion of women in the database. Annex Table A2 shows the basic demographic characteristics of those who were in the database in 1998, those who apparently left it, and those who joined it.
^{10} With PERSAL we are dealing with administrative records where the sample therefore equals the universe, all of our statements have been assessed for statistical significance, and we discuss only results that are valid with at least a 95% confidence interval.
Table 2: Turnover characteristics of various provinces and national level between 1998 and 1999
Turnover characteristics of various provinces and national level between 1998 and 1999  

Leaving 
Joining 
Turnover ratio 
Net change 
EC 
2.6% 
0.7% 
3.3% 
1.9% 
FS 
6.5% 
1.4% 
7.9% 
5.0% 
GT 
8.6% 
3.9% 
12.5% 
4.6% 
KN 
6.3% 
0.8% 
7.1% 
5.6% 
MP 
3.5% 
3.3% 
6.7% 
0.2% 
NC 
6.8% 
3.2% 
10.0% 
3.6% 
NP 
6.0% 
1.4% 
7.4% 
4.6% 
NW 
2.8% 
0.8% 
3.6% 
2.0% 
WC 
5.8% 
5.5% 
11.3% 
0.3% 
National 
5.3% 
1.9% 
7.3% 
3.4% 
Source: calculated by the author from PERSAL database.Notes. 1. Numbers do not always add up perfectly due to rounding error. 2. The turnover ratio used above is not the standard human resources turnover ratio. We use the sum of the “leaving” ratio and the “joining” ratios, whereas the standard measure is the “leaving” ratio. Given the low correlation between the “leaving” and “joining” ratios in this case, we felt that the sum is more indicative of total movement.
The data by population group are interesting but also puzzling. This data contradicts the OHS data, which suggested that about 20% of all teachers were white, as recently as 1999. However, the PERSAL database suggests that only some 12% of teachers are white. It is quite possible that this is largely due to teachers being employed by school governing bodies (SGBs) and by independent schools. It would require that there be some 30000 white teachers employed by independent schools, SGBs, or in informal situations.
Aside from this issue it is clear that Africans are leaving and joining the teaching force in smaller proportions than their proportion of the teaching force. While some 76% of teachers are African, 61% of those leaving are African, and 64% of those joining are African. In contrast, while some 12% of teachers are white, about 28% of those leaving are white, but, interestingly, some 20% of those joining are also white. This evidently deserves some further examination. It is possible that there is a sort of “churning” of white teachers out of and back into the teaching force. The dynamics for coloured teachers were the most extreme, in that they are joining at rates much greater than their proportion of the teaching force.
Despite these levels of turnover and turbulence, an examination of the PERSAL data shows remarkable predictability in the age distribution of both those joining the database and those leaving it. Figures 3 to 6 illustrate the issues quite aptly. Note that for each dimension concept (age and REQV) there are two graphs: the leaving and joining rates, and the distribution of leavers and joiners. The former gives one a sense of how the “average” teacher of a given age or REQV is reacting and behaving, whereas the latter gives a sense of how that behaviour translates into numbers leaving and joining. For example, we see below that teachers with REQV 14 were leaving at a faster rate than were teachers with REQV 13, as a proportion of the total number of REQV 14 teachers. But, because there are more teachers with REQV 13, the total number of teachers with REQV 13 who were leaving was higher than that of teachers with REQV 14 who were leaving.
Figure 3: Leaving and joining
rates by REQV 1998, 1999
Figure 4: REQV distribution of
leavers and joiners 1998, 1999
Figure 5: Leaving and joining
rates by age, 1998 and 1999
Figure 6: Age distribution of
leavers, joiners and stayers, 1998 and 1999
The rate of leaving (those leaving divided by those in the database) by REQV is what one would expect and corresponds to policy: those with the least education are leaving at the fastest rate. There are two other peaks: at REQV 14 and REQV 17. Interestingly, the pattern for the joining rate reflects the pattern for the leaving rate but naturally at a lower level, given that there was net outflow from the database.
If we look at the REQV distribution of the numbers (not rates) of leavers and joiners, we note a much clearer peak at REQV 13. Again, however, the REQV distribution of those leaving and those joining is extremely similar, suggesting stability in the system. In terms of the tabular analysis, the REQV data are also of some interest. The composition of those joining is much more tightly concentrated around REQVs 13 and 14 than the composition of the base (i.e., 46.1% and 36.1% of the joiners are at REQV 13 and 14, as opposed to 38.6% and 25.8% for the base). This suggests that the teaching force is becoming less diverse in terms of training.
The age distribution of leavers and joiners also suggests stability and order, though with some surprises. Above all, these figures suggest that the notion of teachers leaving the profession during their (presumably most productive) middlelevel years is either mythical or is no longer the case. The pattern of leaving suggests that those who leave are mostly either ready for retirement (note the peaks at exactly 65 and 60) or are very young and therefore simply giving the profession a try or joining while awaiting better prospects, and leaving quickly upon finding them. Note that in terms of the age distribution, the peaks for leaving are 65 and 21; for joining, around 20; and for those who stay, 32. Note how closely the distributions for joining and for leaving resemble each other, except for the oldage peak in the latter. Note that the average age of teachers is about 37.
What is very interesting about both the age and REQV data is that the leavers and joiners are more like each other than they are like those who stay or who were in the database to begin with. For example, those who leave and those who join tend to be more concentrated around REQVs 13 and 14 than those who were in the teaching force at the start. Similarly, those who join and those who leave tend to be very young. The fact that joiners are young is expected. However it might be thought surprising that the young would leave at such fast rates, and that so many of the leavers would be young.
Figures 7 and 8 combine the REQV and age data. Because there are insufficient data points in each combination of REQV and singleyear age groups, we utilised 5year age groups. This is why the data appear smoother, from the xaxis point of view, in the following graphic than in those preceding. Points containing less than 25 leavers have been excluded from the graphic, as a simple statisticalvalidity precaution. That is why there is no curve for REQV17, and it is also why some points are missing from some of the other curves, though this is not easily visible to the naked eye. Note that the two graphics present exactly the same information, but from two different perspectives.
It is clear that teachers with different REQVS respond in extremely similar ways to the pressures (or opportunities) for leaving the profession, with the possible exception of the low leaving rate for REQV12 at the younger end of the age spectrum.
It is also clear that the bulge in those who leave at REQV 14 is concentrated amongst young teachers (the 20 and 25 age groups). Otherwise, in other age groups, leavers are concentrated amongst the lesseducated teachers. This is not surprising, since it is the younger and least educated teachers who seem to have enjoyed the best pay advantage (or least disadvantage) in the mid 1990s.
Figure 7: Leaving rates by age and
REQV, 1998
Figure 8: Leaving rates by age and
REQV, 1998
Having a pay advantage, as noted previously, would explain the high joining rates for younger and lesseducated teachers. This high joining rate for the young is also probably explained by the simple fact that, once a young person is trained as a teacher, a fairly logical first step in the job market is a job as a teacher. However, young teachers also leave in high numbers. One explanation might be that once they have been teaching for a year or two, they begin to understand what their lifetime prospects are and that their relative pay advantage is likely to decline; the more highly educated, the more they leave, because the more their lifetime earning prospects appear poor relative to the rest of the labour force. We can see that there are three groups who really represent very high leaving rates: the besteducated young and then the old in general. But note that amongst the young, the besteducated have higher leaving rates, and among the old, the least educated have the higher leaving rates.
One destination for the leavers is, perhaps, employment with school governing bodies. Figure 9 shows how neatly the ratio of school governing body (SGB) employed teachers matches the rate of leaving from the PERSAL (public) database. Also note that the differences in the ratio of SGB to total teachers by age is statistically highly significant: middleaged teachers are much less likely to be teaching in SGB posts than are the young and the old. Unfortunately, we did not have at our disposal similar data for independent school teachers.^{11}
^{11} Note that for 1998 we did not have as many data points as for 1997, so we display a fitted curve through the 1998 data points, whereas the line through the 1997 data points is not a fitted line.
Figure 9: Ratio of SGB to total
teachers, by age, 1997 and
1998