Cover Image
close this bookTurbulence or Orderly Change? Teacher Supply and Demand in South Africa - Current Status, Future Needs and the Impact of HIV/AIDS (CIE, 2000, 36 p.)
View the document(introduction...)
View the documentMulti-Site Teacher Education Research Project (MUSTER)
View the documentList of Acronyms and Abbreviations
View the document1. Overview
View the document2. Background and Introduction
View the document3. The Demographic Characteristics of Teachers
View the document4. Incomes of Teachers and Non-teachers
View the document5. Teacher Turnover - The Dynamics of Teaching Employment
View the document6. Forecasting Basic Numbers
View the document7. Regionality and Micro-Regionality of the HIV/AIDS Epidemic
View the document8. Concluding Remarks
View the documentAnnex

Annex

Table A1 shows in summary form all of the “multivariate” relationships which we have used to double-check more rigorously, and either confirm or deny, trends and differences. This is presented without showing the actual numbers, both to save space and because the numbers individually are not very important. The symbols below should be read as follows: A + implies a positive relationship, so that, for example, being a teacher implies a higher salary. A *** implies that the relationship is statistically significant at the 1/10th of 1% level (an extremely high level of significance), ** at the 1% level, and * at the 5% level. We do not report relationships significant only at the 10% level. NS means that the relationship is “statistically not significant” (that is, more precisely, that the hypothesis that there is no relationship cannot be rejected with a reasonable level of certainty). Thus, for example, the symbols -,*** and -,*** in the columns for “Interaction of time trend and teacher dummy variable” and “Simple time trend for teachers only” and the row for “Hours worked” mean that, assuming the surveys are constructed properly, hours worked per week decreased for teachers and increased for other workers; in other words, there is only a 1 per 1000 chance that we would have measured these relationships as strongly as we did were they not true.

Table A1: “Multivariate” analysis of simple demographic trends, using sampling weights

Equations where the dependent variable is either a quantity (e.g., salary) or the conditional probability of being in a group

Dependent variable

(1) Dummy variable for being a teacher

(2) Simple time trend

(3) Interaction of time trend and teacher dummy variable

(4) Simple time trend for teachers only

Salary

+,***

+,***

+,***

Irrelevant

Salary, controlling for hours worked

+,***

+,***

+,***

Irrelevant

Probability of being African

+,***

+,***

-,***

NS

Probability of being coloured

-,***

+,***

NS

NS

Probability of being Indian

NS

NS

NS

NS

Probability of being white

-,***

-,***

+,***

NS

Probability of being female

+,***

+,***

NS

+,***

Probability of being a union member

+,***

+,*

+,***

+,***

Age

-,***

-,***

+,***

+,***

Hours worked

-,***

+,***

-,***

-,***

Equations where the dependent variable is the conditional probability of being a teacher, as determined by a particular variable

Condition or group

Dummy for the condition or group

Simple time trend

Interaction of time trend and dummy for the condition or group

Union membership

+,***

-,***

+,***

Female

+,***

-,***

-,*

African

+,***

NS

-,***

Coloured

-,***

-,***

NS

Indian

NS

-,***

NS

White

-,***

-,***

+,***

Source: OHS 1995, 1997, and 1999. Author’s tabulation.

Notes: 1) The fourth column in the first panel above arises out of a different equation, in each case, from that in the first three columns. In the fourth column the non-teachers have been filtered from the database and a simple time trend is fit for the relevant quantity or conditional probability. It has to be done this way because the single equation implicit in the first three columns would be over-determined if the database had been filtered for non-teachers. 2) The simple time trends for salaries for teachers only (fourth column) are irrelevant because the salary data are in nominal terms. Thus, in these two rows only the comparisons between teachers and non-teachers are relevant (third column).

Table A2: Basic demographic characteristics of leavers and joiners from the teaching force, as compared to the entire teaching force 1998 to 1999

Characteristic

Leavers as a % of those with same characteristic in 1998 database

Joiners in 1999 as a % of those with same characteristic in 1998 database

Net leavers as a % of those in the 1998 database

Leavers with the given characteristic as a % of all leavers

Joiners with the given characteristic as a % of all joiners

% of those with the given characteristic in the 1998 database

Gender


Male

5.37%

1.79%

3.58%

34.8%

32.4%

34.6%


Female

5.33%

1.97%

3.36%

65.2%

67.6%

65.4%

Population group


African

4.2%

1.6%

2.6%

60.5%

64.4%

76.2%


Coloured

4.8%

2.9%

1.9%

7.5%

12.6%

8.3%


Indian

7.5%

1.4%

6.1%

4.5%

2.3%

3.2%


White

12.0%

3.2%

8.7%

27.5%

20.7%

12.3%

REQV


10

9.1%

1.6%

7.5%

6.5%

4.2%

3.6%


11

5.8%

0.3%

5.5%

4.6%

0.7%

4.0%


12

3.8%

0.2%

3.6%

11.8%

2.5%

15.5%


13

4.6%

1.7%

2.9%

34.8%

46.1%

38.6%


14

6.2%

1.9%

4.2%

31.3%

36.1%

25.8%


15

4.5%

0.8%

3.7%

8.1%

5.3%

9.1%


16

4.3%

0.5%

3.9%

2.5%

1.0%

3.0%


17

6.5%

0.8%

5.7%

0.4%

0.2%

0.3%

Source: calculated by the author from 1998 and 1999 PERSAL database.

Note: we have not presented standard errors in order to minimise overload on the table. However, in general, all of the implicit differences above are statistically significant. For example, the differences between leaving and joining rates for males and females are statistically significant, as is the difference between joining rates at REQV 13 and 14.

Table A3: Assumptions needed to drive a teacher demand and supply projection

Assumption

Importance

Degree of reliability

Discussion of numerical values

Picking out a particular demographic projection

High

Medium, and hides at least another 10-15 assumptions

Uncertainty exists as to whether size of population group of school entry age is already declining after having peaked in 1995 or so. The author believes it is, based on analysis of four data sources.

Picking out orphanhood scenarios

High

Medium

Whether orphans should “drive” teacher demand via a specific caregiver-to-orphan ratio, even if for dialogue purposes, and what that ratio should be, is a matter for discussion. Various orphanhood scenarios can be chosen, from a low of 24.3% of the cohort of 5-9 being orphans by 2015, to a high of 28.7%.

Intake rate into grade 1

Medium

Medium

This has been at least 1.0 for many years. It might conceivably decline. However, it is assumed to be 1.0.

Repetition rate in grade 1

Medium

Low

Repetition in the late 1980s and early 1990s was extremely high-as high as 35%-due to recycling of children admitted to school too young, under the assumption they would repeat. Controlled age of admission can reduce this apparent waste. It is assumed repetition declines to at most 10% (in a high enrolment scenario) or 5% (in a low enrolment scenario).

Grade-to-grade apparent or net flow ratios

High

Medium

These have been relatively low, historically, averaging about 93% (for an apparent loss of 7% between grades). In a high enrolment scenario it is assumed this goes up to 97% on average. In a low enrolment scenario it is assumed it stays at 93% on average. This is not to be confused with a true retention ratio.

Desired class size

High, but policy-driven, not a true assumption

High

This is not so much a matter of empirical analysis as a matter of policy setting, either real policy setting or idealised goals. It is assumed at 38 and 35 for primary and secondary respectively.

Period load (“work effort”) of teachers; proportion periods taught (includes possibility that principals may have a very low teaching load)

Medium, though to some extent policy-driven rather than a true assumption

Low

This is partly a matter of empirical analysis and partly a matter of policy setting. It is assumed constant at 92% and 87% for primary and secondary respectively. This is a fairly good approximation of current reality.

Rate of substitute teacher usage

High, to some degree policy-driven, not a true assumption

Medium

Partly a matter of policy, partly a matter of necessity. It is assumed that in a best case scenario 2% of teachers at any moment need to be substituted. In a worst case it is assumed 5% will need to be substituted.

Usage of a “special” learner: educator ratio for orphans

High, to some degree policy driven

High

Can be set to any level. A 0-1 switch controls whether the ratio is used at all. The switch is 1 in the worst case scenario. If used, a 10 to 1 ratio is set, but can be re-set.

Assumptions about qualifications distribution of teachers

Low for overall balance, high for skills distribution

High

Not explicitly used to drive overall conclusions. Assumed that as many as 25% would remain under-qualified in primary level even by 2015.

Normal attrition rate of teachers

High

High accuracy of measurement in base year, but extremely susceptible to policy measures.

Measured to be 5.5% (depends a bit on qualifications), and staying at that level. Evidently this ratio is a key driver in the projections, and is very responsive to incentives and actual demand for teachers, as noted in the analysis of compensation and entry and exit rates in this document.

Mortality assumptions for teaching force based on HIV/AIDS scenarios

High

Medium

Normal attrition counts normal mortality. This refers to extra mortality from HIV/AIDS. Ranges from 3.5% by 2015 in the best case to 4.6% in the worst case.

Percent of grade 12 who sit for Senior Certificate Exam

Medium

High

This refers to the percentage of learners in grade 12 who opt for entering and sitting for the Senior Certificate Exam. This ratio is measurable with great accuracy at any given moment, but it is highly susceptible to policy and therefore its future variability is inherently difficult to forecast. A focus on increasing the pass rate could encourage schools to discourage learners who are unlikely to pass from sitting for the exam. This ratio has been at somewhere between 90% and 94% in recent years. It is assumed to stay fixed at 92% in any scenario.

Pass rate on Senior Certificate Exam

Low

High

This refers to the well-known “matric pass rate.” It is known with great accuracy in the base year, but it can fluctuate, as was obvious in the results of the year 2000. It is assumed to improve by 3 percentage points in the best scenario and by only 1 percentage point in the worst scenario. The base level is not assumed to be as high as the actual value for 2000, under the cautious supposition that 2000 may have been an exceptional year.

Ratio of headcount enrolment in TTC or tertiary institutions to Senior Certificate passes

High, extremely susceptible to policy

Low

This is a key driver in the system. It is not known with any accuracy whatsoever in the base year, and, on top of this, it is highly susceptible to policy shifts. This ratio drives the enrolment in teacher training institutions. It has been as high as 0.5 in recent memory, when the teacher training institutions were producing bumper crops of teachers, and has sunk to as low as about 0.11 (assuming our best estimates are correct) in 2001. In the best case scenario it is assumed this goes up by 5 percentage points each year. In the worst case it is assumed it decreases to 0.1 and then maintains that level.

Exit or graduation rate from TTC or tertiary institutions

High

Medium

This is a key driver as well. It can be known with reasonable certainty for the base period, based on its value in recent history. It measures, to some degree, the “internal efficiency” of the teacher training institutions, as well as the nominal length of the training programmes. (E.g., with a training programme of four years of nominal length, this ratio would be 0.25.) The ratio is assumed to start at 0.25 and improve to 0.3 in the best case scenario, on the assumption that efficiency can be improved and/or programmes will be shortened. It is assumed to decrease to 0.2 if programmes are long and/or internal efficiency is low.

Percent of those enrolled in TTC or tertiary institutions who are already teachers

High

Medium

Ratio of enrolment that are already teachers and therefore cannot be assumed to drive a replenishing of supply. This has been at approximately 0.45 in recent history. It is intrinsically hard to measure this ratio because the necessary data have not typically been reported in any data documents in the past. Some recent measurements give us some degree of confidence. The ratio is assumed constant at 0.45. Not enough is known about the behaviour of this ratio to justify making any other assumption.

Percent who go on to pursue teaching upon exit from TTC or tertiary institution

High, extremely susceptible to policy

Low

This is a key driver in the system. Its historical value is not known with any certainty because it has never been measured. Furthermore, this ratio is extraordinarily susceptible to policy shifts. It is assumed to have a value of 0.5 in a historical base, to decrease to 0.4 in a worst case scenario, and to improve to 0.9 in a best case scenario.

Base year data on enrolment in all levels of the system

Medium (one might think this is high, but in fact, since in most projections what is interesting is the change over the base, this has only medium importance)

Medium for school level, low for tertiary institutions for 2001

Considerable uncertainty surrounds base year numbers. Discussion of these numbers would take us too far afield. As noted, we are more interested in changes over the base, rather than in absolute numbers per se.

Unit cost data for education programmes in TTCs and tertiary institutions and the effect of economies of scale on such costs

High

Low

This number is not known with much certainty because the cost of places in Colleges of Education in the last few years cannot be taken to be a reliable indicator of true cost, given that they have not operated at maximum efficiency. Costs at tertiary institutions in recent years are somewhat difficult to track because not all institutions report their data. Furthermore, the cost drivers are normally given in terms of Full-Time-Equivalents and other concepts because this is what the funding formula requires, but it is hard to link this up to the headcount concepts that are relevant for projecting teacher supply.


***

The discussion papers are downloadable from the following web address:

http://www.sussex.ac.uk/usie/muster/list.html