1. An Index of Structural Imbalance
in European Labour Markets
G.A.Meagher, J. M. Dixon & R.A.Wilson
Skillsnet Technical Workshop
Skillsnet Workshop, November 2016, Thessaloniki
Institute for Employment Research,
University of Warwick, United Kingdom
Centre of Policy Studies (CoPS)
Victoria University, Melbourne, Australia 1

2. Overview
Background:
Previous analysis has demonstrated value of using CGE methods (the MLME) to help the interpretation of the results based on E3ME /WLME
Present focus is on further developing indicators of labour market pressure and structural imbalance

3. Existing E3ME /WLME approach
Forecasts generated using a modular modelling approach containing two major components:
a multi-sector macroeconomic model of 29 European countries (E3ME), primarily developed and operated by Cambridge Econometrics; and
Various labour market extensions (WLME), primarily developed and operated by IER (with inputs from AM and ROA)

4. Current modules – the Warwick Labour Market Extension (WLME)
EDMOD - determines the forecasts of employment by occupation;
QUALMOD - determines provisional forecasts of employment by qualification;
STOCKMOD - determines labour supply by qualification; and
BALMOD - revises the provisional qualifications forecasts to match the labour supply projections from STOCKMOD.

5. The alternative CGE LME (MLME)
The Melbourne CGE LME (MLME) developed by Tony Meagher et al. at COPS, relies:
less on time series extrapolation & econometric analysis;
more on explicitly modelled economic behaviour, based on theoretical considerations.
Introduces a range of behavioural and technical parameters which offer more scope for modelling developments in the labour market (substitution possibilities, etc)
Enables the implications of the pressures and scale of skills imbalances implicit in the WLME to be investigated more explicitly

6. Supply side
Labour by skill (3 broad levels of qualification (ISCED)) can be converted into labour by occupation according to Constant Elasticity of Transformation (CET) functions The position of the transformation curve is determined by the supply of the skill - If the wage rate of occupation 2 increases relative to that of occupation 1, the isorevenue line becomes steeper, and the owners of the skill can increase incomes by offering more of their time to occupation 1 as opposed to occupation 2. Hence, they change the occupational mix from E1 to E2. Each of the 3 “skills” can be transformed into any of the 27 occupations - This enables us to say something about supply to an occupation

7. Skill Transformations between Occupations

8. Demand for occupations
Labour in different occupations can be converted into effective units of industry specific labour according to Constant Elasticity Substitution (CES) production functions. The position of the isoquant is determined by the demand for labour in the industry. If the wage rate of occupation 2 decreases relative to that of occupation 1, the isocost line becomes flatter, and the producers in the industry can reduce their costs by substituting some of occupation 2 for occupation 1. Hence they change the occupational mix from E1 to E2.

9. Substitution between Occupations in Industries

10. Advantages
The MLME can accommodate different scenarios concerning the operation of the occupational labour markets If relative wage rates are fixed, the MLME determines the skill mismatches (expressed in terms of occupations) which pertain at those wage rates If relative wage rates are flexible, the MLME determines the wage rate changes required to clear the labour markets and eliminate any skills mismatches If relative wage rates are sticky, the MLME determines the residual mismatches after the partial wage adjustment has occurred.

11. Contrast to existing approach
Focuses on skills imbalances for occupations rather than qualifications
The WLME effectively puts the onus on qualification mix within occupation changing to match the available supply
The MLME recognises that a greater or lesser supply of skills (as measured by the 3 broad qualification levels) can affect the balance of supply and demand for particular occupations.

12. Data requirements
Employment measured in persons cross classified by industry, occupation and qualification;
Hours worked per person differentiated by industry,
Wage rates differentiated separately by industry, by occupation and by qualification;
Labour supply differentiated by qualification;
Wage rates differentiated by occupation and qualification, are needed for the base year only;
Other data are required for every year of the forecast;
All of this information is now available - Adding extra detail (more skill categories) would require additional data

13. MLME
The

14. Interpretation

15. The augmented MLME

16. Interpretation 2

17. Changing relative wages

18. Supply-Demand Pressures on Wages

19. The index of structural pressures: Table 1 (UK 2020)
Column 1 - excess demand e.g: occupation 2 Legislators and senior officials, excess supply of million hours
Column 2 - % of employment in 2009 (excess supply of %).
Column 3 - share of total employment in 2009 (Legislators and senior officials was 0.30 %).
Column 4 - contribution each occupation to the overall index ((% from column 2 * share from column 3/100) = (37.96 x 0.30) / 100) = 0.11 for Legislators and senior officials (cf. total of 20.07)

20. Developments over time: Table 2 (2009-2020) = Final column of Table 1
Overall Index increases from zero in 2009 to 3.80 in and in 2020
(in aggregate, the structural imbalances increase & in general contributions of each of occupation follow suit);
Occupations of note: 13 Office Clerks; (12.5 %) ; 27 Labourers in mining, construction, etc (11.9 %); and 4 Managers of small enterprises (11.2%);
Office clerks contribute by way of excess demand but the other two contribute by way of excess supply.

21. Employment growth rates by occupation 2009-2020: Table 3 (WLME)
OECD (2016, p.37) suggests: “Skill needs assessment and anticipation exercises (such as labour market forecasts) are usually linked to labour market needs, so they also proxy skill needs by the growth of specific occupations.”
Table 3 shows fastest growth for - 10 Life science and health associate professionals; 12 Other associate professionals; & 27 Labourers in mining, construction, manufacturing and transport.
They contribute significantly to the imbalances index in 2020, but all are projected to experience excess supply
13 Office clerks is projected to show the largest excess demand in 2020 is 1 (36.3 %, Table 1, col. 2) and employment is projected to fall!
This suggest that OECD criterion may be too simplistic

22. Employment by industry and skill: Table 4
The results in Table 1 are derived from two solutions of MLME each year (2009 to 2020). The first determines the changes in technology and relative wage rates required to reproduce the WLME occupational forecasts. The second determines the excess demands and supplies that would arise if relative wage rates were fixed. Employment by industry in 2009 (1st panel, 1st col. Table 4) is effectively subject to a “shock” (i.e., the percentage changes in employment shown in col. 2). This “shock” yields the employment levels shown in col. 3 in Employment by skill in 2009 (2nd panel, 1st col. Table 4) is subject to a corresponding “shocks” (col. 2), yielding the employment levels shown in col. 3 in 2010.

23. Imbalances by occupation: Table 5
The impact effect of the shocks on the markets for labour by occupation generates the set of excess demands and supplies shown in column 3 of Table 5.
Thus, for example, the occupation 13 Office clerks experiences an excess demand of million hours.

24. Adjustments in relative wages- in response to imbalances : Table 6
For 13 Office clerks, the wage rate increases by 11 % (cf. average for all occupations (6%)
In general, if the impact effect of the shocks is to cause an excess demand for (excess supply of) an occupation (Table 5), its wage rate increases by more than (less than) the average (Table 6).
There is one exception, namely, 11 Teaching associate professionals. In that case, a small excess demand leads to a small reduction in the relative wage rate, reflecting the fact that the occupational labour markets do not operate independently of one another.

25. Response of supply and demand: Table 7
For13 Office clerks, increases in relative wages (Table 6) causes Demand to fall ( million hours, Table 5 to million hours, Table 7); and Supply to rise ( million hours, Table 5 to million hours, Table 7) eliminating initial excess demand ( million hours in Table 5 to zero in Table 7); This process is repeated for each year The information on excess demands and supplies in Table 5 provides important information about training needs not available from the information on equilibrium employment levels (i.e., from the standard Cedefop forecasts (Table 7).

26. Office clerks: a case study
13 Office Clerks is forecast to contract sharply (-38% Table 3, col. 4) suggesting this group should be allocated a declining share of VET education & training resources;
But Table 1, col. 4 suggests Office clerks makes a major contribution to the structural imbalance index for the UK in 2020 (2.50 of 20.07, 12½ % of the total)
Moreover, Table 1, col. 2 shows the occupation exhibits a significant tendency towards excess demand (36 %).
Table 1 indicates the allocation suggested by Table 3 (WLME), based on increasing employment levels may significantly underestimate the resources needed for Office Clerks if training resources are to be allocated to avoid skill shortages & surpluses at the existing structure of relative occupational wage rates.

27. Comparing WLME and MME results in detail: Tables 8 & 9
Compare employment of Office clerks by industry & skill categories in Tables 8 (for WLME) and 9 (for MLME):
The two tables match aggregate employment (row 42, first panel of the tables, row 4 in the second), but diverge over time for particular industries & skills;
The sense in which MLME reproduces the WLME forecasts is limited: On the demand side of the market the MLME forecasts reproduce the row & column totals but not the forecasts for the individual elements of the matrices;
Similarly, on the supply side, MLME reproduces the row & column totals of annual (3x27) matrices of employment by skill and occupation, but not the individual elements.

28. Shift-share analysis: Table 10 (Demand)
The change in employment of Office clerks in the WLME forecasts can be decomposed - the first panel of Table 10 concerns the demand side);
In 2009, Office clerks labour demand in industry 1 Agriculture etc was million hours (mhrs);
If this changed at the same rate as total labour demand in the industry, it would have fallen by 3.23 mhrs (shift effect);
It actually fell by 3.84 mhrs to mhrs. - the difference (0.61 mhrs), is the share effect (the reduction in labour demand due to the reduction in its share of total labour demand in the industry Agriculture etc).
The share effect can be attributed to changes in technology and changes in relative occupational wage rates (not explicitly identified in WLME)

29. Shift-share analysis: Table 10 (Supply)
A similar decomposition can be applied to the supply side results from WLME are shown in the 2nd panel of Table 10.
In 2009, supply of Office clerks belonging to the category 1 Low skill was mhrs;
If this changed at the same rate as total supply in that category, it would decline by mhrs (shift effect).
It actually fell by mhrs to mhrs. - the difference ( mhrs) is the share effect (the reduction in employment of Office clerks due to the reduction of its share in the total supply of workers with Low skill).
The share effect can be attributed to changes in preferences for working in particular occupations & changes in relative occupational wage rates (implicit in WLME).

30. Corresponding MLME results: Table 11
Table 11 shows the same kind of shift/share analysis for the MLME forecasts for the employment of Office clerks.
A comparison of column 2 with that in Table 10 reveals that the shift effect is identical. This result follows directly from the definition of the shift effect.
The share effect is also the same in aggregate, BUT it differs for particular industries and skills.
Moreover, in MLME, the contributions to the share effect of changes in technology (demand side) and preferences or “tastes” (supply side) can be identified separately from the contributions of changes on relative wage rates. The former (designated TCH), appear in column 3 and the latter (designated RWR) in column 4.

31. Interpretation 3
In further work the regime of technological and “taste” changes driving the differences between the WLME & MLME forecasts will be explored in further detail.
However, note a limitation on the interpretation of Table 11, while shift-share analysis is useful for ex post analysis:
it cannot be inferred that, if relative wage rates are held fixed, the resultant MLME forecasts can be obtained by adding columns 1, 2 & 3 in Table 11. Entries in column 4 would be reduced to zero, but the entries in columns 2 & 3 would also be affected.
This property of Table 11 means that it cannot be used to decompose the excess demand for 13 Office clerks in Table 1 into its industry and skill components.

32. Assumptions about the role of technological change
The augmented version of equation T1 assumes the rate of occupation-specific technical change is common across industry, similarly, the equations describing the supply of labour assume that the rate of occupation-specific preference change is the same regardless of skill; These assumptions are responsible for the differences between MLME and WLME forecasts in Tables 10 & 11; If these assumptions were made occupation, industry and skill speific, the MLME model could be configured to reproduce WLME employment forecasts exactly; This possibility & experiments with other key assumptions (CES substitution elasticities & the CET transformation elasticities) is the subject of ongoing analysis.

33. Main results and findings
E3ME /WLME results are combined with MLME to reproduce the WLME forecasts & reveal underlying structural pressures (“balance” mean that labour markets clear);
Training agencies focus on the education & training mechanisms that respond to emerging shortages and surpluses aiming to promote a workforce that is equipped with the skills needed in jobs in the future;
Emerging mismatches are usually inferred from analyses of the current situation and are often only qualitative - the measures developed here provide new quantitative insights into future labour market pressures;
The structural imbalance index can be used to compare labour market balance in different countries, & identify the sources of any differences. .

34. Conclusions and policy recommendations
Judgements about allocation of resources for training requires a view about wage adjustment:
the usual approach assumes fixed relative wages - the entire adjustment is borne by training response;
but if markets clear, pay contributes.
Longer term - changes in wages induce a training response;
Policy should aim to ensure wage differentials reflect working conditions (work intensity, social prestige, etc);
But the “Correct” “compensating wage differentials” are unknown – the current pattern is usually “desirability”;
Deviations from existing differentials are often met with complaints of “skills shortage” & demands for more training- The role of wage rate adjustment is generally ignored.

35. Developing alternative projections using the new approach
Forecast employment by industry using E3ME, including associated industry wage rates Extract other relevant data on numbers and pay Determine the corresponding forecast of employment by occupation and qualification using the CGE LME Determine implied average industry wage rates (weighted sum of the occupational wage rates in the CGE LME) Recompute the E3ME forecasts based on the average industry wage rates determined by MLME Iterate until convergence has been achieved Modify workbooks to incorporate CGE LME outputs

36. Institute for Employment Research
Contact Details
Rob Wilson
Institute for Employment Research
University of Warwick
COVENTRY, CV4 7AL
Tel: +(44)