El problema del filólogo como educador

Despite the importance of the true nature of the ‘lost’ state to truthfully reflect the

treatment impact on actual employment (as opposed to as observed by the employ­

ment offices), some conclusions can be drawn as to the treatment effect on unem­ ployment. A large positive effect on compensated unemployment, together with the absence of an effect on (registered) unemployment, would entail a negative effect on uncompensated unemployment, which is likely to be further reinforced by the ob­ served negative effect on ‘lost’ rates (since ‘lost’ individuals who are in reality unem­ ployed cannot be drawing compensation). Focusing on registered unemployment, what programmes seem to affect is thus merely the type o f the unemployment spell experienced, in particular the treatment ‘swaps’ uncompensated unemployment for compensated one. The treated also have a sizeably larger long-term probability of be­ ing on programmes than comparable individuals who have postponed their participa­ tion decision, so that the treated remain significantly more likely to be registered at an unemployment office over time.

To gain some insight as to the treatment effect on the overall labour market situa­ tion of participants, let us assume that, conditional on the X* s, the employment status of the lost individuals is missing at random for those who join and those who wait longer, so that we can impute their employment probability. Let us further abstract from the inactivity state (possibly a rather strong assumption for the lost individuals). These two assumptions allow us to completely ‘attribute’ the observed treatment ef­ fect on lost rates partly to the observed effect on employment and partly on unem­ ployment rates. Joining a programme compared to waiting longer would then ‘swap’ uncompensated, unregistered unemployment not only for compensated unemploy­ ment as well as for further programme participation, but also for employment in the regular labour market. This is visualised in Figure 2.7, based on the 5-year average of the probability of being in various labour market states with and without the treat­ ment.

Figure 2.7 Expected labour market status of individuals in the case of (A) waiting longer in open unemployment and (B) joining a programme, broken down into prob­ ability of being on a programme, in employment or in unemployment, the latter in turn broken down into compensated and uncompensated one.

(Average over the 5-year horizon since the start of the programme)

(A) without the treatment (B) with the treatment

□ programme □ employed □ compensated unemployment □ non-compensated unemployment

To roughly fix ideas, further focus on that treatment effect which after 4-5 months seems more stable and permanent in nature (thus ignoring the initial lock-in effect and the endogenous duration of the treatment). The figures resulting from this series of simplifying assumptions are presented in Table 2.2, where "de facto' (as opposed to ‘observed’) highlights that we are trying to account for the lost spells. Joining a programme would thus make participants on average 3 percentage points more likely to be in employment over time, compared to similar individuals who have postponed their participation decision. However, participants would also be 3 percentage points more likely to be in compensated unemployment and another 3 percentage points more likely to be on programmes over time. Overall, participants would thus have on average a 9% lower probability of being in an unemployment spell not supported by benefits and not registered either.

T ab le 2.2 A crude summary of the effect of participation versus waiting:

The more permanent treatment effect on the probability o f being in the various states, six months after joining a programme (% points)

de fa cto unemployed

compensated " non-compensated ^ on programmes ^ de facto employed *'

+ 3 - 9 + 3 + 3

Notes: Out of the labour force state is ignored. * The more long-term effect is informally gauged from Figures 2.3(A) and (C). ^ Employment probability is imputed for the lost individuals (i.e. assuming missing at random); the more long-term effect is then informally gauged from Figure 2.6. ' As implied by the other calculations.

T re a tm e n t effects by m o n th o f placem ent

Further interesting insights are gleaned when looking at the various outcomes for dif­ ferent sub-groups of the treated based on the time they have spent in unemployment before being placed on a programme. Table 2.3 reveals that for those individuals join­ ing a programme immediately (within their first month) or very late (in their 18**' month) as well as around the time benefits expire (in their 14**' month) the various treatment effects are considerably worse than those for individuals entering a pro­ gramme in intermediate periods (3'^**-6**' months). W hile the differential effect for ‘im­ m ediate’ joiners may be explained by these individuals being possibly rushing the choice o f the appropriate type of programme as well as locking themselves too soon, thus foregoing initial jo b offers, selection issues characterising individuals joining at different months are likely to play a considerable role in these findings (see footnote 28); in particular, for individuals entering a programme exactly at the time of benefit exhaustion, the main incentive may not be in leaving the unemployment system soon, an issue that will be the focus of Section II.7.

T able 2.3 Average treatment effects by month of placement into the programme (averaged over the 5-year horizon since the start of the programme)

Rates/Probabilities

(% points) _{1 to 18}

Placem ent in month:

1 3 6 14 1 8 '

Employment:

- observed 4.8 2.3 4.4 6.4 3.6 3.1

- imputed 2.5 0.3 1.9 4.1 1.6 1.1

- lower; upper bounds -2.0; 8.0 -3.8; 6.2 -2.8; 8.3 0.1; 9.0 -0.6; 5.0 -0.8; 4.2

Deregisteredjempl. -2.7 -1.4 -3.2 -1.5 -5.6 -7.3 Lost -5.8 -4.6 -7.0 -5.5 -4.8 -6.1 Inactivity -1.7 0.9 -0.7 -2.4 -4.4 -4.3 Education 0.0 0.7 0.2 -0.3 -0.3 0.5 On Programmes 7.4 6.6 7.2 7.7 8.2 12.2 Unemployment -4.7 -5.2 -4.0 -6.1 -2.6 -4.9 Benefit receipt 1.9 1.3 1.3 1.0 4.8 3.3

Notes: * averaged over 4.5 years.