2.2.2.2. Teorías a favor del proteccionismo
2.2.2.2.2. Teoría de Krugman
The results from this two-way matching methodology are shown in Table 5.14. As before, the difference in post-16 education participation rates between young persons in pilot and control areas is 3.9 percentage points in the unmatched sample. Among the unmatched older siblings there is a higher participation rate among those in control areas compared to pilot areas, which suggests that the area effect may be placing a downwards bias on initial
estimates of the impact of EMA. Taking into account area specific effects without matching would, therefore, lead to an upward revision of estimates of the overall impact of EMA. However, as discussed when explaining the methodology, these unmatched differences do
not provide the best estimates of relevant differences between areas, since they do not take into account differences in demographic composition.
The effect of matching makes the results somewhat less clear-cut. When young people are matched to their own older sibling only, post-16 education participation rates are actually higher in the pilot areas than in the control areas, so that the area effect estimated is positive; this may suggest that original estimates of the overall impact of EMA were too high.
Importantly, however, this difference is not statistically different from zero. Taking into account this area effect, the estimate of the overall effect of EMA is reduced to some 3.4 percentage points amongst this matched sample. However, by extension, this is not
significantly different from the estimated ‘raw effect’ of 5.2 percentage points amongst this matched group. It should also be remembered that this group of young people who have an older sibling is an unrepresentative subgroup of the overall population, so it is possible that the area effects are different for this group than for the population at large.
The results where young persons are matched to any older sibling again show estimated area specific effects that are negative but not significantly different from zero. This again
suggests that initial estimates of the overall impact of EMA may be biased downwards, by an estimated 3.0 percentage points when the 0.7 matching rule is applied, and as much as 6.6 percentage points using the 0.3 rule. Amongst these relatively small matched samples, the estimated ‘raw effect’ of EMA is also somewhat higher than the original estimates of the impact of EMA would suggest. Clearly the impact of EMA on these subgroups (whose characteristics are described in Table A5) is higher than its impact on the population at large. This need not alter the estimates of area specific effects, so long as we assume that these pre- EMA effects are not different for this segment of the population.
Table 5.14 The Impact of EMA on Those with an Older Sibling, Using the Older Siblings Education Decisions to Control for Area Effects
Any sibling: closeness
Unmatched Own
sibling 0.7 0.3
1 Young persons in pilot areas 71.7 (0.9) 68.1 (1.0) 70.4 (1.2) 67.7 (2.0) 2 Young persons in control areas 67.8
(0.7) 63.0 (2.2) 63.3 (1.8) 61.6 (2.8) 3 Estimate of ‘raw’ EMA effect (1–2) 3.9
(1.1) 5.2 () 7.0 (2.1) 6.1 (3.5)
4 Older siblings in pilot areas 43.1 (0.8) 49.5 (1.0) 45.9 (1.9) 41.9 (3.1) 5 Older siblings in control areas 44.6
(1.0) 47.8 (2.4) 48.9 (4.0) 48.5 (4.7) 6 Estimate of area effect (4–5) -1.3
(0.6) (1.4)1.7 -3.0(4.3) -6.6(5.4)
EMA effect allowing for area
effects (1.7)5.2 3.4 (1.7) 10.0 (4.5) 12.7 (6.0) Observations 4,716 2,427 2,605 1,148
Note: Standard errors for the ‘own siblings’ matchings are based on 1,000 bootstraps; those for the 0.7 and 0.3 matching rules are based on 170 bootstraps.
The sibling analysis contains sufficient sample sizes and close enough matches to look at whether EMA has a differential impact on certain subgroups of the population.
Unfortunately, as discussed in Appendix 5.4, it is not possible to use older siblings to look at differential area specific effects between young men and young women at this stage. This is because the data do not currently contain the gender of those older siblings who have already left the home.24 However, urban and rural areas can be analysed separately and the results of this analysis are shown in Table 5.15. The results for urban areas again suggest that the main estimates of the effect of EMA set out in earlier sections of this chapter if anything under- estimate the impact of EMA in urban pilot areas. The possible downward bias on our original estimates range from 6.6 percentage points (if the 0.7 matching rule is used) to 10.4 percentage points using the stricter matching rule. Again, these results are not significantly different from zero.
In rural areas, however, the estimated area effects are positive, although again not
significantly different from zero. This would suggest that our original estimates of the impact of EMA in the rural pilot may be too high, by as much as 8.2 to 9.7 percentage points
amongst these matched samples. This accords well with evidence from Careers Services data presented in previous chapters which showed that participation rates in the rural pilot area were considerably higher than in the rural control areas before EMA was introduced.
Upward revisions to the estimated impact of EMA in urban areas and downward revisions in the rural pilot would bring estimates of the impact of EMA in these two area types closer into line, although the exact degree of convergence would depend on which of the matched samples were used.
Table 5.15 The Impact of EMA on all Eligible Young Persons Using Closest Matched Siblings to Assess Area Effects, by Urban/Rural
Unmatched Closeness
0.7 0.3
Urban
1 Young persons in pilot areas 68.7
(0.8)
67.2 (1.3)
64.7 (2.1)
2 Young persons in control areas 64.5
(1.1)
63.2 (2.1)
60.8 (3.2) 3 Estimate of ‘raw’ EMA effect (1–2) 4.2
(1.3) (2.3)4.0 (3.9)3.9
4 Older siblings in pilot areas 39.5
(0.9)
41.0 (2.1)
37.3 (3.3) 5 Older siblings in control areas 41.5
(1.2)
47.6 (4.7)
47.7 (5.3) 6 Estimate of area effect (4–5) -2.1
(1.5)
-6.6 (5.2)
-10.4 (6.0)
EMA effect allowing for area effects (3-6) 6.3
(2.0) 10.6 (5.3) 14.3 (6.7) Observations 3,627 2,030 916 Rural
1 Young persons in pilot areas 81.8
(1.2)
81.6 (2.2)
79.3 (5.0)
2 Young persons in control areas 73.9
(1.4)
63.1 (3.4)
64.7 (6.1) 3 Estimate of ‘raw’ EMA effect (1–2) 7.9
(1.8) 17.5(4.2) 14.7(8.0)
4 Older siblings in pilot areas 55.3
(1.6)
63.1 (3.8)
59.9 (7.0) 5 Older siblings in control areas 50.9
(1.8)
53.4 (5.2)
51.7 (8.1) 6 Estimate of area effect (4–5) 4.4
(2.4)
9.7
(6.5)
8.2
(10.7)
EMA effect allowing for area effects (3-6) 3.5
(3.0) 7.8 (7.3) 6.4 (12.7) Observations 1,089 575 232