Describe the Brewer et al. treatment correctly

README.md

@@ -36,7 +36,7 @@ Placing imputed entrant earnings at full-time equivalent, as policyengine-uk doe
 
 The participation response is a floor in several ways — whole-year ages exclude the 9-to-12-month cohort, and the response is confined to parents whose youngest child is eligible — but it is also not robust: see the audit findings below.
 
-**The intensive margin is much larger, and overlaps with it.** Parents already in work and paying for childcare see their out-of-pocket cost fall by about 58% under both legs together, and at a childcare price elasticity of hours of −0.042 that is 2.4% more hours: about 30,565 full-time equivalents, £1.67bn of earnings and **£0.67bn** back to the exchequer once the model has taxed the extra earnings and withdrawn benefits on them. That takes the combined dynamic cost to **£5.95bn**. The elasticity was proposed by Max Mosley and is derived from Brewer et al. (+0.600 hours a week over a mean of 14.319, against a 100% price fall); it is a total-hours effect measured over all mothers including those not working, so it already contains the participation channel and the two margins should not be read as additive. The treatment it comes from was 12.5 extra free hours a week, not a price fall of 100%. See `hours_response.py` and `sources.HOURS_PRICE_ELASTICITY`.
+**The intensive margin is much larger, and overlaps with it.** Parents already in work and paying for childcare see their out-of-pocket cost fall by about 58% under both legs together, and at a childcare price elasticity of hours of −0.042 that is 2.4% more hours: about 30,565 full-time equivalents, £1.67bn of earnings and **£0.67bn** back to the exchequer once the model has taxed the extra earnings and withdrawn benefits on them. That takes the combined dynamic cost to **£5.95bn**. The elasticity was proposed by Max Mosley and is derived from Brewer et al. (+0.600 hours a week over a mean of 14.319, against a 100% price fall); it is a total-hours effect measured over all mothers including those not working, so it already contains the participation channel and the two margins should not be read as additive. The treatment it comes from was a move from the part-time offer (12.5-15 hours a week) to full-time school hours (30-35), not a price fall of 100%. See `hours_response.py` and `sources.HOURS_PRICE_ELASTICITY`.
 
 **Distributionally, the average gain broadly rises with income, but the free-hours leg does not.** Among families with a child under 5 in 2027-28, on the central labour supply assumption, the average annual gain runs Q1 £1,066, Q2 £942, Q3 £2,106, Q4 £2,393, Q5 £3,035 — the bottom two quintiles are close, Q2 is the lowest, and the gradient establishes itself from Q3 up. Low-income families gain less in cash because Universal Credit already covers 85% of their childcare costs, which this reform keeps, and because they use fewer paid hours.
 

dashboard/public/data/free_childcare_reform_results.json

@@ -10,11 +10,11 @@
   "policyengine_uk_version": "2.94.0",
   "policyengine_version": "5.3.0",
   "provenance": {
-    "analysis_commit": "8a97cd0152b6fc6e0d7c2b649312ee6e830e5796",
-    "working_tree_dirty": false,
-    "source_digest": "b7e3225e8452812e70fcb80b741ff68c1834d9c7f9bbd4bf0f1afcd5c7458404",
+    "analysis_commit": "51504865a1e7d1fc9e6fc68d577247461ff75cef",
+    "working_tree_dirty": true,
+    "source_digest": "a61b3784e1981e3b3c4f7022e79651285e9e5eb40ed4de2be594aaad90d40179",
     "python_version": "3.13.15",
-    "generated_at": "2026-09-02T12:07:49+00:00"
+    "generated_at": "2026-09-02T15:37:19+00:00"
   },
   "reform_definition": {
     "free_hours": "15 hours a week of free childcare for every child from 9 months to school age, plus a further 15 hours where parents work and earn under \u00a3100,000 \u2014 replacing 30 hours for working parents under \u00a3100,000 and 15 hours for 3-4 year olds.",
@@ -97,7 +97,7 @@
     },
     "brewer_hours": {
       "label": "Brewer, Cattan, Crawford and Rabe \u2014 Does more free childcare help parents work more? (IFS WP20/09)",
-      "description": "England. Table A.3 (file page 41): full-time eligibility raised mothers' usual weekly hours by 0.600 (s.e. 0.264), measured over all mothers with zeros for non-workers, against the Table 1 (file page 17) sample mean of 14.319 hours. The -0.042 hours elasticity used here is that +4.19% taken against a 100% price fall; the treatment was 12.5 extra free hours a week (file page 8), not a 100% price fall. It is a total-hours effect that contains the participation channel, so it overlaps with the extensive-margin result.",
+      "description": "England. Table A.3 (file page 41): full-time eligibility raised mothers' usual weekly hours by 0.600 (s.e. 0.264), measured over all mothers with zeros for non-workers, against the Table 1 (file page 17) sample mean of 14.319 hours. The -0.042 hours elasticity used here is that +4.19% taken against a 100% price fall; the treatment was full-time eligibility (30-35 hours a week) relative to the part-time offer (12.5-15), file pages 8-9, not a 100% price fall. It is a total-hours effect that contains the participation channel, so it overlaps with the extensive-margin result.",
       "url": "https://ifs.org.uk/sites/default/files/output_url_files/WP202009-Does-more-free-childcare-help-parents-work-more.pdf#page=41"
     },
     "akgunduz_plantenga": {

dashboard/src/components/CostTab.jsx

@@ -390,7 +390,8 @@ export default function CostTab({ data, year, bound, intensive, area }) {
                   all mothers, zeros included, so it already contains people moving into work
                   — the paper's own employment effect accounts for most of it. Adding it to
                   the participation response counts that channel twice, and the treatment
-                  was 12.5 extra free hours a week, not a 100% price fall.
+                  was a move from part-time (12.5-15 hours a week) to full-time school hours
+                  (30-35), not a 100% price fall.
                 </li>
               ) : null}
             </ul>

data/free_childcare_reform_results.json

@@ -10,11 +10,11 @@
   "policyengine_uk_version": "2.94.0",
   "policyengine_version": "5.3.0",
   "provenance": {
-    "analysis_commit": "8a97cd0152b6fc6e0d7c2b649312ee6e830e5796",
-    "working_tree_dirty": false,
-    "source_digest": "b7e3225e8452812e70fcb80b741ff68c1834d9c7f9bbd4bf0f1afcd5c7458404",
+    "analysis_commit": "51504865a1e7d1fc9e6fc68d577247461ff75cef",
+    "working_tree_dirty": true,
+    "source_digest": "a61b3784e1981e3b3c4f7022e79651285e9e5eb40ed4de2be594aaad90d40179",
     "python_version": "3.13.15",
-    "generated_at": "2026-09-02T12:07:49+00:00"
+    "generated_at": "2026-09-02T15:37:19+00:00"
   },
   "reform_definition": {
     "free_hours": "15 hours a week of free childcare for every child from 9 months to school age, plus a further 15 hours where parents work and earn under \u00a3100,000 \u2014 replacing 30 hours for working parents under \u00a3100,000 and 15 hours for 3-4 year olds.",
@@ -97,7 +97,7 @@
     },
     "brewer_hours": {
       "label": "Brewer, Cattan, Crawford and Rabe \u2014 Does more free childcare help parents work more? (IFS WP20/09)",
-      "description": "England. Table A.3 (file page 41): full-time eligibility raised mothers' usual weekly hours by 0.600 (s.e. 0.264), measured over all mothers with zeros for non-workers, against the Table 1 (file page 17) sample mean of 14.319 hours. The -0.042 hours elasticity used here is that +4.19% taken against a 100% price fall; the treatment was 12.5 extra free hours a week (file page 8), not a 100% price fall. It is a total-hours effect that contains the participation channel, so it overlaps with the extensive-margin result.",
+      "description": "England. Table A.3 (file page 41): full-time eligibility raised mothers' usual weekly hours by 0.600 (s.e. 0.264), measured over all mothers with zeros for non-workers, against the Table 1 (file page 17) sample mean of 14.319 hours. The -0.042 hours elasticity used here is that +4.19% taken against a 100% price fall; the treatment was full-time eligibility (30-35 hours a week) relative to the part-time offer (12.5-15), file pages 8-9, not a 100% price fall. It is a total-hours effect that contains the participation channel, so it overlaps with the extensive-margin result.",
       "url": "https://ifs.org.uk/sites/default/files/output_url_files/WP202009-Does-more-free-childcare-help-parents-work-more.pdf#page=41"
     },
     "akgunduz_plantenga": {

src/free_childcare_reform/sources.py

@@ -87,9 +87,11 @@ class Source:
 # an intensive effect for those in work. The paper's own +3.5pp employment
 # effect at typical part-time hours accounts for most of it, so this
 # component overlaps with the participation response reported alongside it
-# and the two should not be read as additive. Second, the treatment was 12.5
-# extra free hours a week in term time, not a 100% price fall, so as a
-# per-unit-price elasticity this is a floor (treatment: file page 8).
+# and the two should not be read as additive. Second, the +0.600 is the
+# effect of full-time eligibility (school hours, 30-35 a week) relative to the
+# part-time offer (12.5-15 a week) — a move of roughly 15-20 free hours a
+# week in term time, not a 100% price fall — so as a per-unit-price
+# elasticity this is a floor (treatment: file pages 8-9).
 HOURS_PRICE_ELASTICITY = -0.042
 
 # Scale applied to the OBR participation elasticities for the low and high
@@ -466,8 +468,9 @@ class Source:
     "weekly hours by 0.600 (s.e. 0.264), measured over all mothers with zeros for "
     "non-workers, against the Table 1 (file page 17) sample mean of 14.319 hours. "
     "The -0.042 hours elasticity used here is that +4.19% taken against a 100% "
-    "price fall; the treatment was 12.5 extra free hours a week (file page 8), not "
-    "a 100% price fall. It is a total-hours effect that contains the participation "
+    "price fall; the treatment was full-time eligibility (30-35 hours a week) "
+    "relative to the part-time offer (12.5-15), file pages 8-9, not a 100% price "
+    "fall. It is a total-hours effect that contains the participation "
     "channel, so it overlaps with the extensive-margin result.",
     "https://ifs.org.uk/sites/default/files/output_url_files/WP202009-Does-more-free-childcare-help-parents-work-more.pdf#page=41",
 )