Figures and tables in LifeSim Childhood: Extrapolating Intervention Effects and Public Cost Savings from Birth to Adolescence in the UK

Abstract

Economic evaluation of early childhood interventions is challenging because it is hard to extrapolate the full range of long-term benefits and public cost savings from short-term effectiveness evidence. One way to address this issue is through the use of microsimulation. This paper introduces a closed birth cohort childhood microsimulation model, LifeSim Childhood, that is capable of extrapolating the long-term effects of various childhood circumstances on a broad range of health, educational, and social outcomes and public cost savings up to age 17 in the UK. We use regression-based long-jump microsimulation that simulates outcomes at each age directly from total-effect parameters estimated using bespoke longitudinal birth-cohort regressions, without tracing indirect mediating pathways. The aims of this paper are to describe the general modelling approach and methods underpinning LifeSim Childhood, to present a simple illustration of how the model can be used to simulate the effects of reducing early childhood poverty, and to compare our illustrative results with estimates from quasi-experimental studies. We use data from the Millennium Cohort Study which follows children born in the UK around the year 2000. We focus on exposure–outcome relationships that can be plausibly interpreted as causal based on interdisciplinary theory and evidence, and estimate total effects conditional on an explicitly justified set of confounding variables. As well as describing our data inputs, regression analysis methods and simulation methods, we apply the model to four hypothetical income-shifting scenarios in early childhood and estimate the general magnitude of long-term public cost savings and wellbeing benefits, alongside a battery of more specific outcomes.

This article contains 9 figures and 16 tables.

Figures

**Figure 1.** Structure of LifeSim Childhood. Ongoing work will extend this to include risk factors up to age 14. The current list of risk factors includes early years household income, having a teenage mother, low birth weight, disability, delayed school readiness, etc. Outcomes are estimated from the age of the risk factor up to age 17, and will be extended to age 23 when the new sweep of the MCS is available. Our current outcomes include cognitive ability, socio-emotional scores, hospitalisation, exclusion from school, etc. The important adverse outcomes at age 17 we currently focus on are poor GCSE performance, psychological distress, obesity, smoking and poor health.

**Figure 2.** Simplified directed acyclic diagram (DAG) for the effect of early childhood income. The preferred model controls for everything in the “Confounders” list. We also run an “extra conservative” model which controls for additional variables about which there is room for disagreement because they are partly mediators as well as partly confounders - we call these “partly confounding mediators”. These additional control variables in the extra conservative model are: IMD, maternal MH, HH employment, grandparent’s education, and child’s disability. HH education is highest education in the household, HH disability is an indicator for any parental disability in the household, IMD is the index of multiple deprivation, maternal MH is Maternal malaise measured at 9 months, HH employment is an indicator for labour force participation in the household and the child’s disability is observed at age 3.

**Figure 3.** Percentage reduction in adverse outcomes with each intervention. Income increase scenarios - Scenario 1 - Poorest to second poorest, Scenario 2 - Poorest two to middle, Scenario 3 - Poorest to richest, Scenario 4 - All to richest. Outcomes are measured at MCS sweep 7 when respondents are around age 17. Poor GCSEs also includes N5 results for Scotland, Psychological distress is identified using the Kessler scale, Obesity is based on UK90 thresholds for sex and age, Regular smoking is more than 6 cigarettes a week.

**Figure 4.** Estimated gain in annual average WELLBYs per child in the general population cohort. Income increase scenarios - Scenario 1 - Poorest to second poorest, Scenario 2 - Poorest two to middle, Scenario 3 - Poorest to richest, Scenario 4 - All to richest. Life satisfaction is based on a simple linear transformation of parent reported SDQ emotional symptoms score.

**Figure 5.** Baseline annual cost per child in each MCS sweep broken down by source. Truancy, exclusion and special education needs are not measured at age 17 so age 14 values are carried over. Conduct disorder does not include justice costs and the unit cost is lower from age 11.

**Figure 6.** Annual cost savings per child for each scenario in each MCS sweep broken down by source. The bar plots clustered at each age represent scenarios 1, 2, 3, and 4. Income increase scenarios - Scenario 1 - Poorest to second poorest, Scenario 2 - Poorest two to middle, Scenario 3 - Poorest to richest, Scenario 4 - All to richest. Truancy, exclusion and special education needs are not measured at age 17 so age 14 values are carried over. Conduct disorder does not include justice costs and the unit cost is lower from age 11.

**Figure A1.** Density plot of early years OECD equivalised weekly household income in the MCS split into quintiles. The means for each quintile or fifth of the distribution are £123.47, £207.40, £304.54, £419.29 and £659.05 respectively.

**Figure A2.** Baseline prevalence of adverse outcomes at age 17. Outcomes are measured at MCS sweep 7 when respondents are around age 17. In Scotland, Poor GCSEs refers to N5 exam results, psychological distress is identified using the Kessler scale, obesity is based on UK90 thresholds for sex and age, regular smoking is more than 6 cigarettes a week.

**Figure A3.** Baseline life satisfaction proxied by SDQ emotional symptoms score in each MCS sweep. Life satisfaction is based on a simple linear transformation of parent reported SDQ emotional symptoms scale.

Tables

Table 1. Simulated effects of scenarios for a UK birth cohort of 700,000 children

Outcome		Scenario 1	Scenario 2	Scenario 3	Scenario 4
Adverse outcomes (Number of cases prevented by age 17)
	Poor GCSEs	10,909	36,366	34,831	76,999
	Psychological distress	697	5,576	7,685	22,916
	Obesity	-785	4,066	8,875	31,353
	Regular smoker	3,411	12,410	11,251	22,638
	Poor health	3,937	10,344	10,897	24,396
Public cost savings (Cost savings between ages 3 and 17, in millions of 2023 £s)
	Total savings	197	1,017	1,012	2,298
By source	Any hospitalisation	-4	-3	30	127
	Disability	7	46	57	155
	Conduct disorder	26	78	84	178
	Special education needs	107	771	711	1,621
	Persistent truancy	1	6	5	13
	Permanent exclusion	61	118	124	205
Wellbeing improvement (WELLBYs gained between ages 3 and 17)
	WELLBYs	282,840	733,257	918,563	2,249,815
	Value (in millions of 2023 £s)	3,677	9,532	11,941	29,248
Scenario cost (Cost to increase income to quintile minimum, in millions of 2023 £s)
	Cost estimate	-2,134	-9,717	-20,240	-52,020

The umber of births in the UK in 2021 was around 700,000. Costs and wellbeing values are discounted at 3.5% per year calculated at birth. Tables with baseline levels and unadjusted figures can be found in the appendix. Income increase scenarios - Scenario 1 - Poorest to second poorest, Scenario 2 - Poorest two to middle, Scenario 3 - Poorest to richest, Scenario 4 - All to richest. Negative values in the table are increases in cases or public costs. Poor GCSEs also includes N5 results for Scotland, Psychological distress is identified using the Kessler scale, Obesity is based on UK90 thresholds for sex and age, Regular smoking is more than 6 cigarettes a week. The program cost is based on minimum the amount that would need to be transferred to each household over five years to move them across fifths of the income distribution.

Table 2. Simulated effects from three models of increasing the income of families with children in the poorest fifth to that of the second poorest fifth for a UK birth cohort of 700,000 children

Outcome		Preferred Model	Extra Conservative	Unadjusted
Adverse outcomes (Number of cases prevented by age 17)
	Poor GCSEs	10,909	5,965	15,068
	Psychological distress	697	-152	-298
	Obesity	-785	-1,013	2,740
	Regular smoking	3,411	1,527	5,837
	Poor Health	3,937	2,499	5,254
Public cost savings (Cost savings between ages 3 and 17, in millions of 2023 £s)
	Total savings	197	-155	350
By source	Hospitalisation	-4	-17	13
	Disability	7	-7	10
	Conduct disorder	26	15	43
	Special education needs	107	-179	212
	Truancy	1	1	2
	Exclusion	61	32	68
Wellbeing improvement (WELLBYs gained between ages 3 and 17)
	WELLBYs	282,840	180,847	435,380
	Value (in millions of £s)	3,677	2,351	5,660

Number of births in the UK in 2021 was around 700,000. Costs and wellbeing values are discounted at 3.5% per year calculated at birth. Tables with baseline levels and unadjusted figures can be found in the appendix. All results are the estimated impact of income increase shift scenario 1 - Poorest to second poorest. Negative values in the table are increases in cases or public costs. Poor GCSEs also includes N5 results for Scotland, Psychological distress is identified using the Kessler scale, Obesity is based on UK90 thresholds for sex and age, Regular smoking is more than 6 cigarettes a week. The additional control variables in the extra conservative model are: IMD, maternal mental health, household employment, grandparent’s education, and child’s disability.

Table 3. Effects of increasing income by $1000 (£676) in Lifesim compared to estimates from Cooper and Stewart (2021)

	Cooper and Stewart (2021)			LifeSim Childhood
Paper	Outcome	Age	Effect	Outcome	Age	Effect
Blau (1999) (All families in US)
	PIAT Maths	5+	0.01	NFER Progress in maths	7	0.02
	PIAT Reading	5+	0.01	BAS Word reading	7	0.02
	PPVT	3+	0.01	BAS Naming vocabulary	3-5	0.01
Votruba-Drzal (2006) (All families in US)
	PIAT Maths	5+	0.02	NFER Progress in maths	7	0.02
	PIAT Reading	5+	0.02	BAS Word reading	7	0.02
Fernald et al. (2008) (Poor households in Mexico)
	PPVT	4-6	0.21	BAS Naming vocabulary	3-5	0.03
Milligan and Stabile (2011) (Low education households in Canada)
	PIAT Maths	6-10	0.07	NFER Progress in maths	7	0.03
	Hyperactivity	4-10	0.07	SDQ Hyperactivity	5-11	0.0
	Conduct disorder	4-10	0.10	SDQ Conduct problems	5-11	0.03
Dahl and Lochner (2012) (Poor families in US)
	Maths	8-14	0.21	NFER Progress in maths	7	0.03
	Reading	8-14	0.21	BAS Word reading	7	0.03
Gennetian and Miller (2002) (Poor families in Minnesota)
	BPI Internalising	5-13	0.12	SDQ Internalising	5-14	0.03
	BPI Externalising	5-13	0.11	SDQ Externalising	5-14	0.02
Dearing et al. (2006) (Poor households in US)
	CBCL Internalising	2-5	0.02	SDQ Internalising	3-5	0.03
	CBCL Externalising	2-5	0.03	SDQ Externalising	3-5	0.01
Zachrisson and Dearing (2015) (All families in Norway)
	CBCL Internalising	2-3	0.02	SDQ Internalising	3	0.01

Effect sizes are based on standard deviation increases in outcome for a $1000 increase in income. The most similar available test, based on concept measured and age at measurement, in the MCS is used for comparison with those presented in Cooper and Stewart (2021). The age range for the LifeSim estimates are based on the sweep age ratherthan actual age at time of testing. All MCS measures, except the SDQ scores, are age and ability adjusted measures that are standardised within sample. PIAT - Peabody individual achievement test, NFER - National foundation for educational research, PPVT - Peabody picture vocabulary test, BAS - British ability scales, SDQ - Strength and difficulties questionnaire, BPI - Basic personality inventory, and CBCL - Child behaviour checklist. Ages in MCS correspond to MCS sweeps and not age at observation (age 3 - sweep 2, age 5 - sweep 3 and so on until age 17 - sweep 7).

Table A1. Summary of unit costs currently used in LifeSim Childhood

		Public Cost Sector
Simulated Outcomes	Ages Available in MCS	NHS	Social	Education	Criminal Justice	Outcome Costed	Annual Unit Cost(2023 £)
Hospital Admission	3, 5, 7, 11, 14, 17	X				Inpatient visit for children	1,587
Disability	3, 5, 7, 11, 14, 17	X				Per child cost to NHS	1,008
Conduct Disorder	3, 5, 7, 11, 14, 17	X	X	X		Conduct disorder	3,092 (5 to 10) 1,963 (11 to 16) 236 (17+)
Special Education Needs	7, 11, 14			X		EHC Plan	25,500
Any Truancy	11, 14			X		Persistent Truancy	943
Any Exclusion	11, 14			X		Alternate Provision	21,848

Ages in MCS correspond to MCS sweeps and not age at observation (age 3 - sweep 2, age 5 - sweep 3 and so on until age 17 - sweep 7).

Table A2. Sources of unit costs currently used by LifeSim childhood

Simulated Outcome	MCS Source	Cost Outcome	Ages costed	Cost year	Cost source
Hospital Admission	Parent admission	Hospital	0-18	2020	Average cost of a single inpatient hospitalisation for a child aged between 0 and 17 in the UK in 2003/2004 at 2019/2020 prices. (Dale et al., 2024)
Disability	Parent	Healthcare		2016	average annual cost of healthcare for a child in the UK. (Kelly et al., 2018)
Conduct disorder	Parent	Conduct disorder	5 to 18	2008	Costs from 4 papers published between 2000 and 2007. (Bonin et al., 2011)
Special Educational Needs	Teacher	EHC Plan	5 to 18	2023	Average incremental cost to DfE of student with EHC plan in 2023 (Acton et al., 2024)
Persistent truancy	CM	Persistent Truancy	11 to 12	2005	Cost education welfare services in England and number of persistent truants in 2005. (Brookes et al., 2007)*
Permanent School Exclusion	Parent	Alternative Provision	5 to 18	2018	Survey of 101 LAs by ISOS for the Department for Education in FY 2017-2018. (Bryant et al., 2018)

also source for costs in the Greater Manchester unit cost database.

Table A3. Simulation population descriptive statistics for confounders and exposure

	Characteristic	Mean/Proportion	Standard Error
Sex	Male	0.503
Ethnicity	White	0.860
	Mixed	0.032
	Indian	0.018
	Pakistani or Bangladeshi	0.044
	Black	0.027
	Other	0.013
Country	England	0.826
	Wales	0.047
	Scotland	0.091
	Northern Ireland	0.037
Region	North East	0.037
	North West	0.103
	Yorkshire and the Humber	0.089
	East Midlands	0.071
	West Midlands	0.081
	East of England	0.088
	London	0.130
	South East	0.148
	South West	0.078
Household Highest Education	No qualifications	0.119
	NVQ level 1	0.081
	NVQ level 2	0.340
	NVQ level 3	0.098
	NVQ level 4	0.293
	NVQ level 5	0.037
Household Characteristics	Disability in household	0.029
	Single parent	0.146
	Age of mother (at birth)*	28.723	0.041
	Annual early years income*	18226.072	74.836

mean and standard error, all other numbers are proportion of population in category.

Table A4. Simulated baseline outcome means

Outcome	Age 3	Age 5	Age 7	Age 11	Age 14	Age 17
Poor GCSEs						0.38 (0.045)
Psychological distress						0.25 (0.026)
Obesity						0.26 (0.046)
Regular smoker						0.13 (0.044)
Poor health						0.10 (0.036)
Life satisfaction	7.89 (0.125)	7.90 (0.125)	7.73 (0.146)	7.41 (0.166)	7.23 (0.173)	7.14 (0.181)
Any hospitalisation	0.22 (0.034)	0.16 (0.033)	0.14 (0.032)	0.15 (0.031)	0.15 (0.041)	0.19 (0.043)
Disability	0.05 (0.022)	0.07 (0.024)	0.08 (0.026)	0.11 (0.033)	0.14 (0.033)	0.16 (0.046)
Conduct disorder		0.09 (0.005)	0.09 (0.005)	0.09 (0.006)	0.09 (0.007)	0.08 (0.006)
Special education needs			0.05 (0.030)	0.06 (0.028)	0.06 (0.030)
Any truancy				0.04 (0.031)	0.05 (0.025)
Any exclusion				0.06 (0.066)	0.09 (0.043)

Terms in brackets are the standard deviation of the distribution of means across simulations. Ages correspond to MCS sweeps and not age at observation (age 3 - sweep 2, age 5 - sweep 3 and so on until age 17 - sweep 7).

Table A5. Simulated baseline cost means by source

Outcome	Age 3	Age 5	Age 7	Age 11	Age 14	Age 17
Any hospitalisation	349.49 (54.33)	246.68 (51.84)	215.45 (51.02)	243.71 (49.26)	237.38 (65.29)	308.90 (67.54)
Disability	46.67 (21.74)	73.89 (24.25)	84.12 (26.00)	114.43 (33.64)	138.30 (33.13)	161.57 (46.54)
Conduct disorder		283.86 (16.04)	273.90 (16.19)	209.21 (13.89)	175.65 (13.49)	67.39 (4.48)
Special education needs			1221.48 (758.85)	1583.28 (697.62)	1630.01 (770.40)	1630.01 (770.40)
Permanent exclusion				137.77 (148.40)	154.34 (73.18)	154.34 (73.18)
Persistent truancy				7.79 (5.37)	8.91 (4.38)	8.91 (4.38)

Costs for special education needs, permanent exclusion, and persistent truancy are carried over from age 14 to age 17 as those outcomes are not measured at age 17. Terms in brackets are the standard deviation of the distribution of means across simulations. Ages correspond to MCS sweeps and not age at observation (age 3 - sweep 2, age 5 - sweep 3 and so on until age 17 - sweep 7).

Table A6. Simulated undiscounted effects of scenarios for a UK birth cohort of 700,000 children

Outcome		Scenario 1	Scenario 2	Scenario 3	Scenario 4
Adverse outcomes (Number of cases prevented by age 17)
	Poor GCSEs	10,909	36,366	34,831	76,999
	Psychological distress	697	5,576	7,685	22,916
	Obesity	-785	4,066	8,875	31,353
	Regular smoker	3,411	12,410	11,251	22,638
	Poor health	3,937	10,344	10,897	24,396
Public cost savings (Cost savings between ages 3 and 17, in millions of 2023 £s)
	Total savings	296	1,574	1,512	3,394
By source	Any hospitalisation	−5	−3	44	181
	Disability	11	69	82	217
	Conduct disorder	36	109	117	248
	Special education needs	155	1,197	1,062	2,404
	Persistent truancy	2	10	9	21
	Permanent exclusion	97	192	198	324
Wellbeing improvement (WELLBYs gained between ages 3 and 17)
	WELLBYs	398,629	1,042,533	1,326,571	3,293,455
	Value (in millions of 2023 £s)	5,182	13,553	17,245	42,815
Scenario cost (Cost to increase income to quintile minimum, in millions of 2023 £s)
	Cost estimate	−2,364	−10,760	−22,414	−57,607

Number of births in the UK in 2021 was around 700,000. Tables with baseline levels and unadjusted figures can be found in the online appendix. Income increase scenarios - Scenario 1 - Poorest to second poorest, Scenario 2 - Poorest two to middle, Scenario 3 - Poorest to richest, Scenario 4 - All to richest Negative values in the table are increases in cases or public costs. Poor GCSEs also includes N5 results for Scotland, Psychological distress is identified using the Kessler scale, Obesity is based on UK90 thresholds for sex and age, Regular smoking is more than 6 cigarettes a week.

Table A7. Simulated effects of scenarios for a UK birth cohort of 700,000 children with SDQ internalising as the wellbeing measure

Outcome		Scenario 1	Scenario 2	Scenario 3	Scenario 4
Adverse outcomes (Number of cases prevented by age 17)
	Poor GCSEs	10,909	36,366	34,831	76,999
	Psychological distress	697	5,576	7,685	22,916
	Obesity	−785	4,066	8,875	31,353
	Regular smoker	3,411	12,410	11,251	22,638
	Poor health	3,937	10,344	10,897	24,396
Public cost savings (Cost savings between ages 3 and 17, in millions of 2023 £s)
	Total savings	197	1,017	1,012	2,298
By source	Any hospitalisation	−4	−3	30	127
	Disability	7	46	57	155
	Conduct disorder	26	78	84	178
	Special education needs	107	771	711	1,621
	Persistent truancy	1	6	5	13
	Permanent exclusion	61	118	124	205
Wellbeing improvement (WELLBYs gained between ages 3 and 17)
	WELLBYs	197,879	629,453	678,858	1,656,902
	Value (in millions of 2023 £s)	2,572	8,183	8,825	21,540
Scenario cost (Cost to increase income to quintile minimum, in millions of 2023 £s)
	Cost estimate	−2,134	−9,717	−20,240	−52,020

Number of births in the UK in 2021 was around 700,000. Costs and wellbeing values are discounted at 3.5% per year calculated at birth. Tables with baseline levels and unadjusted figures can be found in the online appendix. Income increase scenarios - Scenario 1 - Poorest to second poorest, Scenario 2 - Poorest two to middle, Scenario 3 - Poorest to richest, Scenario 4 - All to richest Negative values in the table are increases in cases or public costs. Poor GCSEs also includes N5 results for Scotland, Psychological distress is identified using the Kessler scale, Obesity is based on UK90 thresholds for sex and age, Regular smoking is more than 6 cigarettes a week.

Table A8. Simulated effect size of scenarios per child in birth cohort

Outcome		Scenario 1	Scenario 2	Scenario 3	Scenario 4
Adverse outcomes (Reduction in probability by age 17)
	Poor GCSEs	0.02	0.05	0.05	0.11
	Psychological distress	0.0	0.01	0.01	0.03
	Obesity	0.0	0.01	0.01	0.04
	Regular smoker	0.0	0.02	0.02	0.03
	Poor health	0.01	0.01	0.02	0.03
Public cost savings (Cost savings between ages 3 and 17, in 2023 £s)
	Total savings	281	1,452	1,445	3,283
By source	Any hospitalisation	−6	−4	43	181
	Disability	10	65	81	221
	Conduct disorder	36	111	120	254
	Special education needs	153	1,102	1,015	2,316
	Persistent truancy	2	9	8	19
	Permanent exclusion	86	169	178	292
Wellbeing improvement (WELLBYs gained between ages 3 and 17)
	WELLBYs	0.4	1.05	1.31	3.21
	Value (in 2023 £s)	5,253	13,618	17,059	41,782

Costs and wellbeing values are discounted at 3.5% per year calculated at birth. Tables with baseline levels and unadjusted figures can be found in the online appendix. Income increase scenarios - Scenario 1 - Poorest to second poorest, Scenario 2 - Poorest two to middle, Scenario 3 - Poorest to richest, Scenario 4 - All to richest Negative values in the table are increases in cases or public costs. Poor GCSEs also includes N5 results for Scotland, Psychological distress is identified using the Kessler scale, Obesity is based on UK90 thresholds for sex and age, Regular smoking is more than 6 cigarettes a week.

Table A9. Simulated effect size of scenarios per recipient in birth cohort

Outcome		Scenario 1	Scenario 2	Scenario 3	Scenario 4
Adverse outcomes (Reduction in probability by age 17)
	Poor GCSEs	0.08	0.13	0.26	0.14
	Psychological distress	0.01	0.02	0.06	0.04
	Obesity	−0.01	0.02	0.07	0.06
	Regular smoker	0.03	0.05	0.08	0.04
	Poor health	0.03	0.04	0.08	0.04
Public cost savings (Cost savings between ages 3 and 17, in 2023 £s)
	Total savings	1,482	3,762	7,618	4,128
By source	Any hospitalisation	−33	−10	229	227
	Disability	53	169	429	278
	Conduct disorder	192	288	633	320
	Special education needs	806	2,854	5,350	2,912
	Persistent truancy	8	22	42	24
	Permanent exclusion	455	438	936	367
Wellbeing improvement (WELLBYs gained between ages 3 and 17)
	WELLBYs	2.13	2.71	6.92	4.04
	Value (in 2023 £s)	27,682	35,274	89,903	52,537

Table A10. Effects of increasing income by $1000 (£676) in Cooper and Stewart (2021) compared to three LifeSim models

Cooper and Stewart (2021)			LifeSim Childhood
Paper	Outcome	Effect	Outcome	Preferred Model	Extra Conservative	Unadjusted
Blau (1999) (All families in US)
	PIAT Maths	0.01	NFER Progress in maths	0.02	0.01	0.02
	PIAT Reading	0.01	BAS Word reading	0.02	0.01	0.03
	PPVT	0.01	BAS Naming vocabulary	0.01	0.01	0.03
Votruba-Drzal (2006) (All families in US)
	PIAT Maths	0.02	NFER Progress in maths	0.02	0.01	0.02
	PIAT Reading	0.02	BAS Word reading	0.02	0.01	0.03
Fernald et al. (2008) (Poor households in Mexico)
	PPVT	0.21	BAS Naming vocabulary	0.03	0.02	0.05
Milligan and Stabile (2011) (Low education households in Canada)
	PIAT Maths	0.07	NFER Progress in maths	0.03	0.03	0.05
	Hyperactivity	0.07	SDQ Hyperactivity	0.0	0.0	0.0
	Conduct disorder	0.10	SDQ Conduct problems	0.03	0.02	0.05
Dahl and Lochner (2012) (Poor families in US)
	Maths	0.21	NFER Progress in maths	0.03	0.03	0.05
	Reading	0.21	BAS Word reading	0.03	0.03	0.05
Gennetian and Miller (2002) (Poor families in Minnesota)
	BPI Internalising	0.12	SDQ Internalising	0.03	0.02	0.05
	BPI Externalising	0.11	SDQ Externalising	0.02	0.01	0.04
Dearing et al. (2006) (Poor households in US)
	CBCL Internalising	0.02	SDQ Internalising	0.03	0.02	0.05
	CBCL Externalising	0.03	SDQ Externalising	0.01	0.01	0.03
Zachrisson and Dearing (2015) (All families in Norway)
	CBCL Internalising	0.02	SDQ Internalising	0.01	0.01	0.02

Effect sizes are based on standard deviation increases in outcome for a $1000 increase in income. The most similar available test, based on concept measured and age at measurement, in the MCS is used for comparison with those presented in Cooper and Stewart (2021). The age range for the LifeSim estimates are based on the sweep age rather than actual age at time of testing. All MCS measures, except the SDQ scores, are age and ability adjusted measures that are standardised within sample. PIAT - Peabody individual achievement test, NFER - National foundation for educational research, PPVT - Peabody picture vocabulary test, BAS - British ability scales, SDQ - Strength and difficulties questionnaire, BPI - Basic personality inventory, and CBCL - Child behaviour checklist. Ages in MCS correspond to MCS sweeps and not age at observation (age 3 - sweep 2, age 5 - sweep 3 and so on until age 17 - sweep 7).

Table A11. Coefficients - marginal effects on life satisfaction (0-10)

	Life satisfaction age 17
Intercept	8.456 (0.114)	Mixed	0.041 (0.085)
Poorest fifth	-0.65 (0.058)	Indian	0.172 (0.096)
Second poorest fifth	-0.451 (0.048)	Pakistani or Bangladeshi	0.204 (0.07)
Middle fifth	-0.273 (0.042)	Black	0.158 (0.092)
Second richest fifth	-0.198 (0.039)	Other	0.118 (0.132)
Wales	-0.033 (0.065)	Smoked during pregnancy	-0.198 (0.034)
Scotland	0.014 (0.062)	Age of mother at birth	0.004 (0.003)
Northern Ireland	0.219 (0.071)	No qualifications	-0.237 (0.07)
North East	0 (0.08)	NVQ level 1	-0.233 (0.072)
North West	0.007 (0.058)	NVQ level 2	-0.103 (0.057)
Yorkshire and the Humber	-0.051 (0.065)	NVQ level 3	-0.108 (0.064)
East Midlands	-0.009 (0.067)	NVQ level 4	-0.02 (0.053)
West Midlands	-0.125 (0.059)	Disability in household	-0.516 (0.082)
East of England	-0.107 (0.057)	Single parent	-0.056 (0.046)
South East	-0.048 (0.052)
South West	-0.044 (0.063)

Life satisfaction is a linear transformation of SDQ emotional symptoms scale. The estimates presented are from an OLS regressions using the preferred specification on Life satisfaction on a 0 to 10 scale.

Table A12. Marginal effect on life satisfaction of a one unit change in log household income in early childhood

Dependent variable	Age 17		Age 14		Age 11
Dependent variable	Coef (s.e.)	Ratio	Coef (s.e.)	Ratio	Coef (s.e.)	Ratio
Parent reported SDQ emotional symptoms	0.39 (0.05)	1	0.40 (0.04)	1	0.34 (0.04)	1
Parent reported SDQ internalising score	0.50 (0.05)	1.28
Self reported SDQ emotional symptoms	0.32 (0.07)	0.82
Self reported SWEMWBS	0.21 (0.04)	0.54
Self reported LS1–7			0.23 (0.04)	0.57	0.18 (0.04)	0.53

Coefficients and standard errors are calculated using the confounder list from the preferred model. Ratio is the ratio of the coefficient of our main wellbeing to other available measures in the MCS.

Table A13. Effect of Scenario 1 on early childhood outcomes based on income at 9 months and incomes 9-months to age 5

Outcome	Income Ages 0 - 5	Income at 9 months
Hospitalisation Age 3	0.01 (0.02)	0.01 (0.02)
Hospitalisation Age 5	0.01 (0.02)	0.01 (0.02)
Disability Age 3	0.00 (0.02)	0.01 (0.02)
Disability Age 5	0.00 (0.02)	0.01 (0.02)
Parent reported SDQ emotional symptoms Age 3	-0.28 (0.10)	-0.23 (0.10)
Parent reported SDQ emotional symptoms Age 5	-0.20 (0.14)	-0.09 (0.13)

Coefficients and standard errors are calculated using the confounder list from the preferred model.