Technical appendix: statistical methods and detailed results

A1. How the indices are constructed

General principles

All indices follow a standardized construction methodology:

Index_t = Σ(w_i × normalize(X_it)) / Σw_i where: X_it = Raw value of component i at time t w_i = Weight assigned to component i normalize() = Min-max normalization to [0,1]

Normalization formula

normalize(X) = (X - X_min) / (X_max - X_min) For reverse indicators: normalize_reverse(X) = 1 - normalize(X)

A2. Statistical Suppression Index (SSI) in detail

Component weights and graded severity

SSI is a 0-10 weighted sum of graded suppression severities: SSI_t = Σ(weight_i × severity_it), severity ∈ [0,1] (1.0 = full suppression / discontinuation, ~0.7 = methodology dispute, ~0.3 = delay or post-resolution scar). There are six documented streams, and the weights sum to 10. The UPA years (2004-2013, shaded) record nothing, so SSI = 0. Both eras are computed identically in data/compute_indices.py. Cell values are the annual severity. No published third-party data is involved. These are severities coded by the author and anchored to datable events (see notes).

Year	Census (2.5)	Consumption (1.5)	Employment (1.0)	GDP back-series (1.5)	Institutional (1.5)	Mortality undercount (2.0)	SSI Score
2004	0	0	0	0	0	0	0.00
2005	0	0	0	0	0	0	0.00
2006	0	0	0	0	0	0	0.00
2007	0	0	0	0	0	0	0.00
2008	0	0	0	0	0	0	0.00
2009	0	0	0	0	0	0	0.00
2010	0	0	0	0	0	0	0.00
2011	0	0	0	0	0	0	0.00
2012	0	0	0	0	0	0	0.00
2013	0	0	0	0	0	0	0.00
2014	0	0	0	0	0	0	0.00
2015	0	0	0	.7	0	0	1.05
2016	0	0	0	.7	0	0	1.05
2017	0	0	0	.7	0	0	1.05
2018	0	0	.3	1	0	0	1.80
2019	0	1	1	1	1	0	5.50
2020	1	1	.3	1	1	0	7.30
2021	1	1	0	1	1	1	9.00
2022	1	1	0	1	1	1	9.00
2023	1	1	0	1	1	.6	8.20
2024	1	.5	0	1	1	.4	7.05
2025	1	.3	0	1	1	.3	6.55
2026	.3	.2	0	1	1	.3	4.65

Key finding: SSI was 0 across the UPA decade, then climbed to a peak of 9.0 in 2021-2022, when the COVID-mortality undercount (official ~0.5M deaths against 3-5M excess) stacked on top of the census delay, the withheld consumption survey, the shelved GDP back-series and the permanent NSSO→NSO merger. It was 4.65 in 2026. The acute events eased, and the census now runs on a notified schedule (houselisting April to September 2026, population enumeration February 2027, Ministry of Home Affairs), so it is coded 0.3 as a delay. The merger and the unresolved back-series still count at full weight, and the data scars persist.

Correction, 10 October 2026. The employment stream had been scored from 2017 and the consumption stream from 2018, before either event happened. The PLFS 2017-18 report was approved in December 2018 and withheld until 31 May 2019, and the 2017-18 consumption survey was scrapped in November 2019. Both streams are now scored from those dates. SSI 2017 moves from 2.05 to 1.05, 2018 from 4.00 to 1.80, 2019 from 4.80 to 5.50 and 2020 from 7.00 to 7.30, and the NDA mean from 4.95 to 4.78.

A3. Fiscal Centralisation Index (FCI) components

Component calculation

FCI_t = mean of the components published for year t Where: C1_t = minmax(Cess_and_surcharge_share_of_GTR_t) C2_t = 1 - minmax(Devolution_share_of_GTR_t) C3_t = 1 - minmax(States_own_revenue_share_of_revenue_receipts_t) C4_t = minmax(Non_FC_grants_share_of_Union_transfers_t) C5_t = Borrowing_Restrictions_t ∈ {0, 0.5, 1} C6_t = GST_Centralisation_t ∈ [0,1] (0 pre-2017; full once GST compensation ended 30 June 2022) C1-C4 are normalised over the years in which each is published.

Year	Cess and surcharge % of GTR	Devolution % of GTR	States' own revenue % of receipts	Non-FC grants % of transfers	Borrowing	GST	FCI Score
2004	–	25.8	62.9	32.8	0.0	0.00	0.28
2005	–	25.7	60.4	30.2	0.0	0.00	0.29
2006	–	25.4	59.5	30.7	0.0	0.00	0.32
2007	–	25.5	58.3	31.6	0.0	0.00	0.35
2008	–	26.6	58.1	35.3	0.0	0.00	0.39
2009	–	26.4	58.9	38.4	0.0	0.00	0.43
2010	–	27.7	59.1	34.5	0.0	0.00	0.34
2011	10.4	28.8	59.8	32.2	0.0	0.00	0.23
2012	11.7	28.1	61.6	29.9	0.0	0.00	0.20
2013	12.4	27.9	61.7	29.0	0.0	0.00	0.19
2014	13.5	27.1	58.0	40.0	0.0	0.00	0.42
2015	12.2	34.8	54.6	29.0	0.0	0.00	0.19
2016	13.5	35.4	52.9	27.0	0.0	0.00	0.20
2017	13.9	31.5	56.4	31.0	0.0	0.60	0.36
2018	19.9	35.9	54.7	29.2	0.0	0.70	0.38
2019	17.5	32.4	55.6	34.7	0.0	0.80	0.48
2020	24.4	29.4	52.1	37.1	1.0	0.85	0.87
2021	21.8	32.6	53.3	27.6	0.5	0.90	0.58
2022	20.4	31.1	55.9	30.3	0.0	1.00	0.51
2023	18.6	32.6	57.8	31.6	0.0	1.00	0.45
2024	18.0	33.9	56.9	31.4	0.0	1.00	0.44
2025	16.4	34.2	58.2	32.6	0.0	1.00	0.41
2026	14.3	34.7	–	–	0.0	1.00	0.35

Note: Year is the fiscal year starting April (2014 = 2014-15). Sources: cess share, 15th Finance Commission Vol I Table 3.8 (2011-2018, GST compensation cess included from 2017-18) and the Union Receipt Budgets (2019-2026, every cess and surcharge line summed; 2025 RE, 2026 BE), with no figure on this definition before 2011; devolution, 16th Finance Commission Vol II Annexure 7.1 and Budget at a Glance 2026-27; states' own revenue, RBI Handbook of Statistics on the Indian Economy 2025-26, Table 94 (2024 RE, 2025 BE); non-FC grants, 16th Finance Commission Annexure 7.1 (mostly centrally sponsored schemes, plus GST compensation grants 2017-22). The 2014 grants share (40.0%) is an accounting break, because centrally sponsored scheme funds began to flow through state budgets that year. Borrowing is 1 for 2020-21, when extra borrowing of up to 2% of GSDP was tied in part to reforms (15th FC para 10.100). A year with a component missing averages the components it has. FCI averages 0.30 under UPA and 0.43 under NDA, with a peak of 0.87 in 2020 and 0.35 in 2026. On the four measured components alone the UPA years average 0.49 and the NDA years 0.45, so the NDA's higher score comes from the GST and borrowing components.

Correction, 10 October 2026. This table previously held component values described as budget and Finance Commission anchors that matched no published series: devolution at 31-37% of gross tax revenue and falling under the NDA, states' own revenue 15-20 points below the RBI's figures, conditional transfers rising steadily to 42%, and a cess share interpolated from 8.0% (2004) to 29.0% (2024). The index has been rebuilt from the published figures above. The UPA average rises from 0.09 to 0.30 and the NDA average falls from 0.57 to 0.43; the 2020 peak moves from 0.92 to 0.87 and the 2026 value from 0.68 to 0.35. The error in the UPA years ran in this project’s favour.

A4. Democratic Quality Index (DQI) calculations

Four-component geometric mean

DQI_t = (V_t × F_t × R_t × C_t)^(1/4) Where: V_t = V-Dem Liberal Democracy Index (0-1) F_t = Freedom House aggregate score / 100 R_t = (180 - RSF_rank_t) / 180 C_t = V-Dem Core Civil Society Index (0-1)

All four are published third-party measures. The Civil Society Index is added because press-freedom rank barely separates the eras (India was mediocre throughout), while civil-society space collapsed from 0.669 to 0.302 under the NDA. A headline democracy score flattens that dimension. V-Dem first classified India as an electoral autocracy in its Democracy Report of March 2021. The values shown for 2026 are V-Dem's 2025 observations carried forward, because V-Dem has published no 2026 figure.

Year	V-Dem LDI	Freedom House	RSF Rank	Civil Society	DQI Score
2004	0.578	76/100	120/180	0.826	0.59
2005	0.573	76/100	106/180	0.760	0.61
2006	0.569	76/100	105/180	0.760	0.61
2007	0.569	76/100	120/180	0.760	0.58
2008	0.569	76/100	118/180	0.760	0.58
2009	0.539	76/100	105/180	0.760	0.60
2010	0.532	76/100	122/180	0.744	0.56
2011	0.532	77/100	131/180	0.744	0.54
2012	0.525	76/100	131/180	0.743	0.53
2013	0.522	77/100	140/180	0.735	0.51
2014	0.488	78/100	140/180	0.669	0.49
2015	0.417	77/100	136/180	0.576	0.46
2016	0.406	77/100	133/180	0.555	0.46
2017	0.371	77/100	136/180	0.523	0.44
2018	0.363	75/100	138/180	0.471	0.42
2019	0.324	71/100	140/180	0.454	0.39
2020	0.292	67/100	142/180	0.404	0.36
2021	0.307	66/100	142/180	0.421	0.37
2022	0.291	66/100	150/180	0.421	0.34
2023	0.273	66/100	161/180	0.346	0.28
2024	0.281	63/100	159/180	0.302	0.28
2025	0.260	62/100	151/180	0.302	0.30
2026	0.260	62/100	157/180	0.302	0.28

Key finding: DQI fell from a UPA-era average of 0.57 (peak 0.61, in 2005 and 2006) to 0.28 in 2026, and from 0.49 in 2014. The steepest fall came from 2019 onward. V-Dem first classified India as an electoral autocracy in its March 2021 Democracy Report. The geometric mean is now pulled down hardest by the collapse in civil-society space (0.669 → 0.302), more than by press freedom.

Correction, 15 August 2026. The two V-Dem columns above previously held approximated values whose pre-2014 entries were too high (Liberal Democracy Index 0.555 for 2014 against a published 0.488; Core Civil Society Index 0.87 against a published 0.669). That inflated the baseline and therefore exaggerated the measured decline, an error running in this project’s own favour. Both columns are now the published V-Dem series verbatim (v2x_libdem and v2xcs_ccsi). The UPA-decade mean falls from 0.59 to 0.57 and the 2014 value from 0.54 to 0.49. The 2026 value was unchanged at 0.29.

Correction, 10 October 2026. The Freedom House and RSF columns were checked against the publishers. The Freedom House scores for the UPA years had been 78-79 against published scores of 76-77, which raised the earlier baseline and ran in this project’s favour. The scores for 2024 and 2025 had been 66 against published scores of 63 and 62, which flattered the later years and ran against the project’s argument. The RSF rank for 2012 had been 140, the 2013 rank; RSF’s 2011-12 index gives 131. The UPA mean stays at 0.57, 2009 moves from 0.61 to 0.60, and 2026 moves from 0.29 to 0.28. The V-Dem columns match V-Dem’s March 2026 release.

A4b. Unpaid household work in the Time Use Survey

National accounts place own-account household services outside the production boundary, so none of the GDP figures in this appendix include them. The hours are nonetheless measured. Source: NSO Time Use Survey, January to December 2019 and January to December 2024 (MoSPI press release, 25 February 2025). The later round covered 4.54 lakh persons in 1.39 lakh households. The age group differs by row and is stated in each row, because MoSPI publishes some figures for persons aged 6 and above and others for ages 15-59. Minutes are the daily average among participants only, and are not averaged across the whole population.

Indicator	Women 2019	Women 2024	Men 2019	Men 2024
Unpaid domestic services, minutes/day, aged 6 and above	299	289	97	88
Unpaid domestic services, minutes/day, aged 15-59	315	305	n/a	n/a
Participation in unpaid domestic services, %, aged 6 and above	81.2	81.5	26.1	27.1
Participation in employment, %, aged 15-59	21.8	25	70.9	75
Caregiving, minutes/day, aged 15-59 (2024)	n/a	140	n/a	74
Caregiving participation, %, aged 15-59 (2024)	n/a	41	n/a	21.4

Derived result: the participant gap in unpaid domestic services was 299 − 97 = 202 minutes in 2019 and 289 − 88 = 201 minutes in 2024, both for persons aged 6 and above. Male participation rose from 26.1% to 27.1% over the same period while minutes among participants fell for both sexes, so the gap closed by 1 minute in five years. Female employment participation among those aged 15-59 rose from 21.8% to 25%.

A4b.1 Published valuations

These are estimates of what this work would be worth, as a share of GDP. The spread reflects the valuation method, and there is little uncertainty in the hours. This project publishes the range and adopts none of the estimates. No GDP figure anywhere in this assessment is adjusted.

Estimate	Method	% of GDP
SBI Ecowrap (2023)	minimum-wage valuation	7.5%
Ministry of Women and Child Development, 2024 submission (as reported)	not fully specified	15-17%
Opportunity cost, 2019-20 (study on CMIE household data, reported November 2024)	opportunity cost	24.6%
Economic and Political Weekly (2024), 2022-23, as reported; article not traced	replacement and opportunity cost	26-36%
Replacement cost, 2019-20 (same study)	replacement cost	32.4%

Bearing on the employment series: female labour force participation, 40.0% in 2025 on PLFS usual status after rising from 23.3% in 2017-18, is customarily read as the share of women who work, while the Time Use Survey records 81.5% of women aged 6 and above in unpaid domestic work on the reference day. The participation figure measures only the work the accounts recognise. This appendix does not resolve how much of the gap is behavioural and how much is definitional.

A5. Published estimates of employment elasticity

Definition

Employment Elasticity = (%ΔEmployment) / (%ΔGDP) Log-linear specification: ln(E_t) = α + β×ln(Y_t) + γ×X_t + ε_t Where: E_t = Employment at time t Y_t = Real GDP at time t X_t = Control variables β = Employment elasticity

This project does not estimate employment elasticity itself. The table lists published estimates for India with their source, period and method. They use different employment data and different methods, so figures from different rows are not directly comparable.

Source	Period	Estimate	Method
Planning Commission, Twelfth Five Year Plan, as cited by Misra and Suresh (2014) and Goswami (Carnegie, 2014)	1999-2000 to 2004-05	0.44	Plan sub-group estimate from NSSO employment data
Planning Commission, Twelfth Five Year Plan, as cited by Misra and Suresh (2014) and Goswami (Carnegie, 2014)	2004-05 to 2009-10	0.01	Plan sub-group estimate from NSSO employment data
Misra and Suresh, RBI Working Paper 06/2014, Table 4	1999-2000 to 2004-05	0.50	Ratio of compound annual growth rates of employment (NSSO) and GDP
Misra and Suresh, RBI Working Paper 06/2014, Table 4	2004-05 to 2009-10	0.01	Ratio of compound annual growth rates of employment (NSSO) and GDP
Misra and Suresh, RBI Working Paper 06/2014, Table 4	2009-10 to 2011-12	0.18	Ratio of compound annual growth rates of employment (NSSO) and GDP
Misra and Suresh, RBI Working Paper 06/2014	1993-94 to 2011-12	0.18 to 0.20	0.18 as arc elasticity (growth rates) and 0.20 as point elasticity (regression)
Ghosh and Roy Bardhan, ORF Expert Speak, 2 October 2024	2011-16	0.01	The authors' linear econometric model of employment on value added
Ghosh and Roy Bardhan, ORF Expert Speak, 2 October 2024	2017-23	1.11	The authors' linear econometric model of employment on value added

Note: Ghosh and Roy Bardhan use employment from the RBI KLEMS database and PLFS, and use the 2017-23 estimate to argue that growth in that period created jobs. They note that the KLEMS job estimates have been questioned because, with no census since 2011, the population projections behind them may be too high. The ILO and IHD India Employment Report 2024 finds that nearly two thirds of the employment added after 2019 was self-employment, mostly unpaid family work by women, and that nearly 90% of workers are informally employed.

Sources: Misra, S. and Suresh, A. K. (2014), Estimating Employment Elasticity of Growth for the Indian Economy, RBI Working Paper Series 06/2014, https://www.rbi.org.in/Scripts/PublicationsView.aspx?id=15763. Goswami, O. (2014), India's Quest for Jobs, Carnegie Endowment, 6 May 2014, https://carnegieendowment.org/posts/2014/05/indias-quest-for-jobs. Ghosh, N. and Roy Bardhan, A. (2024), Busting the myth of jobless growth, ORF Expert Speak, https://www.orfonline.org/expert-speak/busting-the-myth-of-jobless-growth-insights-from-data-theory-and-logic. ILO and IHD (2024), India Employment Report 2024, https://www.ilo.org/publications/india-employment-report-2024-youth-employment-education-and-skills.

A6. Measuring inequality

Income share calculation

Top 1% Share = Σ(Y_i) / Y_total for i ∈ top percentile Gini Coefficient = (1/2n²μ) × ΣΣ|y_i - y_j| Palma Ratio = Income share of top 10% / Income share of bottom 40%

Year	Top 1%	Bottom 50%
2004	18.4%	18.8%
2005	19.3%	18.4%
2006	19.7%	17.9%
2007	20.1%	17.5%
2008	20.4%	17.0%
2009	20.8%	16.5%
2010	21.2%	16.0%
2011	21.1%	15.3%
2012	21.3%	15.1%
2013	21.6%	15.0%
2014	21.3%	14.7%
2015	21.7%	14.5%
2016	22.2%	14.1%
2017	22.3%	13.9%
2018	22.4%	14.4%
2019	21.6%	14.7%
2020	21.5%	15.5%
2021	21.7%	15.4%
2022	22.6%	15.0%

Per-adult pre-tax national income shares from the Statistical Appendix of Bharti, Chancel, Piketty and Somanchi (2024), Income and Wealth Inequality in India, 1922-2023, World Inequality Lab Working Paper 2024/09, as carried in data.json. The 2022 row is the paper's 2022-23. No annual figure is published for 2023 to 2026, so the table stops at 2022. The same paper gives the top 10% share as 57.7%, the middle 40% as 27.3% and the top 0.1% as 9.6% in 2022-23. The dataset does not carry those as annual series, nor the Gini or the Palma ratio.

A7. Robustness tests

Sensitivity analysis results

Test	Base Result	Alternative Specification	Difference
SSI with ±20% weights	7.05 (2024)	5.6-8.5	±20%
FCI excluding borrowing	0.44 (2024)	0.53	+20%
DQI arithmetic mean	0.28 (2024)	0.33	+18%

Note: No significance test lies behind these rows; each reports how much the 2024 value moves under the alternative. The SSI is a weighted sum, so scaling every weight by 20% scales the index by the same proportion and leaves the cross-era comparison unchanged. Excluding the borrowing component raises the 2024 FCI because that component is zero in every year except 2020 and 2021. Dropping both coded components (borrowing and GST) reverses the era ranking. On the four measured fiscal ratios alone the UPA years average 0.49 and the NDA years 0.45. A geometric mean penalizes weak dimensions more heavily than an arithmetic mean, which is why the arithmetic DQI is higher.

A8. Complete results tables

Master results table, 2014-2026

Year	GDP Growth	CPI Inflation	Unemployment	Top 1%	SSI	FCI	DQI	Press Rank	Devolution (% of gross tax revenue)
2014	7.4%	6.7%	–	21.3%	0.00	0.42	0.49	140	27.1%
2015	8.0%	4.9%	5.0%	21.7%	1.05	0.19	0.46	136	34.8%
2016	8.3%	5.0%	–	22.2%	1.05	0.20	0.46	133	35.4%
2017	6.8%	3.3%	6.0%	22.3%	1.05	0.36	0.44	136	31.5%
2018	6.5%	3.9%	5.8%	22.4%	1.80	0.38	0.42	138	35.9%
2019	3.9%	3.7%	4.8%	21.6%	5.50	0.48	0.39	140	32.4%
2020	-5.8%	6.6%	4.2%	21.5%	7.30	0.87	0.36	142	29.4%
2021	9.7%	5.1%	4.1%	21.7%	9.00	0.58	0.37	142	32.6%
2022	7.6%	6.7%	3.2%	22.6%	9.00	0.51	0.34	150	31.1%
2023	7.2%	5.6%	3.2%	–	8.20	0.45	0.28	161	32.6%
2024	7.1%	4.9%	3.2%	–	7.05	0.44	0.28	159	33.9%
2025	7.7%	2.4%	3.1%	–	6.55	0.41	0.30	151	34.2%
2026	–	3.8%	5.2%	–	4.65	0.35	0.28	157	34.7%

Summary statistics

GDP growth (2014-15 to 2025-26, 12 years): mean = 6.20%, SD = 4.01%; 7.29% excluding 2020-21. UPA decade (2004-05 to 2013-14): mean = 6.82%; 7.23% excluding 2008-09. Including the pandemic year the NDA average is lower, and excluding each era's worst year it is higher. 2023-24 onward is on the 2022-23 base released on 27 February 2026, and the earlier years on the 2011-12 base
CPI inflation (calendar-year averages): UPA 7.98%, NDA 4.82% (2014-2026; 2026 is January to August on the new 2024 base). Inflation was in double digits in 2009, 2010 and 2013 only
Unemployment: 2014 and 2016 have no published figure. 2015 (5.0%) is the Labour Bureau's 2015-16 survey, a different survey from PLFS. On PLFS usual status the rate fell from 6.0% (2017-18) to 3.1% (2025). The 2026 value (5.2%) is the average of the monthly bulletins on current weekly status, a different measure, so no mean is taken across the series
Top 1% Share: 21.3% (2014) to 22.6% (2022-23), +1.3pp, and 18.4% in 2004; the 2022-23 level is the highest in a series that begins in 1922. Bottom 50% Share: 18.8% (2004) to 14.7% (2014), a low of 13.9% (2017), then 15.0% (2022-23). Era averages over published years: top 1% 20.39% (2004-2013) against 21.92% (2014-2022); bottom 50% 16.75% against 14.69%. No income share is published for 2023-2026
SSI: 0 across UPA; peaked 9.0 (2021-22, COVID undercount); 4.65 in 2026, with the census coded as a scheduled delay; NDA mean 4.78
FCI: UPA avg 0.30, NDA avg 0.43; 0.42 in 2014 (an accounting break in grants), peak 0.87 (2020), 0.35 (2026); on the four measured fiscal ratios alone, UPA 0.49 and NDA 0.45
DQI: fell from 0.49 (2014) to 0.28 (2026), a decline of 42.9%; UPA avg 0.57, NDA avg 0.37 (incl. civil-society index)
Press freedom: fell 17 places, from 140 (2014) to 157 (2026)
Devolution: rose from 27.1% of gross tax revenue (2014-15) to 33.9% (2024-25) (16th Finance Commission Annexure 7.1; Budget at a Glance 2026-27)

Technical appendix

Contents