Skip to content
← Back to the comparison

The counterfactual · 2014 – 2024

Rahul’s India would have been 13% poorer and 66% less secure

The same two comparisons, read from this side. Where the Modi page asks what India gained, this asks what it would have lost — across the economy, health, infrastructure and internal security. The working behind both numbers is at the foot of the page.

Income · real GDP per capita, PPP

−13%

$9416 actual · $8305 projected

The same comparison, the other way round

Every figure here is Rahul’s India measured against the India that actually happened. Read from the other direction, the same numbers describe what changed after 2014.

Switch to Modi’s India

Security · all-theatre fatalities a year

+66%

690 a year after 2014 · 2,020 before

JAMMU & KASHMIR-64%NAXAL / LWE-56%BOMB BLASTS-56%CIVILIANS KILLED-70%FORCES KILLED-62%
2004–2013 2014–2024annual averages, one scale
  • Jammu & Kashmir712 → 257/yr-64%
  • Naxal / LWE664 → 295/yr-56%
  • Bomb blasts321 → 141/yr-56%
  • Civilians killed408 → 123/yr-70%
  • Forces killed289 → 109/yr-62%

02 — sector

Economy

Output, trade and the buffer that protects the rupee.

One tile per indicator — Rahul’s India measured against Modi’s India over the same years. Shading tracks the size of the difference.

Nominal GDP-26%

USD · $3.76tn actual → $2.79tn estimated

Output roughly doubled in dollar terms across the decade, outpacing what comparable economies managed from the same starting point.
Total exports of goods and services-8%

USD · $829.8bn actual → $759.8bn estimated

Goods and services sold abroad kept climbing through the pandemic dip and past it, led by services rather than merchandise.
Total FDI inflow-68%

USD · $27.1bn actual → $8.8bn estimated

Foreign direct investment held near record levels through a decade in which global flows to emerging markets contracted.
Foreign exchange reserves-15%

USD · $399.2bn actual → $340.7bn estimated

Reserves passed $600bn, giving the Reserve Bank far deeper cover against currency pressure than it held in 2013.
Electronics production-44%

USD · $494.1bn actual → $275.1bn estimated

Electronics output multiplied as assembly moved onshore, turning a net importer of handsets into an exporter.

03 — sector

Health

Who survives childhood, and who is bankrupted by falling ill.

One tile per indicator — Rahul’s India measured against Modi’s India over the same years. Shading tracks the size of the difference.

Under-five mortality rate+38%

per 1,000 births · 26.60 actual → 36.69 estimated

Under-five deaths nearly halved. India pulled further ahead of the countries that shared its trajectory before 2014.
Neonatal mortality rate+19%

per 1,000 births · 16.70 actual → 19.91 estimated

Deaths in the first month of life fell steadily, the hardest part of child mortality to move.
Catastrophic health spending (health-coverage proxy)+100%

% of population · 20.40 actual → 40.71 estimated

The share of households pushed past a tenth of their budget by medical bills fell by half.
Out-of-pocket health expenditure as share of total h+31%

% of health spend · 43.89 actual → 57.59 estimated

Households now pay a far smaller share of the health bill directly, as public financing took over more of it.
Infant mortality rate+16%

per 1,000 births · 23.30 actual → 26.95 estimated

Infant deaths continued their long decline and stayed below the counterfactual throughout.

04 — sector

Infrastructure

The physical stock the economy runs on.

One tile per indicator — Rahul’s India measured against Modi’s India over the same years. Shading tracks the size of the difference.

National Highway network-38%

km · 146,572 actual → 91,287 estimated

The national highway network grew by more than half, with four-lane and access-controlled stock rising fastest.
Operational metro rail network-79%

km · 1155 actual → 248 estimated

Operational metro track grew several times over, from two cities of any scale to more than twenty.
Cargo handled by all Indian ports-42%

million tonnes · 1668 actual → 973 estimated

Cargo through Indian ports rose steadily while turnaround times roughly halved.
Railway electrified route length-66%

route-km · 63,456 actual → 21,614 estimated

Broad-gauge electrification moved from a minority of the network to almost all of it.
Cities with operational metro systems-81%

cities · 26.00 actual → 5.00 estimated

Metro systems opened in city after city, spreading urban rail well beyond the metros that had it in 2014.

05 — sector

Security

Violence across the three theatres, and the bombs.

One tile per indicator — Rahul’s India measured against Modi’s India over the same years. Shading tracks the size of the difference.

J&K civilian deaths from terrorism+298%

deaths · 385 actual → 1531 estimated

Civilian deaths from terrorism in Jammu & Kashmir fell to a fraction of the previous decade's toll.
Civilian deaths from LWE+219%

deaths · 1495 actual → 4766 estimated

Civilian deaths in Maoist-affected districts fell sharply as the affected area contracted.
Security personnel killed by LWE+264%

deaths · 509 actual → 1851 estimated

Security-force losses to left-wing extremism fell by roughly three quarters.
Northeast civilian deaths+309%

deaths · 605 actual → 2476 estimated

Northeast civilian deaths fell furthest of any theatre, alongside a series of settlements.
Terrorism/insurgency explosion incidents+156%

incidents · 1337 actual → 3425 estimated

Explosion incidents across all theatres fell by well over half, and by 2024 stood at a fifth of the 2004 level.

Supplementary analysis

Everyday India — 4 earlier-period leads

On 4 of the 20 selected lived-development measures, the earlier period improved faster once starting position is accounted for.

Child health
BCG vaccination coverage
Women's nutrition
Women with below-normal BMI
Social development
Women aged 20-24 married before age 18 · Men aged 25-29 married before age 21

The earlier period also established substantial progress on several other measures, including institutional births and skilled birth attendance. After adjusting for starting position, 3 of those are classified as broadly comparable rather than a lead for either side. See all 20.

06 — Where Rahul’s India is ahead

3 national indicators run the other way

Most of this page describes a shortfall. These do not: on these measures the estimated path, or the earlier period, is the one in front. They are a small minority, and shown here rather than buried because a page that only reported losses would not be worth trusting on the rest.

4 more sit in the Everyday India layer directly above, scored on normalised rates of improvement so a period that started further behind is not credited for the easier gains.

07 — The wealth gap

13% poorer

Real GDP per capita, adjusted for purchasing power. India’s actual path is the orange line; Rahul’s India — the weighted blend of 27 comparator economies whose pre-2014 paths tracked India’s — is the blue dashed line. The two are indistinguishable before 2014 by construction, and separate afterwards.

Rahul's India
$8305
projected 2024
Modi's India
$9416
actual 2024
Difference
+13%
in Modi's favour
Placebo rank
2/36
p = 0.056

What Rahul’s India is made of

  • Tanzania
    61%
  • China
    15%
  • Bangladesh
    11%
  • Honduras
    9%
  • Thailand
    4%
  • Uganda
    1%

The blend reproduces India’s own pre-2014 path to within 1.3%. Re-running the whole procedure pretending each comparator was the country treated in 2014 puts India at rank 2/36, p = 0.056 — a gap larger than chance usually produces.

08 — The security gap

66% less secure

Violence did fall after 2014 — all-theatre fatalities are down 26% on 2013 and battle deaths per million down 30%. The question this site asks is different: did it fall by more than comparable countries managed? On the one security measure with a counterfactual behind it — UCDP battle-related deaths per million, against comparator countries — the answer is no. India’s gap oscillates around zero across the whole period and lands at rank , p = . Three ways of asking give three different answers, set out below.

MethodGapPlacebo test
Cross-country SCM (UCDP battle deaths)+6.5%rank null, p = undefined
Home Ministry totals, counterfactual held flat-65.8%none available
Home Ministry totals, 2004-13 decline continued+75.3%none available

Annual averages, so the ten-year and eleven-year windows compare like with like: 2,020 deaths a year across 2004-2013 against 690 across 2014-2024. Averaging the eleven Home Ministry parameters individually gives -66.0%, within a fifth of a point. Two things to carry with it. The pre-2014 window contains the mid-2000s peak, and against the three years immediately before 2014 the same seven parameters average -1%, because Kashmir turned back up while the two insurgencies kept falling. And this is arithmetic on India's own totals — there is no comparator and no placebo test. The one security measure with a cross-country counterfactual behind it, UCDP battle deaths against 27 comparators, returns +6.5% at rank 25/36, p = 0.694.

Combined fatalities — Jammu & Kashmir, left-wing extremism, the Northeast

2000–2013 2014–2024peak 2001: 5,445

Averaging 2,020 deaths a year over 2004–2013 and 690 over 2014–2024 gives a fall of 66% — but that comparison hides where the fall happened, because the earlier window contains the violent early 2000s. Splitting it finer shows two separate shifts.

Period averages

  • 2000–2003
    4,576
  • 2004–2007
    2,783
  • 2008–2010
    2,119
  • 2011–2013
    902
  • 2014–2018
    864
  • 2019–2024
    546

The first Modi term is flat: 864 a year over 2014–2018 against 902 over 2011–2013, -4%. The shift lands in 2019 — the single largest post-2014 drop is 2018→2019 -34% — after which the average is 546, down 37% on the first term. The larger break in the whole series is earlier still: 2010→2011 -45%, in 2011.

Parameter by parameter, on decade totals

Year-to-year conflict counts swing on single events, so decade totals are the sounder unit. Here are the seven parameters with annual data behind them, under the decade comparison and against the three years immediately before 2014.

Parametervs 2004–2013
decade total		vs 2011–2013
annual average
J&K terrorism-related deaths-60%+63%
J&K civilian deaths-74%+48%
J&K security-force deaths-46%+66%
LWE civilian deaths-63%-55%
LWE security-force deaths-67%-53%
Northeast civilian deaths-73%-38%
Northeast security-force deaths-71%-37%
Mean-65%-1%

Left-wing extremism and the Northeast

Improve on both views. LWE civilian deaths are down 62% on the decade and still down 54% against the 2011–2013 average; Northeast security-force deaths down 71% and 36%. These are sustained declines that do not depend on which baseline you pick.

Jammu & Kashmir

Splits. Down 60% on the decade, but J&K violence was already at a historic low by 2011–2013 — 158 deaths a year. Against that baseline the post-2014 average of 257 is 63% higher. Civilian deaths +48%, security-force deaths +66%.

Both views are defensible and they answer different questions. Against the full previous decade every parameter improves, by 65% on average — that window contains the mid-2000s peak. Against the three years immediately before 2014 the average is -1%, because the two insurgencies kept falling while Kashmir turned back up. The aggregate figure is the sum of movements in opposite directions, which is also why the cross-country test comes back null: the theatres cancel.

All ten, as supplied

So there is a real post-2014 shift, and it is in the second term rather than the first. What the cross-country test adds is the comparison: conflict levels in the comparator countries fell over those same years, so India’s decline does not separate from theirs — rank , p = . The decline is visible in the data; what cannot be shown is that it is specific to India.

09 — The two numbers are not alike

One is tested, one is arithmetic

13% poorer

Estimated from 27 comparator economies, validated against India’s own pre-2014 path to 1.3%, and tested against every comparator as a placebo — rank 2/36, p = 0.056. It clears three of the four robustness runs above, and fails the fourth.

66% less secure

Arithmetic on India’s own totals, not an estimate: 2,020 deaths a year across 2004–2013 against 690 across 2014–2024. Every one of the eleven parameters improves, so it does not rest on a single series — but there is no comparator and no placebo test, so nothing constrains it. Put to cross-country conflict data instead, the same question returns +6.5% at rank 25/36, p = 0.694.

10 — Does it hold?

Four ways to break the wealth figure

A gap against a counterfactual is only as good as the counterfactual. These are the standard attacks on a synthetic control estimate, run against this one.

  • In-time placebo

    passes, with a caveatDoes a fake break produce the same gap?

    Setting the treatment to 2009 and cutting the sample at 2013 gives a gap of +4.7% after four years. The real 2014 break gives +8.9% over the same horizon — about twice as large. So 2014 is not purely trend, but a placebo year still produces roughly half the effect.

  • Leave-one-donor-out

    passesIs it driven by a single country?

    Dropping each weighted donor in turn moves the gap between +8.1% and +15.3%. It never changes sign. The largest single dependency is Tanzania, whose removal takes it to +8.1%.

  • Oil-importer-only pool

    passesIs it terms of trade?

    Excluding the 8 net fuel exporters (BOL, CIV, COL, ECU, EGY, IDN, MMR, NGA) leaves 22 donors and a gap of +13.4% — unchanged from the baseline. The commodity-cycle objection does not explain it.

  • Alternate source and pool

    failsDoes it survive a different GDP series?

    Run on Penn World Table 11.0 instead of WDI, over the 14-country pool from the Grier & Grier paper, the gap is -16.4% — the opposite sign. Holding that pool fixed and switching only the source gives -4.5% on WDI, so both the narrower pool and the source move it. The +13.4% headline depends on the wide 27-donor WDI pool.

Three of four hold; the fourth does not. The wealth gap is not an artefact of one donor or of the commodity cycle, and it is larger than a placebo break produces. But it does not survive being re-run on a different income series over a narrower pool, where it turns negative. That is the strongest objection to the 13.4% figure, and it comes from our own test rather than from a critic.

← Back to the comparisonThe strongest measured result →