Latest · Does privatisation work? Previous · Do markets care about democracy? Earlier · The de-rating of Indian markets
Delhi · Odisha · Mumbai · Chandigarh · Haryana · Hindustan Zinc · VSNL · Maruti · BALCO · Air India

Does privatisation work?

Haryana's power workers struck for three days this week over a licence that lets a private company compete with the state distributor in Gurugram and Nuh. We went back through every major Indian electricity privatisation and four strategic sales, before and after, with the consumer's bill in the middle of the table. Losses fell wherever the contract made them fall. Consumers got supply, and a deferred bill. And the stakes the government kept turned out to be the most profitable thing it ever sold.

24 OCTOBER 2026 DATA TO 23 OCTOBER 2026 FULL REPORT · 17 EXHIBITS

Eight findings

The finding

Delhi's distribution losses fell from 50.7% in 2002 to 5.4% at Tata Power-DDL in FY2025-26 and 8.36% across the city in FY2024-25. Consumers got round-the-clock supply and a first-slab tariff of ₹3 a unit. They also acquired a deferred bill: the ₹38,552 crore of regulatory assets the Appellate Tribunal ordered recovered over about 7 years in April 2026, roughly ₹55,074 per connection.

The government that sold 44.92% of Hindustan Zinc for ₹769 crore in 2002-03 has since collected ₹36,876 crore in dividends on the 29.54% it kept, sold a sliver for ₹3,449 crore, and still holds a stake worth ₹58,513 crore. Sell the control, keep a stake: on the Indian record, that is the trade that worked.

  1. Haryana is not selling anything, and its discoms are not failing. The Haryana Electricity Regulatory Commission granted a second, parallel distribution licence for Gurugram and Nuh on 19 October 2026. The state keeps DHBVN; a new company may build its own wires and compete. The Commission's stated reason was 108,590 pending connection applications at DHBVN, 1,643 MW of demand waiting. Haryana's two public discoms cut AT&C losses from 29.58% in FY2014-15 to about 10% now, and have been profitable since FY2017-18, after the state took ₹25,950 crore of their debt. The fight is about who serves Gurugram's growth, not about rescuing a utility.
  2. Delhi worked on losses and supply. The regulator's opening AT&C figure in February 2002 was 50.7%, 57.2% in the east. The Delhi government paid ₹3,452 crore of transition support over four years and then nothing. Today the three private discoms are below 6.5% on CSEP's figure, against an all-India 15.04%. Peak demand served has gone from 2,879 MW to 7,409 MW, on an industry figure; the six-to-seven-hour outages that were routine before 2002 are not disputed by anyone.
  3. Delhi's consumers paid, and part of the bill is still coming. Tariffs rose 21.77% in FY2011-12 and 26% plus an 8% surcharge in FY2012-13. Regulatory assets grew from ₹697 crore in FY2004-05 to ₹33,961 crore by FY2014-15; the discoms now claim ₹91,516 crore, two thirds of it carrying cost, and the Tribunal has ordered ₹38,552 crore recovered. From 1 June 2026 a monthly surcharge reached 17.94% of a BSES Rajdhani bill. The state spends about ₹3,353 crore a year making the first 200 units free. And about 75% of a Delhi bill is the cost of power bought from generators, which no owner of wires controls.
  4. Odisha failed, then worked, with the same consumers. The 1999 contracts carried no loss-reduction obligation; losses were about 44% in 2001 and still 36 to 43% when the licences were revoked in March 2015. The 2020 design put a year-by-year loss trajectory and a ₹5,640 crore capital commitment into the vesting orders, gave the state 49%, and kept retail tariffs frozen for 5 years. Losses have fallen from 29.5% to 14.6%, profit after tax doubled to ₹809 crore in FY2025-26, and the four discoms paid their first dividends. The variable was the contract, not the owner.
  5. Franchisees fail more often than they work. Of 28 appointed since 2006, 12 are running and 16 were terminated or never started. The two celebrated ones, Bhiwandi (58% to 9.64%) and Agra (58.77% to 9.16%), share one parent with a balance sheet. The failures defaulted on input power within a year or two.
  6. Public ownership can do it too. Kerala's state utility is at 6.61%, Andhra Pradesh at 7.87%, Gujarat at 8.25%, all state-owned and all better than Delhi's 8.36%. Six of the fifteen A+ utilities in the Ministry of Power's latest rating are state-owned. The 12 private discoms together sit at 10.05%; the national figure is 15.04% and Purvanchal, which Uttar Pradesh is trying to sell, is at 26.33%. Ownership is not destiny. Tolerance of theft and of tariffs below cost is.
  7. Mumbai is the precedent for Gurugram, and it is a warning about design, not competition. 5.74 lakh suburban consumers, 19% of the base, had switched supplier by FY2015-16. The large consumers who fund the cross-subsidy went first, 49% of them by FY2012-13, and went back once the regulator imposed the surcharges it should have set on day one, 21 months late. Neither licensee cut costs; regulatory assets grew on both sides. The Haryana order rests on the Tribunal's March 2026 Mumbai judgment, so the same questions, cross-subsidy and network duplication, arrive with it.
  8. Sell the control, keep a stake. Hindustan Zinc: ₹769 crore for 44.92%; the kept 29.54% has produced ₹98,839 crore in dividends, a part-sale and current value, 129 times the control price, and ₹36,228 crore of those dividends came after the government refused Vedanta's ₹15,493 crore in 2012. VSNL: ₹1,439 crore for control in 2002, then ₹10,998 crore from the 26.12% kept for nineteen years. Maruti, where nothing was kept: the ₹3,927 crore raised in 2006-07 would be ₹64,602 crore today in shares and dividends, 2.8 times what the Sensex would have done with the cash. BALCO's 49% share of one year's profit, ₹679 crore, exceeds the ₹552 crore the 51% fetched in 2001. Delhi and Odisha hold 49% of their private discoms, and Tata Power-DDL earned ₹1,299 crore last year. The caveat is survivorship: the 49% Odisha kept in the 1999 discoms was worth nothing when the licences were revoked, and Air India had nothing worth keeping.
50.7% → 5.4%
Delhi AT&C loss, February 2002 to FY2025-26 (Tata Power-DDL)
₹38,552 crore
Regulatory assets APTEL ordered Delhi's consumers to repay, April 2026
₹98,839 crore
What the Hindustan Zinc stake the government kept has produced, against ₹769 crore for the control sale
12 of 28
Distribution franchisees still operating, of all appointed since 2006

That is the executive summary of the report, in full. Everything under it is the evidence: 17 exhibits across Delhi, Odisha, Mumbai, Chandigarh and Haryana, the consumer's tariff and the regulatory asset bill year by year, the franchisee scoreboard, the public utilities that cut losses without a sale, and the full arithmetic on the stakes the government kept in Hindustan Zinc, VSNL, Maruti and BALCO.

Sign up and read the evidence

You have just read the whole summary. The report is where it is proven. Leave your name and email and you get it here, plus our future research. No spam, no card.

  • 17 exhibits, each with its sources and definitions stated beneath it
  • Delhi, 2002 to 2026: the regulator's opening loss by circle, the tariff and support record year by year, the regulatory assets, and what share of a bill the private operator actually controls
  • Odisha twice: the 1999 contracts that failed and the 2020 vesting orders that are working, discom by discom against their targets
  • 28 franchisees, 12 running and 16 gone, and the state utilities that beat Delhi without privatising
  • Mumbai's parallel licence, year by year, which is the model Haryana has just adopted
  • Hindustan Zinc, VSNL, Maruti and BALCO: what the government sold, what it kept, and what each is worth today, with the counterfactuals priced
  • Our view on how to privatise, what to keep, and what the Haryana appeal should actually be about
31 markets · October 2021 to October 2026 · The 2025-26 protest wave

Do markets care about democracy?

We measured 31 stock markets on a single dollar basis from 15 October 2021 to 16 October 2026, sorted them by the Economist Intelligence Unit's democracy score, and dated every major protest of the past fourteen months, including India's. Democracies outperformed autocracies by about 81 points. Protests, including India's, barely moved prices. India, a democracy, returned like an autocracy.

17 OCTOBER 2026 PRICES TO 16 OCTOBER 2026 FULL REPORT · 18 EXHIBITS

Seven findings

The finding

$100 invested on 15 October 2021 was worth $173 in the median full democracy on 16 October 2026, and $92 in the median authoritarian market. Over the last twelve months the gap is +21.4% against -13.1%.

India had the largest youth protest movement in the world from 6 June to 25 July 2026, and a second phase has been under way since 10 October. Its market moved +0.7% against world equities in the first week of the first phase, +2.2% after a month, and +5.5% in the first week of the second. $100 invested in India on 15 October 2021 is worth $101, 25th of 31.

  1. Democracies outperformed autocracies, in every window we tested. From 15 October 2021 to 16 October 2026, the median full democracy returned +72.9%, the median flawed democracy +70.1%, and the median authoritarian market -7.8%, all in dollars with dividends. The ranking holds over 10 and 15 years and over 2011 to 2021. The correlation between democracy score and return across the 31 markets is +0.42.
  2. Oil did not create the gap. It hid part of it. Four of the six authoritarian markets are Gulf hydrocarbon exporters and they returned a median +12.2%: UAE +61%, Kuwait +32%, Qatar -7%, Saudi Arabia -8%. The two authoritarian markets without oil, China and Vietnam, returned -18.6% and -9.6%. Remove every hydrocarbon exporter from every bucket and the authoritarian median falls from -7.8% to -14.1%. Two markets are a thin base; the direction is not in doubt.
  3. Autocracy does not buy stability. Authoritarian markets had the lowest day-to-day volatility in the set, 17.1% annualised, and a maximum drawdown of -30.6%, within a point of the full democracies' -33.9%. Calmer in ordinary weeks, identical in the bad ones, negative over the period.
  4. The table flatters autocracies, because the worst cases are no longer in it. MSCI priced every Russian security at 0.00001 on 9 March 2022; the MSCI Russia index returned -99.94% in the five years to its discontinuation on 31 May 2022, and holders of the iShares Russia ETF have recovered $2,262 per $10,000 invested in 2015. Egypt's and Nigeria's ETFs were closed for poor performance in 2024 and 2023. Add Russia at its index value and the authoritarian bucket's average return falls from +8.2% to -7.2%.
  5. Protests had no consistent effect on prices. Peru removed its president and beat world equities by +36.9% over the following four months. Indonesia's protests produced a ministerial resignation and a -13.8% shortfall against the world. South Korea's martial-law declaration cost -8.1% and was recovered in 31 trading days. What separated the outcomes was whether the political question was settled, not whether it was raised. Indonesia's real shock arrived on 16 October 2026, when MSCI proposed stripping it of emerging-market status and the Jakarta index fell 11% in a day.
  6. The winners are mostly what their indices are made of. Peru's ETF is 53% materials and, with world equities stripped out, its returns correlate 0.50 with gold and 0.47 with copper, the strongest commodity exposures in the set. Taiwan's partial correlation with semiconductors is 0.53. Remove Taiwan and South Korea altogether and democracies still lead autocracies by 72 points. India has no measurable exposure to any of the four drivers beyond world equities, and a negative one to oil.
  7. India's market ignored India's protests, and that is the diagnosis. A market that does not react to its largest political mobilisation in a decade is not pricing political risk. It is pricing a multiple and a currency, which is what our previous report found. Among the five markets with a major protest since August 2025, India's -9.6% is second worst; only Indonesia, at -41.9%, did worse. Youth unemployment does not predict returns either: the correlation across 30 markets is +0.12.
$173
$100 in the median full democracy, Oct 2021 to Oct 2026
$92
$100 in the median authoritarian market, same period
$101
$100 in India, same period. 25th of 31.
+2.0%
Median Indian stock, 6 June to 25 July 2026, the protest window

That is the executive summary of the report, in full. Everything under it is the evidence: 18 exhibits across 31 markets, an interactive day-by-day view of every protest against world equities, the oil and AI controls, the survivorship record with Russia quantified, India's live October phase, and the Indian sector table through India's own protest window.

Sign up and read the evidence

You have just read the whole summary. The report is where it is proven. Leave your name and email and you get it here, plus our future research. No spam, no card.

  • 18 exhibits, each with its sources, symbols and definitions stated beneath it
  • 31 markets on one dollar basis, October 2021 to October 2026, sorted by the Economist Intelligence Unit's Democracy Index
  • The oil control that widened the gap, and the AI control that did not change the order
  • Eight dated political events measured against world equities, day by day, with an interactive chart
  • Russia, Egypt and Nigeria: the markets that left the universe, with Russia's exit quantified from MSCI and iShares figures
  • India's two protest phases, the sector table through the first, and what the market's indifference actually tells you
  • Youth unemployment against returns, and fifteen years of history outside the AI cycle
NIFTY-500 · Earnings versus price study

The de-rating of Indian markets

The Indian market peaked on 26 September 2024 and has gone nowhere since. Its companies kept growing throughout. This is a study of what actually happened, what did not cause it, and where the money went instead.

20 SEPTEMBER 2026 PRICES TO 18 SEP 2026 FULL REPORT · 18 EXHIBITS

What actually happened

The context, then the finding

Indian markets have been de-rating for two years. The NIFTY 500 peaked at 24,496.9 on 26 September 2024 and is 6.8% below that level today, almost exactly two years later.

Over those same two years its companies grew operating profit 29.8%. Over the last twelve months all twelve NIFTY-500 sectors grew operating profit, and nine of the twelve still fell in price. The market did not stop earning. It stopped paying.

  1. Indian markets have been de-rating since 26 September 2024. The NIFTY 500 peaked at 24,496.9 on that date and has never exceeded it. The twelve months before returned +38.7%. The two years since have returned -6.8%. Nothing has been earned since.
  2. Earnings did not break with the price. Median operating profit growth across the NIFTY-500, in rupees, ran +21.1%, then +13.4%, then +13.6% across the last three fiscal years. FY24's 21% was the outlier, not the 13%. The FY19 to FY23 median was 14.9%, so roughly 13 to 15% is India's normal and growth returned to trend rather than deteriorating. Sales growth never moved at all: 12.4%, 11.2%, 12.2%.
  3. Foreign outflows were a symptom, not the cause. Over two years foreigners sold 2.96% of Korea's market value, 1.78% of Taiwan's and 0.83% of India's, and those markets returned +187.3%, +105.5% and +2.1%. However the selling is scaled, its size does not track the outcome across these three markets, though with three observations that is an absence of evidence rather than proof of absence. A direct test on 43 months of Indian data agrees: monthly FPI flows do not predict next month's returns (r = -0.125), while past returns do predict next month's flows (r = +0.296). FPIs sell a median of ₹22,748 crore in the month after the market falls, and buy ₹11,946 crore after it rises, and the month after that is mildly weaker than average, though not separably so on ten observations. Capital left India because the earnings were elsewhere. That is the cause; the outflow is what it looked like.
  4. The money went into AI, and into more than semiconductors. Since 26 September 2024, in USD: grid equipment (GE Vernova) +275.9%, gold miners +133.7%, semiconductors +126.8%, grid construction (Quanta) +115.3%, cybersecurity +69.4%, biotech +60.4%, uranium +53.8%, aerospace and defence +46.4%. The NIFTY 500 returned -6.8% in rupees and -17.7% in dollars. The single best large AI stock was SK Hynix at +927%.
  5. India's index owns almost none of it, and the rupee compounded the damage. The entire listed electrical, electronics and AI-infrastructure complex is 3.3% of NIFTY-500 market value. Separately the rupee cost a dollar investor 13.3% over three years, worse over three years than every currency we compared except the rupiah, while the Chinese yuan appreciated 8.8%.
  6. What India does own in that complex is winning. That 3.3% returned a median 140.1% over three years against 46.3% for the median stock, and 36.1% over the last year against -3.8%. The exposure is small. The signal is not.
  7. The last three months are a rotation, not an ending, and India is on the right side of it for once. Korea's fall is a valuation correction: its P/E went from an all-time high of 33.5 in June 2026 to 18.35 today while earnings kept growing. Meanwhile money moved from AI hardware into AI software: cloud +22.9%, cybersecurity +18.2%, software +17.1%, against semiconductors -13.2%. Indian IT is showing the same split, with Coforge +26.1% and Persistent +12.5% while TCS is -0.4%.
26 Sep 2024
The day the NIFTY 500 peaked
-6.8%
NIFTY 500 since that day
+29.8%
Median operating profit growth over the same two years
+927%
SK Hynix, the best large AI stock, over the same two years
From the full report
Where the money went: three years of market performance, indexed
Every market set to 100 on 20 September 2023, so the lines show cumulative return. Weekly. Click a market to add or remove it, and switch between local currency and dollars. The dashed gold line marks 26 September 2024, the day India peaked.
Price indices, excluding dividends. The dollar view converts at spot on each date. India is the NIFTY 500; Semis is the PHLX Semiconductor index; US is the S&P 500.
Sources and definitions. Closing levels rebased to 100 on 20 September 2023, weekly. Symbols: NIFTY 500 (^CRSLDX), KOSPI (^KS11), TAIEX (^TWII), PHLX Semiconductor (^SOX), S&P 500 (^GSPC), Nikkei 225 (^N225), Shanghai Composite (000001.SS), gold front-month (GC=F). USD view converts at spot on each date. Price indices, excluding dividends.

That is the executive summary of the report, in full. Everything under it is the evidence: eighteen exhibits, three separate tests of the foreign-flows explanation, the sector and company detail, and an appendix naming every stock in every grouping so you can check the work rather than trust it.

Sign up and read the evidence

You have just read the whole summary. The report is where it is proven. Leave your name and email and you get it here, plus our future research. No spam, no card.

  • Eighteen exhibits, each with its sources, symbols and definitions stated beneath it
  • Three separate tests of the foreign-flows explanation, including 43 months of monthly FPI data against index returns
  • The complete sector table, and 72 companies above ₹20,000 crore that grew operating profit and fell anyway
  • India's AI-infrastructure complex, all 21 names, with the valuation counterweight against global comparables
  • What previous semiconductor cycles did next, and where this one appears to be rotating
  • Three places India's strengths meet the AI build, and six things to watch
  • An appendix naming every stock in every grouping, plus what this study cannot tell you and what we could not source
Please enter your name and a valid email address.
The report opens here, we do not email it. No spam, no sharing your address. Information and education only, not investment advice.