Indian IPOs don't lag the market over the long run. The typical one still does

In 1991 Jay Ritter showed that American companies which went public between 1975 and 1984 did badly afterwards. Three years after listing, a rupee’s worth of new issues was worth about 0.83 of what the same money in matched firms would have become. The finding became one of the best-known anomalies in finance, and it gave a name to a common suspicion: that companies sell shares when investors are most willing to overpay for them.

India has had a decade of busy IPO markets since 2016, with 529 mainboard listings on the NSE. That is enough to ask Ritter’s question here, provided the test is fixed before the answer is known. It was: the specification was committed on 1 October 2026, before a single return was computed.

The portfolio test

Averaging each IPO’s three-year return and comparing it with the market sounds like the obvious test, but it double-counts. IPOs that list in the same busy year share the same market, so their returns are not independent, and a few enormous winners dominate any average. The test the literature settled on is a calendar-time portfolio: each month, hold every IPO that is between one and thirty-six months past its listing, equally weighted, and ask whether that portfolio beat the market after allowing for its risk.The listing month itself is left out, so the large first-day gains, which only allottees capture, play no part in the long-run test.

The risk adjustment is a regression of the portfolio’s monthly excess return on the market’s:

Here $a$ is the alpha. Ritter’s result predicts it is negative. The standard errors allow for three months of autocorrelation (Newey-West), because overlapping cohorts make neighbouring months alike.

Over 124 months, from June 2016 to September 2026, the portfolio held a median of 79 stocks. Its alpha against the Nifty 500 was +4.1% a year, with a t-statistic of 0.89. The pre-registered bar for significance was a p-value under 0.0167, because three primary tests were run; this one came in at 0.37. The sign is the opposite of Ritter’s, and it is not distinguishable from zero.

The same holds against the closer benchmarks. Against the Nifty500 Equal Weight and the Smallcap 250, which look more like the companies that list, the alphas are smaller still and none is significant.

The median IPO trails

The portfolio result is not the whole story. Buy-and-hold returns per issue tell a different one, and the gap between them is the most useful thing in the study.

At one year the mean IPO was 6.5 points ahead of the Nifty 500 and the median was 7.6 points behind. At five years the mean was 51.9 points ahead and the median 29.2 points behind. Only 41% of IPOs held for three or five years beat the index. A handful of very large winners carry the average: Mazagon Dock rose 11.4 times in three years.

For an investor this is the practical finding. Buying every IPO and holding it would have done fine. Picking one at random would usually have been worse than buying the index.

The measure behind those numbers is simple:

with both returns taken over exactly the same sessions, from the listing-day close.

def _ols_nw(y, x, lags):
    """OLS coefficients and Newey-West standard errors."""
    X = np.column_stack([np.ones(len(y)), x])
    beta, *_ = np.linalg.lstsq(X, y, rcond=None)
    e = y - X @ beta
    xtx_inv = np.linalg.inv(X.T @ X)
    S = (X * e[:, None]).T @ (X * e[:, None])
    for l in range(1, lags + 1):
        w = 1 - l / (lags + 1)
        G = (X[l:] * e[l:, None]).T @ (X[:-l] * e[:-l, None])
        S += w * (G + G.T)
    cov = xtx_inv @ S @ xtx_inv
    return beta, np.sqrt(np.diag(cov))

What it means

Ritter’s anomaly does not travel to India in this decade. That is a null result, and it is the headline. The second finding is the one investors feel: the typical IPO trails the market, and the portfolio only keeps up because a few issues multiply.