IMAGE: SHUTTERSTOCK
IN THE late 1990s the Nepal Stock Exchange briefly became a model of market rationality. For a few months, past prices could not predict future prices. Statistical tests that normally reject efficiency came back empty-handed. Then the moment passed. Inefficiency returned and stayed for two decades. Two years ago, it happened again. The same tests began to show that NEPSE was once more behaving as if all available information was already in the price.
A new study by Saurav Karki, using daily trading data from July 1995 to February 2025, makes the case that efficiency in Kathmandu’s only stock exchange is not a fixed trait. It arrives, disappears and sometimes arrives again. That pattern fits the Adaptive Market Hypothesis, an idea that blends behavioural finance with evolutionary biology. Investors try strategies that work. Those strategies attract imitators. The edge erodes. Then a shock resets the game. Markets do not trend towards permanent efficiency. They cycle through efficient and inefficient phases.
The study, published in the NRB Economic Review, covers nearly 30 years of daily returns, divided into two-year slices and also examined through rolling five-year windows. The most striking finding is how rare clean efficiency has been.
From 1999 to 2019, NEPSE was consistently inefficient. Prices moved in ways that could be predicted from past prices. Then, from 2021 to 2025, signs of efficiency began to emerge. The timing points to two technological changes: screen-based trading introduced in August 2007 and a fully automated online system launched in January 2021. The pandemic pushed the last holdouts onto digital platforms. By 2024-2025 the same three statistical tests that had signalled efficiency in 1997-1999 once again failed to find predictable patterns.
But the two efficient episodes had different roots. The first, in 1997-1999, coincided with political instability and a drop in trading volumes. Fewer participants meant less competition to find and exploit mispricing. Efficiency emerged not because information travelled faster but because the market had become too slim to matter. The second episode, by contrast, arrived with a surge in demat accounts, better regulatory oversight and automated settlement. That kind of efficiency is more likely to last.
What has not changed is volatility. Even as return predictability has weakened, the tendency for turbulent days to cluster together has persisted. The study estimates that roughly 90% of a day’s volatility carries over to the next day. A sharp fall or rise is followed by another. Shocks decay slowly. Investors cannot assume that a bad day will be a one-off.
Nor does the market distinguish between good news and bad news. In developed stock exchanges, negative surprises typically drive up volatility more than positive surprises of the same size. That asymmetry reflects two things: falling share prices increase debt-to-equity ratios, and investors feel losses more acutely than gains. NEPSE shows no such effect. A 2% rise and a 2% fall produce similar jumps in volatility. The explanation lies in who trades. Retail investors dominate. Short selling is absent; so are derivatives. Margin trading is minimal. Both a surprising gain, which triggers a fear of missing out, and a surprising loss, which triggers panic selling, generate the same emotional response.
The study also identifies two distinct market regimes. In the calm state, daily volatility runs at about half a per cent and average returns are near zero. In the turbulent state, volatility jumps to around 1.7% daily, and average returns become modestly positive. That last detail matters. High volatility in NEPSE is not just about falling prices. It tends to arrive with speculative rallies and heavy trading.
The two regimes have different staying power. Calm periods last about 15 trading days on average. Turbulent periods last about nine days. The asymmetry suggests that while shocks hit frequently, the market has developed some ability to stabilise. It recovers faster than it destabilises. That is a small vote of confidence in the resilience built over three decades.
For investors, the lesson is not to adopt a single strategy and hold on. Strategies that work in high-volatility periods—riding momentum, betting on reversals—may fail when the market turns calm. In low-volatility states, passive indexing (which is not available in Nepal) becomes more attractive because arbitrageurs have already squeezed out the easy mispricings. For regulators, the finding supports continued investment in digital plumbing and liquidity programmes. The fact that turbulent episodes are shorter than they used to be is good news. But the high persistence of volatility means circuit breakers and market surveillance should be designed for weeks of instability, rather than a day or two.
The stock exchange that briefly became efficient in the late 1990s later unlearned the habit. The efficiency gains of 2024-25 will be tested by the next political or economic shock, and by how well the market has truly learned to adapt. ■
