Graham Net-Net on 7 Global Markets: One Works, Two Are Cautionary
We ran Graham's net-net strategy, buying stocks trading below their net current asset value, across seven global markets over 24 years. Japan works. Most don't. Hong Kong was a disaster.
Contents
- The Strategy
- The Summary Table
- Japan: Why It Works
- United States: The Eroded Premium
- Hong Kong: The Cautionary Tale
- India: Data Coverage Note
- Korea, Canada, UK
- What to Take From This
- Excluded Markets
Here's the full cross-market breakdown with year-by-year data.
The Strategy
A net-net stock trades below its liquidation value. Specifically:
NCAV per share = Current Assets − Total Liabilities − Preferred Stock (per share)
If a stock's price is below this figure, you're theoretically buying the liquid assets at a discount before the business operations factor in at all. Graham called these "bargain issues", the margin of safety is in the balance sheet, not in earnings projections.
Backtest setup: - Annual rebalancing (April) - Equal weight all qualifying stocks - Hold cash if fewer than 5 qualify - 45-day lag on financial data (no look-ahead bias) - Market cap thresholds set per exchange to match net-net universe (all are small/micro-cap) - Benchmark: S&P 500 Total Return (SPY)
The Summary Table
| Exchange | CAGR | SPY | Excess | Sharpe | MaxDD | Avg Stocks | Cash Periods |
|---|---|---|---|---|---|---|---|
| Japan (JPX) | 8.85% | 8.84% | +0.02% | 0.416 | -33.7% | 27.9 | 6/24 |
| Canada (TSX) | 7.49% | 8.84% | -1.34% | 0.133 | -55.9% | — | 0/24 |
| India (NSE)* | 8.28% | 8.84% | -0.55% | 0.042 | -57.7% | 26.0 | 8/24 |
| Korea (KSC) | 6.60% | 8.84% | -2.24% | 0.162 | -39.3% | — | 7/24 |
| US (NYSE+NASDAQ+AMEX) | 5.02% | 8.84% | -3.81% | 0.078 | -54.4% | 28.5 | 0/24 |
| UK (LSE) | -0.62% | 8.84% | -9.46% | -0.149 | -43.5% | — | 7/24 |
| Hong Kong (HKSE) | -3.17% | 8.84% | -12.01% | -0.199 | -82.1% | 24.1 | 2/24 |
*India (NSE): data coverage starts 2009. See note below.
Six of seven markets underperformed the S&P 500. Japan is the exception.
Japan: Why It Works
Japan averaged over 100 qualifying net-net stocks per year. The structural reasons: companies hoard cash, governance reform was slow, and decades of deflation meant many profitable businesses traded at distressed valuations.
Japan net-nets returned 8.85% CAGR over 18 active years (2007-2024), matching the S&P 500 but with better risk characteristics:
- Max drawdown: -33.7% vs -40.3% for SPY
- Down capture: 47.4% (when SPY falls, Japan net-nets fall less than half as much)
- Sortino ratio: 1.167 vs 0.889 for SPY
The strategy beat the S&P 500 in 8 of 18 active years. Notable outperformance: 2011-2014 (Abenomics governance reform), 2022-2024 (Japan equity renaissance).
The risk-adjusted case for Japan net-nets is real. It's not a guarantee, 2018-2021 were rough, but the structural supply of net-nets combined with an identifiable catalyst (governance pressure to return cash) creates a genuine edge.
United States: The Eroded Premium
Academic research from the 1980s found substantial net-net premiums in US markets. Oppenheimer (1986) documented 29% annual returns from 1970-1983. That premium has largely disappeared.
Our data: 5.02% CAGR over 24 years, underperforming SPY by 3.81% annually. The strategy had zero cash periods, stocks were always available, but holding them didn't pay.
The exception was crash-recovery years: 2001 (+30.8%), 2003 (+82.4%), 2009 (+97.3%), 2020 (+103.8%). In years following severe market dislocations, deeply discounted stocks bounced hard. The problem: the losses in non-crash years were severe enough to offset those recoveries over the full period.
US net-nets today are predominantly struggling companies in secular decline: retail, energy micro-caps, burn-rate bio-pharma. They're cheap for real reasons that often don't resolve within a year.
Hong Kong: The Cautionary Tale
Hong Kong net-nets performed reasonably well through 2010, the strategy delivered 48.4% in 2003 and 66.5% in 2006. Then the deterioration began.
Year-by-year after 2010:
| Year | HKSE Net-Net | SPY | Excess |
|---|---|---|---|
| 2011 | -34.3% | +8.8% | -43.1% |
| 2012 | +9.5% | +12.5% | -3.0% |
| 2013 | +16.8% | +23.1% | -6.2% |
| 2014 | +7.0% | +11.4% | -4.4% |
| 2015 | -5.1% | +2.7% | -7.9% |
| 2016 | -1.0% | +16.1% | -17.1% |
| 2017 | -6.8% | +11.5% | -18.3% |
| 2018 | -21.5% | +13.2% | -34.7% |
| 2019 | -48.3% | -12.1% | -36.2% |
| 2020 | +22.5% | +65.4% | -42.8% |
| 2021 | -23.2% | +14.6% | -37.7% |
| 2022 | -16.4% | -7.7% | -8.6% |
| 2023 | -17.1% | +28.9% | -46.0% |
| 2024 | -3.9% | +8.8% | -12.7% |
Thirteen years of consistent underperformance. This isn't a data artifact. It reflects real-world events: the 2014-2015 Occupy Central movement, the 2019 protests (months of street-level disruption), the 2020 National Security Law, and ongoing regulatory crackdowns extending into 2023.
Hong Kong net-nets lost money even as underlying companies had positive NCAV. The discount widened rather than closing. When political risk is extreme and persistent, value investing doesn't work, there's no catalyst to close the gap between price and fundamental value.
The -82.1% peak-to-trough drawdown is the starkest number in this dataset.
India: Data Coverage Note
India (NSE-only) shows CAGR of 8.28% over 24 years, with 8 cash periods. But those 8 cash periods were all from 2001-2008.
The FMP data warehouse didn't have sufficient Indian stock coverage before 2009 to reliably compute historical NCAV figures. From 2009 onward, the NSE data is solid.
Effectively, India's track record covers 2009-2024, 16 active years. Over that period the results are: CAGR 8.28%, MaxDD -57.7%, Sharpe 0.042.
Key years: - 2009: +133% (crash recovery. India's net-nets bounced hard) - 2014: +40.9%, 2015: +40.8% (Modi election, India market rally) - 2016: +27.5% (demonetization created distress, then recovery) - 2019: -48.2% (liquidity crunch, shadow banking crisis, pre-COVID) - 2020: +105.8% (COVID recovery) - 2021: +52.7% (India growth optimism) - 2023: +80.9% (India outperformance year)
India net-nets produce episodic outsized returns but with significant volatility. The -57.7% max drawdown reflects how deeply distressed the underlying companies are. For an investor with high risk tolerance and a long horizon, India could work. The Sharpe of 0.042 says you're not being compensated much for the volatility on a risk-adjusted basis.
Note: We ran BSE+NSE combined initially, which creates a methodological problem, many Indian companies are dual-listed on both exchanges. A company like Ganesh Housing appears twice in the portfolio, doubling the position without diversification. The NSE-only results above correct for this.
Korea, Canada, UK
Korea (KSC): 6.60% CAGR, -2.24% excess, Sharpe 0.162. Decent Sharpe relative to the group but still trails SPY. Korean corporate governance has similar issues to Japan (chaebols hoard cash), but the market is more export-dependent and politically sensitive than Japan's domestic net-net universe.
Canada (TSX): 7.49% CAGR, -1.34% excess, Sharpe 0.133. Zero cash periods. Canada produces consistent net-net candidates but the premium is thin. The energy and resource concentration in TSX small-caps creates correlation with commodity cycles.
UK (LSE): -0.62% CAGR, -9.46% excess, Sharpe -0.149. Flat-to-negative returns over 24 years. UK small-caps have faced persistent headwinds: Brexit uncertainty, post-financial-crisis deleveraging, and structural decline in sectors like retail and media that produce the most net-nets.
What to Take From This
The net-net premium documented in academic research from the 1970s-1980s exists in different forms in different markets today:
- Japan: Structural supply from corporate culture + identifiable catalyst from governance reform. Premium is real and risk-adjusted.
- Most markets: Premium has eroded. Stocks are cheap for real reasons. Holding them for a year typically doesn't close the discount.
- Hong Kong: When systematic political risk overrides fundamental value, the strategy fails entirely.
- Crash years: Universally, net-nets bounce hard after market dislocations. The strategy works best as a post-crash recovery play, not as a steady-state strategy.
If you're going to run net-nets, Japan is the data-supported answer. For US, UK, and Hong Kong, the historical evidence no longer supports it.
Excluded Markets
Australia (ASX) and Brazil (SAO): Excluded due to adjusted-close price data quality issues. Historical price splits aren't consistently reflected in the adjClose field, which creates systematic errors in return calculations. Data quality, not strategy performance, drove these exclusions.
China (SHH/SHZ): Average 4 qualifying stocks per year, too thin for a 30-stock portfolio. Results would be driven by idiosyncratic company outcomes, not strategy performance.
Taiwan (TAI/TWO): Borderline ~20 qualifying stocks per year, thin data coverage, and results didn't produce sufficient signal to justify dedicated analysis.
Run the current global net-net screen on Ceta Research Data Explorer.
Data: FMP financial data warehouse via Ceta Research. Annual rebalance April, equal weight, 2001-2024. India (NSE) covers 2009-2024 due to data availability. Backtest code: github.com/ceta-research/backtests
