I was curious to better understand how Japanese Net-Nets have performed over the recent period, so I decided to do some work on it myself.

Of course, plenty of academic work suggests that implementing a strategy of buying Net-Nets has positive risk-adjusted returns. However, it is difficult to get a clear picture from the academic work as each study is nuanced by various time periods, different markets, and investment criteria. Also, there are some mental potholes that one can easily fall into as academic work likes to use mean returns and monthly mean returns. As Gaurang Merani has pointed out, using mean returns to get to geometric returns can be a misleading journey, leading to precise calculations that are precisely wrong.

Another issue is that in many academic works, it is unclear, at least to me, if dividends were included in the return calculations. For example, multiple studies evaluate performance based on dividend versus non-dividend-paying stocks. Still, it is unclear if the dividends from the dividend-paying stocks were included in the calculated returns.

Anecdotally, folks like Nate Tobik have stated the Japanese Net-Nets have worked out very well in the past, while Monish Pabrai states (albeit hearsay) that his venture into Japanese Net-Nets in 2012 was a mistake.

In the end, I decided it was worth the effort to size things up for myself.

To get there, I first needed to pull data from LSEG. LSEG supports point-in-time data so there is reasonable assurance that I’m avoiding survivorship bias and look-ahead bias. I also decided to screen for companies trading at or below their net current asset value, as opposed to the traditional 67% of NCAV, because I wanted a larger sample size. (Spoiler alert: it had no real effect on performance.)

I used the traditional equation: NCAV = (Current Assets – (Total Liabilities + Preferred Stock) ) ÷ (Closing Stock Price * Shares Outstanding)

The period of study is April 20, 2010, to April 20, 2024. This is a somewhat arbitrary date. The length is 14 years, which is reasonably long. I could have gone back further but decided against it. I was also curious to focus on and around the time of the start of Abenomics in 2012 and the collective push towards more efficient use of capital since then.

The screen generated 760 stocks over that period in reasonably sized buckets, aside from the extreme low in 2018 of only 7 stocks. The high was in 2020, with 100 stocks. Aside from 2018, the portfolio would consist of at least 20 names.

The portfolio is equal-weighted and the number of positions varied each year. For example, in 2010 the portfolio held 60 positions, equal-weighted, and in 2013 the portfolio consisted of 95 holdings, equal-weighted.

On April 20th of each year (or thereabouts depending on it being a trading day), the portfolio was rebalanced as of the end-of-day closing price. I did not make any adjustments for transaction fees or slippage and assumed all orders were filled.

I also did not take into account the available volume. However, Japanese Net-Nets typically have material market caps, unlike in America where you can easily find a Net-Net with a market cap of $100K with tiny volume. The lowest market cap stock purchased was GFA Co Ltd (8783.T) in 2011 with a market cap, at the time of purchase, of $3.2 million (USD). So it’s reasonable to assume, with a portfolio of $1 million (USD), that all of the stocks could have been purchased in the study.

Although these sectors are typically left out of studies, I decided to include Financials and Real Estate mainly due to curiosity.

The returns are priced and calculated in yen and do not take into account currency fluctuations. Dividend payments were included in the returns of the Net-Net portfolio. I used the price return of the MSCI Japan ETF (EWJ) as a relative proxy — dividends were not included.

The results are somewhat to be expected. The Net-Net portfolio was up over 12x during that period while the EWJ was up only 1.6x.

Below is a table of cumulative and geometric mean returns over the 14 year period.

	Japanese Net-Nets at or below 1x NCAV	Japanese Net-Nets at or below 0.67x NCAV	MSCI Japan ETF (Yen)
Cumulative Return	1,247.32%	1,275.52%	160.04%
Geometric Mean Return	19.75%	19.94%	3.41%

Some of the academic work that touts +30% annualized returns I find to be a bit suspect, given my experience. But these results are more in line with what I was expecting. Strangely, because these results are from another country, it does fit Ben Graham’s comments that a Net-Net strategy does about 20% a year.

What about sectors? It is difficult to compile a horserace based on sectors because of the lack of adequate representation of names per sector. For example, Health Care only had one name during the entire period and was down -14%. (If you held that name for two years, however, you would have been up 51%.) Only Consumer Discretionary, Consumer Staples, and Information Technology had at least one stock in every portfolio vintage.

In brief, Financials, Industrials, and Real Estate had above-average returns, while Consumer Staples and Consumer Discretionary had below-average returns.

What about those big winners? Below is a boxplot by sector showing the 25th and 75th percentiles, the median (listed numerically in bold), and the outliers. Based on median returns, dispersion, and the opportunity to catch some high-fliers, Industrials, IT, Financials, and Real Estate are interesting to look into further.

As Jim O’Shaughnessy would say, the results here are more likely the ceiling rather than the floor.