The best investments of every year, in hindsight

How the ranking works

A window starts at the end of the year before the one you pick and lasts the number of years you choose: “2009, 10 years” runs from the end of 2008 to the end of 2018. Everything is ranked by total return: the price change plus dividends reinvested, as if $1,000 was put in once. The multiple (×) is what each dollar became; the rate a year is the steady growth that gives the same result.

Stocks are the 500 companies in today's S&P 500 (503 tickers: two share classes of one company count once). Their returns come from the companies' own filings with the SEC: every year a company must show a graph of $100 invested five years earlier, dividends reinvested, and most print its numbers, which this site chains into one record per company. The records run by fiscal year, so a company whose year ends in June or September is measured from that month. They reach 2025 for most companies; the stocks tool starts in 1992, the first year with at least 30 records.

Assets are the 23 of the portfolio builder, December to December: the S&P 500 and 10-year Treasuries with their income, cash at the federal funds rate, cash under the mattress, gold, silver, platinum, US home prices and 15 cryptocurrencies, each from its first price. That tool shows the same numbers month by month.

This is hindsight. Nobody knew in 2009 which companies would lead, and a list of winners picked afterwards says nothing about which ones will lead next.

Famous windows

After the 2008 crash. From the start of 2009, held 10 years, the leaders of the 161 ranked companies were Booking Holdings (×23.4), TransDigm Group (×22) and Regeneron Pharmaceuticals (×20.3). The median company grew ×3.95 and the S&P 500 with dividends ×3.6; 61% of the companies beat it over the same months. Open this window.

From the dot-com peak. Put in at the start of 2000 and held 20 years, the S&P 500 grew ×3.24, +6.1% a year. The best of 99 ranked companies did far better: Old Dominion ×135, Tractor Supply ×99.5, Idexx Laboratories ×64.3; 2 of them ended below the money put in. Open this window.

The best single year in these records was Axon Enterprise in 2003: $1,000 became $20,400. The worst was Las Vegas Sands in 2008, down 94%. See 2003.

The same top 10 rarely repeats

Move the start by one year and the one-year top 10 is almost new: on average only 1.4 of its 10 names were also in the previous year's, and in 9 of 33 pairs of years none were. The 34 one-year windows had 26 different winners, and no company came first more than twice (8 did it twice).

Longer windows look steadier only because neighbours share most of their years: two 10-year windows a year apart have 9 years in common, and on average 6.4 of the 10 names stay. The 25 ten-year windows had 12 different winners. In the grid, each cell is the winner of one start year (across) and holding period (down).

How many of the top 10 stay

The bars count, for every start year, how many names of the previous start year's top 10 are in its top 10 again. On average 1.4 of 10 stay for one-year holding, 4.8 for five years and 6.4 for ten: the longer the window, the more years two neighbours share.

Which asset won each decade

Among the assets, the winner changes with the era: S&P 500 in the 1960s, Gold in the 1970s, S&P 500 in the 1980s, S&P 500 in the 1990s, Gold in the 2000s, S&P 500 in the 2010s and BNB in the 2020s. The table puts each decade's winner next to the S&P 500 and the weakest asset of the same years.

Survivorship bias, plainly

The stocks are the companies in the S&P 500 today. To be there, a company had to survive and grow. Those that went bankrupt, were sold cheaply or shrank out of the index are not in the data, so every list of best buys here looks better than the choice an investor really had in 2009, and the median company flatters the typical stock of that time.

The further back, the fewer companies have a record at all. At the start of 2000, 99 of today's 500 companies (20%) have one; 229 were already filing annual reports, but their filings give no numbers that far back (the largest: Nvidia, Apple and Microsoft); 172 had not filed an annual report on EDGAR yet, mostly because they were not listed, and EDGAR itself begins in 1993–1996.

That is why some of the largest companies, such as Microsoft, Amazon and Alphabet, are missing from earlier windows: they are ranked only from the year their filings start to give numbers. Missing records and missing failures pull in opposite directions, and neither is random, so read every list as “of the companies that can be measured”.