If AI makes companies more productive, can it predict which stocks I should buy?
Markets are too efficient for individuals to use this strategy
So here’s an interesting question: is there a way to determine how well a publicly traded company is incorporating AI into its workflows? And if there is, is there an investment advantage to be had—alpha, in the argot—from buying those stocks? Someone asked me this, more or less, the other day.
The question arises from the general notion that AI will increase employees’, and therefore companies’, productivity. By itself, this isn’t an objectionable position to take. But I don’t think we can conclude that identifying those companies which are integrating AI into their workflows better than their competitors will yield much of an investment advantage. And the reason that I don’t think this will be a fruitful investment strategy is that the public markets are fairly efficient1. Public markets rapidly aggregate and incorporate information about a company when settling on the price of its stock.
And, to the extent that AI does improve a company’s productivity, that increased productivity will appear in various financial ratios. All of the information contained in those ratios is quickly synthesized by the markets, and prices are reset on a continuous basis as traders2 buy and sell positions in the stock. Stock prices constantly incorporate new information.
In any event, the financial ratios which measure a company’s productivity primarily focus on how efficiently the company uses its assets to generate revenues and profits. Here are some of the ratios commonly used by financial analysts to assess a company’s productivity:
Total Asset Turnover Ratio: This measures how effectively a company uses its assets to generate revenues. Revenues / Total Assets. A higher ratio indicates more efficient uses of assets.
Inventory Turnover Ratio: This assesses how quickly a company sells and replaces its inventory. Cost of Goods Sold (COGS) / average inventory. A higher ratio indicates more efficient inventory management.
Accounts Receivable Turnover Ratio: This measures how efficiently a company collects its receivables. Total Sales / Average Accounts Receivable. A higher ratio implies efficient credit and collection processes.
Fixed Asset Turnover Ratio: This ratio evaluates how well a company uses fixed assets, like machinery and equipment, to generate sales. Net Sales / Net Fixed Assets.
Operating Cycle: This is the amount of time between purchasing inventory and collecting cash from sales. A shorter operating cycle indicates more efficient operations.
Return on Assets (ROA): This measures how profitable a company is relative to its total assets. Net Income / Total Assets. Higher ROA indicates more efficient use of assets to generate profit.
Return on Equity (ROE): This measures how effectively a company uses its equity to generate profits. Net Income / Shareholder Equity.
Labor Productivity Ratios: An example would be Revenue Per Employee. They measure how efficiently labor is used to generate revenue.
The problem with using any one of these measurements as an indication of a company’s AI capabilities is that these ratios can all be calculated from the company’s financial statements and footnotes to those statements. This data is rapidly ingested by algorithmic trading firms and other institutional investors. This means that by the time you assess this information, any AI-influenced investment advantage to be had in a given stock has been arbitraged away by institutional investors who have analysis and trading capabilities that you don’t.
This is beyond the scope of this post, but the notion of efficient markets is commonly over-simplified. There are three ‘forms’ of the efficient markets hypothesis: strong, semi-strong, and weak. Public markets are fairly strongly efficient, or at least semi-strongly efficient. Private markets are much less efficient.
Not necessarily, and not usually, human traders. The majority of trading in public markets is done by software, not by a person using her Robinhood app.