# The Problem of Induction

Imagine you are a chicken. Each day a farmer comes and feeds you. After a few months of this you conclude that whenever the farmer visits, he will bring you food. All the empirical evidence supports this conclusion. Then, on an otherwise unremarkable day, instead of feeding you the farmer chops your head off.

That is a gruesome introduction to the problem of induction. (Though I have heard the chicken example many times I believe it originated with Bertrand Russell)

I have been noodling around with scientific reasoning and logic as relates to investment due diligence. What grates on me is that I have come to believe much of what people are looking to get out of a due diligence process cannot actually be achieved. For example, when we due diligence an investment manager the emphasis is on proving the manager is skilled. In reality we cannot prove this. At best we can conclude it is highly probable a manager is skilled. Alternatively, we can prove a manager is not skilled, provided we define a measure of “skill” in advance.

The average due diligence process is grounded in inductive reasoning. We make observations about the investment manager, her strategy and her firm. If the observations are favorable, we generalize that the manager is likely to be skilled and will perform well in the future. Logically this process is flawed.

### The Problem of Induction

I first became aware of the problem of induction several years ago via Taleb’s Fooled By Randomness. The issue is we can only use inductive reasoning to conclude something is “likely” or “unlikely.” We cannot use inductive reasoning to prove something is true or false.

The classic example is the black swan. For a long time people believed all swans were white. They did not know all swans were white (they would have to have observed all the swans in existence to prove this). Rather, people inferred an extremely high probability for all swans being white because all the swans observed to date had been white. Then, in 1697, Willem de Vlamingh discovered cygnus atratus in Australia.

In the context of investing we struggle with the naïve extrapolation of past performance into the future. On the basis of past performance I can say, “I believe it is probable this investment manager is highly skilled.” However, I cannot use that data to prove, with certainty, that the manager will continue to outperform in the future. This WSJ article hit piece discussing Morningstar ratings is a practical exploration of the issue (although for the record I believe the WSJ badly misrepresented what Morningstar is trying to achieve with its ratings).

The problem of induction is central to the validity of the scientific method. Science does not prove the truth of hypotheses, theories and laws. It merely verifies they are consistent with empirical results. However, as with inferences about the colors of swans, it only takes one false case to disprove a scientific theory. The philosopher Karl Popper therefore concluded falsifiability is the essential criteria determining whether a theory can be considered scientific.

From Wikipedia:

Among his contributions to philosophy is his claim to have solved the philosophical problem of induction. He states that while there is no way to prove that the sun will rise, it is possible to formulate the theory that every day the sun will rise; if it does not rise on some particular day, the theory will be falsified and will have to be replaced by a different one. Until that day, there is no need to reject the assumption that the theory is true. Nor is it rational according to Popper to make instead the more complex assumption that the sun will rise until a given day, but will stop doing so the day after, or similar statements with additional conditions.

Such a theory would be true with higher probability, because it cannot be attacked so easily: to falsify the first one, it is sufficient to find that the sun has stopped rising; to falsify the second one, one additionally needs the assumption that the given day has not yet been reached. Popper held that it is the least likely, or most easily falsifiable, or simplest theory (attributes which he identified as all the same thing) that explains known facts that one should rationally prefer. His opposition to positivism, which held that it is the theory most likely to be true that one should prefer, here becomes very apparent. It is impossible, Popper argues, to ensure a theory to be true; it is more important that its falsity can be detected as easily as possible.

### Applications To Due Diligence & Investment Analysis

This means you cannot “prove” an investment thesis is correct. At best you can gather evidence to build conviction that your investment thesis is “probably” correct. In my experience much due diligence is conducted with an inductive mindset. This leaves due diligence processes vulnerable to confirmation bias.

Should we invert the process?

In other words, you would organize due diligence with the goal of falsifying an investment thesis. If the thesis cannot be falsified, you invest. As a risk management discipline, you then establish a series of easily falsifiable statements constituting “thesis breaks” (e.g. “Company A will average double-digit revenue growth over the next 3 years”). When a thesis break is triggered, the investment is re-evaluated or removed from the portfolio.

### Practical Considerations

At a high level, evaluating an investment opportunity can almost always be boiled down to the following:

People: Management has integrity and is aligned with investors.

Process: Processes are disciplined, repeatable and based on sound economic principles.

Performance: Past performance supports management’s ability to execute.

The due diligence process should not be structured to verify these statements as accurate. It should be structured to prove they are false. In practice to guide your work you would need to establish a whole series of falsifiable statements underneath these broad headings. For example, under People:

• Management has never committed or been associated with securities-related offenses.
• Management has no prior record of personal or business bankruptcy.
• Management has never been convicted of a felony or misdemeanor offense.
• Management owns >10% of shares outstanding / maintains significant personal investment.

In my view this is a more straightforward, disciplined and logically sound method of organizing a due diligence process. To a non-practitioner the distinction may seem silly. However, the structure is designed to minimize confirmation bias—a common and dangerous cognitive bias in investment research and portfolio management.