Practical Reasoning

 

 

When laboring in the vineyards of Wall Street, we discovered something about bank credits. The approval of every loan quickly hinged on the answering of three main questions, but and this is crucial – they were never the same three questions, even for large companies.

 

What was going on here? The answer to this most practical question is easy. Unlike the stationary, statistical world that economic modelers love, the business world is complex and ever changing. That requires a different sort of reasoning.

 

Practical philosophers and practitioners do not like the simple world posited by modelers, a product of the first stage of the 17^th century Enlightenment, which believed in the theoretical calculability of everything. But Isaiah Berlin wrote:

 

“If I am a statesman faced with an agonising choice of possible courses of action in a critical situation, will I really find it useful – even if I can afford to wait that long for the answer – to employ a team of specialists in political science to assemble for me from past history all kinds of cases analogous to my situation, from which I or they must then abstract what these cases have in common, deriving from this exercise relevant laws of human behaviour?...the retention of that which is common, would produce a very thin, generalized residue, and one far too unspecific to be of much help in a practical dilemma….Obviously what matters is to understand a particular situation in its full uniqueness…” ¹

 

What kind of reasoning appreciates a situation in its, “full uniqueness”?

 

Mervyn King is a trained economist and the former Governor of the Bank of England. He has co-written a book “Radical Uncertainty (2020).” It is the kind of book that sometimes requires a day per page to consider, but in summary seems perfectly obvious. It’s a great book.

 

The central problem that he addresses is this: the basic uncertainty of the world. How consumers will react; how investors will react; how international markets will react. “Radical uncertainty precludes optimising behaviour. In the world as it is, we cope rather than optimize. The numbers…are derived from historic data series and assume a non-existent stationarity of the world. Struggling to cope with a large world which they could only imperfectly understand, the proponents of these calculations invented a small world which gave them the satisfaction of clear-cut answers. And financial regulators claiming to monitor risk in the financial system did the same. It is understandable that people who are given a job which they cannot do find instead a more limited task which they can do.” ²

 

How do people handle a lack of knowledge? There are three reasoning methods to understand the world. Deductive reasoning is from first principles, top-down, and is relevant only in modeled worlds. Inductive reasoning is from the data, bottom-up, but the Gaussian generalizations won’t hold if the world (and the data) changes. The third form of reasoning has a name, its abductive reasoning, where events are one of a kind. It is the form of reasoning used in specific cases: in law, business, and in life.

 

“We are constantly engaged in abductive reasoning, using our knowledge and experience to make sense of complex situations…” ³ The goal of this form of reasoning is to arrive at a best explanation, “What is going on here?” ⁴

 

Consider the legal system. “Because justice is administered not on average but in individual cases, bare statistical evidence in the absence of a narrative, is never enough….We need a story. Narratives are the means by which humans – as judges, jurors, or people conducting the ordinary business of life – order our thoughts and make sense of the evidence given to us. The legal style of reasoning, essentially abductive, involves a search for the ‘best explanation’- a persuasive narrative account of events relative to the case…. In civil proceedings, the narrative must be a good one…. In criminal proceedings, the narrative must be sufficiently compelling that no materially different account of events could be seriously entertained – the prosecution case is established beyond reasonable doubt. A ‘good’ explanation meets the twin criteria of credibility and coherence. It is consistent with (most of) the available evidence and the general knowledge available to judges and jurors…Statistical reasoning has its place. But only when integrated into an overall narrative or best explanation.” ⁵ Abductive reasoning applies, as well, to the case method in MBA programs.

 

Law is a systematic attempt to order human affairs, which are otherwise disorderly. Stories may be many; but they should reflect the facts, in our system. In law or in investing, the facts that abide ultimately determine the real story. Many things we learned in school were modeled facts, facts that are useful only if they fit well into convincing stories of real people, companies, or the economy.

 

Harry Markowitz was the originator of Modern Portfolio Theory. “Markowitz himself, asked about his own portfolio decisions in planning for retirement responded, ‘I should have computed the historical co-variances of the asset classes and drawn an efficient frontier. Instead, I visualized my grief if the stock market went way up, and I wasn’t in it. (At the present time, 11/24, we do not feel any grief about this.) – or if it went way down and I was completely in it. My intention was to minimize my future regret. So, I split my contributions 50/50 between bonds and equities.’ Markowitz’s description of his own behaviour corresponded to…‘risk as feelings’ – his decisions reflecting the hopes and fears he held in anticipation, rather than the maximization of subjective expected utility implied in his own models.” ⁶

 

Mervyn King ends his book with this thought, “We see through a glass, darkly. And we communicate with each other through narratives, not probabilities, to describe our endlessly fascinating world.” ⁷

 

__

 

 

For investors, feelings about their portfolios are very important. But, given the priorities of the incoming Administration, inflation could be a problem in the U.S. The following discusses how both the general OLS (Gaussian) model and generative AI might be useful additions to the abductive reasoning toolkit, to determine how inflation might propagate.

 

In a competitive industry, if one company raises prices it will lose market share and its return on capital will drop. In a smoke-filled room, if a group of companies agree to raise prices, that is illegal collusion. But what if companies in concentrated price-taking markets raise prices all together due to a common shock: climate change, trade wars, supply chain disruptions? The economist Isabella Weber of University of Massachusetts, examined 138,962 earnings call transcripts for 4,823 stock market traded companies. She concluded that large input cost increases provide executives with, “the opportunity to raise prices and protect or even increase profits.” The counter to this requires increased public investment in co-ordination, port capacity and buffer stocks. The paper is titled, “Implicit Coordination in Sellers’ Inflation: How Cost Shocks Facilitate Price Hikes.”

 

We think this is an important, if somewhat complicated, paper because more supply shocks will likely occur in the future, possibly creating greater inflation. Current Keynesian monetary tools affect mainly demand.

 

What is the methodology that this paper, which has not yet been peer-reviewed, uses to arrive at the conclusion that common shocks allow companies to raise prices?

 

1) This conclusion makes intuitive sense.

2) The researchers are from reputable institutions: University of Massachusetts, London School of Economics, The University of Heidelberg.

3) The paper is extraordinarily detailed, but the methodology also includes the extensive use of Large Language Models and ChatGPT to construct and validate indexes. The results of this paper could then fall into some sort of AI lacuna. The problem is results of generative AI models cannot be traced to causes. But the fact that there were also human coders to handle exceptions in the construction of a cost increase sentiment index made the study credible to us. The key question was, how were large cost increases received?

 

It was important to us that there would not be a conclusion, “Because generative AI says so.” So, to proceed:

 

            The authors then create a Sentiment Index = f( S(c) ), to make a long story short. ⁸

 

            Where S(c) = +1 if c is an element of S+

                        S(c) = -1 if c is an element of S-

                           0 otherwise

 

 Here, S+ and S- represent lists of positive and negative-toned words. For example, if the computer encounters, as a short example, the words: “challenged”, “concern”, “pressured” these words reflect a negative sentiment. If it encounters the words: “achieve”, “better”, “confident” these words reflect a positive sentiment to a company’s pricing power.

 

The authors then use OLS regression to regress the Sentiment Index upon the annual change in the Bureau of Economic Analysis Price Index (x). ⁹ The result, we take the simple case where the authors use only one index:

 

Sentiment  Index/      Values for the model y=cx + b

                                   Coefficient, c       0.185*** (very significant)

                                   Constant, b         -0.433***

 

Observations              62

R²                                                0.679 (Which improves if you add more indexes; .66 is the lowest

                                                  model threshold for the social sciences.)

 

 The authors then make a point that caused us to further examine OLS assumptions. “To be clear, our regression framework does not aim to capture causal relations. Instead, our results demonstrate that corporate executives tend to discuss increases in costs more positively in the presence of large economy-wide cost shocks and their co-occurence with supply constraints. Our index validation and qualitative analysis affirm that this positive sentiment is often indicative of a firm’s perceived ability to pass on their increases in input costs. Conversely, negative sentiment is indicative of a firm’s perceived compulsion to absorb firm-specific cost increases. In this context we interpret our regression results as descriptive evidence suggestive that large cost shocks can function as an implicit coordinating mechanism.”

 

This is a very careful study, also considering section 4.2.2 where the authors do index scatter plots to further validate the indexes constructed. Why the above disclaimer and the scatter plots? It is because they run OLS against price data automatically containing very large variances. And therefore, the best model interpretation is correlation rather than cause.¹⁰ The OLS model is derived using Gaussian (normal) data, that does not usually exist in empirical economics, in a non-placid world. This study could be reviewed by well-trained statisticians.

 

The authors then examine the qualitative aspects of price sentiment, going into a few actual cases, using, as we noted above, abductive reasoning. To quote a Hilton Worldwide Holdings, Inc. executive in October, 2021:

 

     “With input costs going up, labor costs going up and all of those

        fun things, they’re going to be – margins are going to be higher

        ultimately when we get past this for all the reasons I talked about

        in terms of the pricing power that we have and the broader inflationary

        environment. That’s very helpful to the business.

 

We have discussed specific abductive reasoning. The OLS quantitative analysis then allows a further generalization from the data. But trying to generalize about people, without specific people, is problematic. The use of both forms of reason, the significant specific and the general, can lead to balanced and realistic policies.

 

 

 

Footnotes

 

 

 

RETURN TO HOME PAGE