Thursday, September 22, 2016

Forgetfulness and financial analysis – Is more memory always better?



Lest we forget - Lest we forget
-Rudyard Kipling

Big historical events, especially tragedies, are committed to memory so we will not forget, yet is it really good to remember everything? Put differently, is forgetfulness useful? Would we be better off if some memories disappeared?

Now a quant would say that more data is always better than less data. Some quants may use a fixed set of years for data analysis and thus cut memory before the start date, but the general rule of thumb is that more data are always better than less.  Similarly, the behavioral finance professor will warn against recency biases with our thinking, and will also comment on vividness and recall biases. More data may be better, but be careful of your memories. 

Daniel Kahneman, the decision theorist and Nobel Prize winner, has talked about humans having two selves, the "experiencing self" and "remembering self". There is a difference from what we experience versus what we will remember. We will make decisions based on our memories and not our experiences. The memory will focus on the results of the experience, so an event that had a negative outcome becomes negative. If the same experience has a positive conclusion, the memory will be positive. 

Events, or more importantly the memories of those events, will skew thinking. Those who went through the '87 crash will act differently than newer traders. Those who remember the tech bubble and lost money may avoid that sector to their detriment. We may want to forget in order to be unbiased. Our mind generates biases that should be forgotten.

More memory is not always better than less. Being forgetful about old regimes and old relationships may actually be a positive. For example, in the long-run monetary policy will have well-defined impact on an economy, but currently a QE regime is different than an inflation targeting environment. Forgetting some of the past lessons or weighting them differently is a positive.

Data analysis will not generate a difference between memory and experience not unless it is programmed. All data are equal weighted unless an explicit weighing function is used. Memory is only lost if a start-date cut-off is employed. We can control our memories and forgetfulness through quant analysis.


More memory is not always better than less. Being forgetful about old regimes and old relationships may actually be a positive. For example, in the long-run monetary policy will have well-defined impact on an economy, but currently a QE regime is different than an inflation targeting environment. Forgetting some of the past lessons or weighting them differently is a positive. Quant work can generate controlled forgetfulness.

No comments: