Whoa of Aggregation

This defines entrepreneur and entrepreneurship - the entrepreneur always searches for change, responds to it, and exploits it as an opportunity.

~ Peter F. Drucker, Innovation and Entrepreneurship

Entrepreneurship is a complicated thing, depending on many social, economical, cultural, and personal factors. If personal factors are in the mix, it should be interesting to figure out the correlates (or better yet, the antecedents) of successful entrepreneurship in the domain of personality. I reported earlier (here) on an interesting study by Don Moore and colleagues (2007), which showed that entrepreneurs tend to be myopically optimistic. They enter markets without sufficiently considering that others may do the same, and that only some of them will last. To understand the social consequences, I suggested the game of chicken as a model for market entry. The idea was, as in the work by Moore et al., that entrepreneurial success emerges in a context of interdependence. If too few dare to entreprend, society as a whole will be worse off; if too many entreprend, many will fail.

Trait psychology seeks to predict (or postdict) outcomes of interest from scores on standardized personality inventories. The so-called Big Five taxonomy assumes that individuals can be described sufficiently well by their scores on Extroversion (E), Agreeableness (A), Conscientiousness (C), Neuroticism (N), and Openness to new experiences (O). Recent research suggests that entrepreneurs score high on E, C, and O, and score low on A and N. A new article by Obschonka, Schmitt-Rodermund, Sibereisen, Gosling & Potter (2013) takes this work a step further and shows that within the U. S., within the U.K., and within Germany, aggregated scores representing the distance from the ideal entrepreneurial personality profile are negatively related to indices of entrepreneurial activity (e.g., rate of self-employment in a state, rate of business start-ups). They show an amazing pair of maps, according to which in the U.S., both entrepreneurial personalities and entrepreneurial activity grow stronger as you move west.

Why treat states or provinces as the units of analysis instead of the individual? Obschonka et al. have little to say about that besides noting that others do this too. There is an interest in the geography of psychological characteristics. The idea is that the average person may not be the same in different parts of the country or the world, and that these differences are measurable. To me, this has always rung true, but many social psychologists still believe that any perception of a between-group difference must be evidence of irrational stereotyping. Why should there not be regional differences in entrepreneurial personality?

Obschonka et al. suggest that their work is a contribution to socioecological psychology, an area of research that “focuses on social habitats and the mutual interplay between these social habitats on the one side, and mind and behavior on the other” (p. 105). With the exception of one ancillary analysis, however, their work is not about interplay, but about linear prediction. The authors assume that if there is a correlation between entrepreneurial personality and activity with the person as the unit of analysis, then the same correlation may appear if data are first aggregated within states before correlating them across states. This hypothesis is not a foregone conclusion. It may not be likely, but it is possible that the over-states correlation is opposite in sign to the within-states correlations. This would be a variant of Simpson’s paradox. Anyway, it did not turn out that way.

Correlation usually raises difficult questions about causation because it enables causation without guaranteeing it. Moreover, the direction of causation – if there is any causation – remains unknown. Still, most people are probably inclined to think it more likely that certain individual personality characteristics cause entrepreneurial activity and perhaps success than the inverse. If Orville scores higher on the entrepreneurial personality index than Ralph, and if he also shows more entrepreneurial activity, we cautiously conclude that his personality is the cause of his activity. If, however, both the average entrepreneurial personality score and the average entrepreneurial activity score are higher in Oregon than in Rhode Island, the inference is much harder (again, because the over-persons correlation does not compel the over-states correlation). If personality is to be considered a cause, then it can only operate at the level of the person. In other words, if we have convinced ourselves of the causal effect at the person level, finding a corresponding correlation over traits is at best consistent with this causal idea, but it is not new evidence for this idea to be true. I hasten to point out that the authors do not claim to have shown evidence for causation, which returns us to the question of what exactly the incremental value of their work is.

If trait psychology is the psychology of individual differences, it is surprising to see individual differences aggregated away within states, so that the states become the new individuals. Why stop there? We can aggregate over states and re-examine the question of the (comparative) wealth of nations. Perhaps there are chapters for Wundt’s Völkerpsychologie yet to be written.

Trait psychology is the psychology of, well, traits. It is therefore surprising to see trait scores aggregated over trait dimensions. Obschonka et al. suggest that “entrepreneurship requires a specific constellation or set of traits” (p. 116). This is to say, the entrepreneurial profile over traits is a unique configuration, or perhaps a meta-trait. The authors write that “given that profile-level analyses delivered a consistent picture but that the dimension-level analyses did not, an argument can be made that the regional-level relationship between the profile and entrepreneurial activity is not a simple linear combination between one aspect of personality and the criterion” (p. 116).

This language is ambiguous. I am not sure what “a simple linear combination between one aspect of personality” might mean. One trait can yield a linear relationship with an outcome variable, but linear combination refers to the aggregation over traits. In this sense, the work shows a linear combination of the individual trait scores. The entrepreneurial personality score is computed by summing the differences between the maximum scores on E, C, and O, and the person’s actual scores and the person’s scores on A and N. If we simply compute E + C + O – A – N, we have a composite score that is perfectly and negatively correlated with the entrepreneurial personality index. The index is simply a linear combination of trait scores with equal weights for each trait. As such, the composite score cannot reveal anything beyond the added contributions of each trait. What the composite score does, however, is to hide the contributions of the individual traits. It also leaves open the question of whether some traits contribute more than others, or whether some traits might be unnecessary.

Obschonka et al. take a look at the individual traits and find that they do not perform well or consistently individually. Hence their argument that aggregation is necessary. But which aggregation? Is E+C+O-A-N a uniquely strong predictor of entrepreneurial behavior and success? Presumably, omitting any of the five factors or reversing its sign would significantly degrade the predictive power of the composite. One may ask what it is about adding hostility (disagreeableness) to the otherwise desirable profile that makes the difference between successful entrepreneuring and not even trying or failing.

Personality Essential Reads

Why Do We Have Different Tastes in Music?

How Much Does Your Personality Determine Your Genes?

Moore, D. A., Oesch, J. M., & Zietsma, C. (2007). What competition? Myopic focus in market-entry decision. Organizational Science, 18, 440-454.

Obschonka, M., Schmitt-Rodermund, E., Silbereisen, R. K., Gosling, S. D., & Potter, J. (2013). The regional distribution and correlates of an entrepreneurship-prone personality profile in the United States, Germany, and the United Kingdom: A socioecological perspective. Journal of Personality and Social Psychology, 105, 104-122.

=============

Although this is not a diaretic post like Divano Letto, I want to offer two words of the day.

[1] Dysgraphia is to writing what dyslexia is to reading. I have it, and probably so do you. Proof-read emails and posts to FB or a blog before putting them out there.

[2] The mereological fallacy is a popular defense against reductionism. It is you that is feeling anxious, not your amygdala. Meros is place or part in Greek. Sometimes you need to consider the whole whole to understand a phenomenon or experience.