A new study just published in the academic journal Economic Development Quarterly looks at the links between big (budget over $2 million) performing arts organizations and the change in the proportion of the metro workforce that is in Richard Florida’s definition of the ‘creative class’. The article, by Arthur Nelson and colleagues is (behind a paywall) here, and Florida discusses here. Artsjournal.com links to it with the clickbait “The study finds substantial evidence that performing arts organizations add to both the growth of the knowledge class and to urban economies broadly.” Let’s take things down a touch.
Let me note, as someone who has tried to do this kind of estimation, that empirical work in this field is challenging. We don’t have the ability to perform experiments – “let’s choose some cities at random and assign them with full symphony orchestras and ballet companies, and see what happens to regional economic growth” – and so we have to work with the historical data we have, making choices on how to measure variables of interest, what to include in the analysis and what to leave out, and how to specify the regression models. I’m glad teams of researchers are making attempts to look hard at arts organizations and regional economies, and it is uncharitable to carp about whether a different researcher would have chosen these variables and estimation methods rather than those.
That said, let’s look at the study itself. First thing to remember, and I know this sounds tedious and banal, but correlation is not the same as causation. Artsjournal and Richard Florida might hope that “the study finds substantial evidence that performing arts organizations add to both the growth of the knowledge class and to urban economies broadly,” but it doesn’t. The authors are more modest, asking “Is there an association between growth in the share of knowledge class employment and the presence of professional performing arts?” [p. 4, my emphasis] and in the conclusion noting “We further recommend that future research test for causality. In this respect, our current work is the platform for future work.” [p. 7, my emphasis again]. Metro areas that had more large performing arts organizations (restricted in this study to orchestras, opera, and ballet/dance companies) tended also to have, during the first decade of this century, a larger increase in the share of employment that are in Florida’s definition of the ‘creative class’, some other variables held equal. Whether A caused B, B caused A, or A and B occurred because of some other cause C, is not known. The authors of the study know this, we should not fall to inflating their limited claims.
I think the study’s authors do go off track when they try to nail down monetary estimates of the effect of arts organizations on local incomes, extrapolating from the growth of the share of employment in the creative class (n.b. the share, not the absolute number – Florida’s CityLab commentary seems to get this mistaken), to the mean salaries of creative class workers, to some economic impact multipliers (unforgivable), and in the end total income growth. This serves to undercut the more nuanced part of the study that recognizes the lack of knowledge regarding causality.
I start to worry about the focus on the share of employment in the creative class. I’ve posted twice on this topic in the past few weeks (here and here) and won’t repeat those points, but I will raise one more general one: why focus on that one kind of employment? It is about a third (or so) of the US workforce, and for cost disease reasons is bound to increase as a share of total employment, relative to jobs in resources, manufacturing, transportation, etc as those sectors take advantage of possibilities in capital-labor substitution. There is no reason to make it a key metric of successful regional economic development. For that we look to indicators of quality of life, and how equally they are made available to all residents. We have lots of metrics for those things, so why focus on this metric of what proportion of the workforce falls under a collection of job classifications from a 2002 popular book?
Goodhart’s Law, in its most general form, holds that “When a measure becomes a target, it ceases to be a good measure.” Anyone who has worked with performance metrics quickly learns this – as soon as you begin to strive for good results on a ‘key performance indicator’ it ceases to become a useful performance indicator, since your efforts are no longer directed towards the truly important outcome, but rather to its specific measure. Historically, it might well be that ‘share of total employment that is creative class’ has been associated with a variety of good economic outcomes, but if local governments adopt “% creative class” as a target, it might well be that good policies for improving the local economy, and prudence in investment, is lost, in pursuit of doing well in the creative class league tables.
Be careful out there.