The role of bipolar disorder and family wealth in choosing creative professions

Data and example

We use register data on the population of Denmark, obtained from Statistics Denmark. Our data includes mental health diagnoses and occupations for 2,524,325 people born between 1946 and 1975. Family identifiers, which we use to match people to their siblings and measure differences in parental wealth, are available for 71 percent of the population. Appendix Table A1 summarizes the variables used in our analysis. Although we are not allowed to share the data ourselves, researchers can access it through an application from Statistics Denmark. (https://sundhedsdatastyrelsen.dk/da/english/health_data_and_registers/research_services/apply/data_statistics_dk).

Mental health diagnoses

Information about diagnoses comes from the Central Psychiatric Register (National Patient Register for Psykiatri Diagnoser), which records all mental health diagnoses in Denmark between January 1, 1995 and December 31, 2015. The registry classifies mental disorders according to the World Health Organization’s International Statistical Classification of Diseases and Related Health Problems (ICD-10; see http://apps.who.int/classifications/icd10/browse/2016/en#/F30-F39).

Implementation of this classification, our variable BD identifies 18,729 people who have received at least one diagnosis of bipolar disorder (ICD-10: F31) or mania (ICD-10: F30). BD is defined as “A disorder characterized by (…) some cases of elevation of mood and increased energy and activity (hypomania or mania) and in other cases of depression of mood and decreased energy and activity (depression) .” Mania is described as “A disorder (…) ranging from carefree joviality to almost uncontrollable excitement, (…) accompanied by increased energy, resulting in overactivity, speech strain and a decreased need for sleep.”

Creative professions

In this analysis, we measure a person’s propensity for creativity by working in creative professions. Existing economic analyzes have measured creativity by outputs, such as patents^8.9publications³⁶and musical compositions³⁷. In contrast, studies of mental disorders and creativity in psychology and medicine have treated creativity as an individual-level characteristic, reflected by a person’s career choice.^1,2,19.

We follow the psychology literature in defining creative professions. Previous studies have classified creative professions as designers, writers, academics, visual artists, architects, exhibition artists, performing artists, composers and musicians.¹⁹. Others have excluded architects, but photographers have¹. We include both architects and photographers as creative professions and report the results separately per profession.

The Danish registers follow the International Standard Classification of Occupations (ISCO) to classify occupations. The years 1995–2009 use the 1988 classification and the years 2010–2015 use the 2008 classification. Using 4-digit ISCO codes we distinguish academics (ISCO code 2310), photographers (3131), visual artists (2452 in 1988; 2651 and 2166 in 2008), designers (3471 in 1988; 3432, 3435, 2163 and 2166 in 2008). ), performers (2454 and 2455 in 1988; 2654 and 2655 in 2008), composers and musicians (2453 in 1988, 2652 in 2008), writers (2451 in 1988; 2431, 2432, 2641 and 2642 in 2008), and architects (2141 in 1998; 2161 and 2162 in 2008; appendix table A2). In multinomial logit regressions, we aggregate ISCO codes down to the three-digit level to reduce the number of choices.

Family identification information and parental assets

To match people with their siblings, we use their mother or father’s Social Security number as family identification. Family identifiers are available for 1,788,166 people (71 percent of the population); 75 percent of them have one or more brothers and sisters. Family identifiers allow us to identify siblings of people with BD.

Parental wealth data is available for people whose mother or father declared assets for at least one year between 1980 and 2015. For people whose parents are on the list but have no financial assets, we set the assets to zero. Assets are reported by banks and other financial institutions and not by the individuals themselves. All results are robust enough to exclude individuals without parental wealth information from the analyses. To determine an individual’s position in the distribution of parental wealth, we calculate the percentile of parental wealth for each year (from 1980 to 2015) and assign each individual to their parents’ median percentile across all years.

Empirical framework

BD and creative employment

First, we test whether people with BD are more likely to work in creative professions than the general population. To prevent differences in labor participation between people with BD and the population, we limit attention to people with a positive income in a given year. We estimate the following equation separately for eight creative professions, including writers, academics, architects, designers, musicians, photographers, visual artists, and performing artists:

where the variable creatively_It equals one if person i has worked in one year in one or a specific creative profession TAnd BD_i equals one if the person has been diagnosed with BD at least once. An indicator for women F_i controls for possible gender differences in occupational choices. A vector of cohort fixed effects θ_c(i) controls for systematic differences in the propensity to hold creative jobs across cohorts. A vector of year fixed effects τ_T controls for differences in the same tendency over time. We cluster standard errors at the individual level. The coefficient β estimates the difference in the likelihood of having a creative job between people with BD and the population, controlling for variation in creative employment across calendar years, birth cohorts, and gender.

Creative employment among siblings of people with BD

If BD is associated with creativity through a genetic link, siblings may have a milder ‘subthreshold’ form of BD, allowing them to be more creative, without experiencing debilitating symptoms. Models of BD in molecular neuropsychiatry have proposed an inverted U-shaped relationship between genetic risk for BD and creativity²⁵. Suspicions that “some aspects of the bipolar spectrum may confer benefits, while more severe manifestations of symptoms negatively impact creative performance.”

To test whether healthy siblings of people with BD are more likely to pursue creative careers, we estimated Eq. (1) with an indicator for BD siblings instead of the indicator for BD:

In this modified equation, the coefficient is β_S estimates the difference in the likelihood of having a creative job between siblings of people with BD and other people of the same sex, in the same birth cohort and calendar year.

Most common occupations for people with BD: multinomial logit

Looking beyond creative professions, we explore what types of work people with BD are most likely to pursue. Formally, we estimate multinomial models of occupational choice, using the broader 3-digit ISCO08 codes to classify occupations. To define occupations consistently over time, we limit the analysis to 2010–2015, when ISCO08 codes are available. Following the psychology literature, we use four-digit codes to examine employment in creative occupations across key specifications. To estimate multinomial choice models of an individual’s choice across occupations, we use 3-digit ISCO codes. (A previous study estimates a multinomial model of a choice between five occupations, using data from a total of 20,861 interviews in five university cities. They find that people with BD and mania are most likely to work in the service sector and exhibit, using a measure of creativity, that service provision is a creative activity ²⁸.)We model the probability that a person is employed J on time T as:

Where Y_It is the profession of person i in year T And F_i is an indicator for women. First we estimate β_J And γ_J for 129 occupations via maximum likelihood. We normalize both parameters to zero for “primary schools and early childhood teachers” (ISCO08 code 234), the most common occupation (with 6.8 percent of workers). We then calculate the excess probability of occupancy J for people with BD as exp(β_J)−1.