Weathering Volatility 2.0

From the entire universe of nearly forty million Chase deposit customers, we assembled six million anonymized families to form a 75-month balanced panel (October 2012 to December 2018), for whom we have detailed transaction-level information on income, spending, and account balances (checking and savings) held at Chase. Our unit of analysis is the primary account holder, which we refer to as a “family.”

To be included in our sample, an account holder must have: 1. At least five transactions (inflows or outflows) from a personal checking account in every month between October 2012 and December 2018. This attempts to ensure the Chase account observed is the account holder’s active bank account. 2. At least $400 in average monthly total income for every twelve-month rolling period. This serves to filter for account holders whose income is likely landing at the Chase account observed. 3. At least $10 in average spending, and at least $1 spent every month. This attempts to ensure we see spending activity for a given account.

Incomes we observe are take-home incomes, meaning after taxes and payroll deductions. Income categories we construct in our data set include labor income (i.e. payroll and other direct deposits) and non-labor income (i.e. government income, capital income, and otherwise).

Line graph showing the median coefficient of variation (CV) for total income between 2013 and 2018. Coefficient of variation (CV) is our measure of month-to-month volatility. CV measures the dispersion of a family’s income in a given month relative to the mean income of the prior twelve-months, including the month measured. The median CV remains constant, around 0.38 between 2013 and 2018.

Bar graph showing the distribution of coefficient of variation for total income. Families vary significantly in levels of income volatility they experience. The distribution of CV has a standard deviation of 0.37 with a long right tail. About eight percent of family-months have a CV above 1.0, which would correspond to a more than 60 percent change in monthly income within a year.

Two scatterplots measuring the magnitude of income spikes and dips as a percent change from the median monthly income during the prior year, and the frequency of income spikes and dips as a percent change from the median monthly income during the prior year. Families whose coefficient of variation (CV) fall in the middle bin of the CV distribution experience spikes that are 51 percent above and 56 percent below their baseline income. As families’ incomes become more volatile, the frequency of income swings they experience flattens out but not the magnitude. Beyond a CV of 1.0, families’ frequency of income swings plateaus. The magnitude of income swings, however does not plateau as families’ incomes become more volatile. As CV increases, the magnitude of income spikes and dips continues to increase, especially that of spikes.

Two dot plots examining the heterogeneity of income volatility n terms of the average number of income spikes and dips by take-home income quintile. Higher-income families experience more frequent income swings. Lower-income families experience the largest dips when dips happen.

Bar graph examining heterogeneity of median spending and income coefficient of variation by demographic groups. We show demographic groups by: age bins of 18-24, 25-34, 35-44, 45-54, 55-64, 65-74, 75+; five income quintiles, gender, and five quintiles of cash buffer month. Spending volatility is lower than income volatility, except among account holders above age 75 and those with the largest cash buffers.