Micro to Macro
In the July issue of FUSE, we examined how household socioeconomic characteristics affect spending in different consumption categories. We found that consumption categories had clear relationships with variables like age and income, enabling us to associate spending level and growth rates with consumer ages and income levels.
In this issue, we take a macro perspective. We look at how consumer categories behave in different economic environments. We also consider the influence of habit formation and income uncertainty. As longer-term investors, stable, through-the-cycle revenues and earnings are especially attractive to us. This revenue stems from consumer spending, so identifying stable and resilient spending trends throughout the cycle is a persuasive component of our research.
Consumption – A Primer
Elasticity is the measure of a variable’s sensitivity to change in another variable. Elasticities are usually measured for the change in a given consumption category (e.g. automobiles) relative to a change in its price (own-price elasticity), a similar category’s price (cross-price elasticity), or income (income elasticity).
Habit Formation is a process where behavior becomes habitual through repetition. Habit formation implies that consumers’ current utility from consumption is determined by individual consumption relative to a baseline reference level. Consumers form expenditure pattern habits based on both past consumption and response to shocks.
The Permanent Income Hypothesis posits that a person’s consumption at a given point in time is determined by current income and expected future income. Permanent income is defined as expected long-term average income. This means that change in income in a given period will not alone determine consumption. Instead, consumers will evaluate consumption on a larger scale of lifetime income, which will ultimately be the main consumption determinant.
Breaking Down Consumption at the Macro Level
In our previous report, we used data from the United States Bureau of Labor Services Household Expenditure Surveys, which provided household-level data on consumption. Now we are using aggregate consumption data from the United States Bureau of Economic Analysis. The data begins in 1960 and is reported on a quarterly frequency, allowing us to evaluate consumption categories in a variety of growth environments over many cycles.
While there are 361 different consumption categories nested within three large areas – Durable Goods, Nondurable Goods, and Services – we are most interested in a select group of approximately 20 specific categories including, among others, Beer, Wine, Coffee & Tea, Restaurants, Soft Drinks, Food, and Jewelry. The multitude of categories reveals great variation in outcomes that contrast with the headline overall consumption data. While overall real consumption growth ranges from -4% minimum to 7% maximum, a simple average of our categories’ minimum and maximum rates yields -14% to 17%.
Our primary goal is to evaluate consumption categories of interest across different economic environments. Our first order of business is to define those differing environments. Our anchor variable is personal disposable income growth, since it has a more direct relation to consumption than the economy’s overall growth rate.
We have partitioned annual real disposable income growth outcomes into thirds, dividing them into Low Growth (bottom third), Medium Growth (middle third), and High Growth (upper third). This is to recognize the fact that the recession/expansion designation suffers from being binary and recessions being relatively few and far between compared to expansions. Our partitioning yields the following ranges: -3.7% to -0.2% for Low, -0.2% to 3.3% for Medium, and 3.3% to 6.8% for High.
Key Consumption Category Correlations
All Growth Phases
We also must understand how the categories of interest relate to each other, both over the entire timeframe and under our different growth outcomes. We conducted a cluster analysis that arranged each category of interest based on its correlations to other categories. This systematically formed clusters of cross-correlated groups, plotted below.
Key Consumption Category Correlations
High Growth Phases
Key Consumption Category Correlations
Low Growth Phases
Looking across all growth periods, we see two main correlation clusters – one that includes real spending on Beer, Wine, Restaurants, Food, and Spirits, and one that includes real spending on Cars, Jewelry and Watches, Clothing and Footwear, and Household Goods. This makes intuitive sense, as the former includes largely Nondurable Goods while the latter includes mostly Durable Goods. We can infer that spending on these specific categories tends to cluster around their greater type (Durable vs. Nondurable).
When we conduct the same analysis over the Low and High Growth periods, we see changes in almost all correlation pairs. This tells us that consumption relationships tend to change in different economic environments.
For example, we observe a moderately strong correlation between Beer and Spirits at 0.64 across all periods. However, we see that correlation declines from Low Growth to High Growth, going from 0.29 to -0.45. One interpretation is that during times of low income growth there is little discrimination between purchasing spirits and beer. In fact, the correlation is positive. However, as income growth increases, consumers begin to discriminate between the two, causing negative correlations.
How Does Consumption React to Economic Conditions, Habit, and Income?
Beyond understanding how our consumption categories relate to each other, we want to know how they relate to economic conditions throughout the cycle. To do this, we ran three different regressions.
First, we regressed our various consumption categories’ real consumption growth on real disposable income growth for different growth periods (Low Growth, Medium Growth, High Growth). The goal was to see how different categories respond to different economic growth environments.
Second, we looked for evidence of habit formation for our consumption categories. We did this by regressing the category’s real growth in the current period on four respective, previous quarters. This created a quick proxy for habit formation – if a certain quarter yields a significant coefficient, it may indicate the presence of habitual spending in that category.
Third, we regressed each category’s real growth on four respective quarters of previous personal disposable income growth. This was to identify any possible relationship between recent past income growth and category spending. If a category does not exhibit a relationship with current personal disposable income growth yet did so with (recent) past growth, it could evidence that spending tends to be associated with a consumer’s evaluation of longer-term income rather than the immediate period’s income.
Below, we present the results from three categories we found particularly interesting and illustrative. The results are presented in a “dashboard” that shows three figures. The leftmost graph shows the category’s elasticity estimates when regressed on the Low/Medium/High growth periods. The top right graph shows the category’s elasticity estimates when regressed on its previous four-quarter spending growth. The bottom right graph shows the category’s elasticity estimates when regressed on previous four-quarter income growth. Estimates that are statistically significant are shaded orange.
For Motion Picture Theaters, we note the very elastic, negative, and statistically significant estimate for Low Growth (-4.79). Given that personal disposable income growth in our Low Growth periods are all negative, this indicates that real consumer spending on the movies during such periods is actually quite strong. Recent evidence supports such results. Movies are intended to transport the consumer temporarily from reality. Such utility may be especially valued during hard economic times.
Motion Picture Theaters
When it comes to Jewelry, we find highly elastic and significant estimates for Medium and High Growth periods, versus inelastic and statistically insignificant estimates for Low Growth periods. This is unsurprising, since consumers tend to spend on jewelry when income growth is strong or average, yet refrain when income growth is mediocre.
Children’s and Infants’ Clothing yields an interesting result. We see a positive, statistically significant elasticity estimate for Low Growth periods. Since low growth periods are negative, this implies that real spending on children and infant clothing decreases during such economic times. One might think that children and infant clothing is a necessity that would be relatively inelastic to economic cycles, even downturns. This may be the case for Medium and High Growth periods, but our estimate is not statistically significant to allow such conclusions. Our results do raise the possibility that baby/child clothes are items that are cut back on during rough times. Perhaps families rely more on hand-me-downs or make do with fewer items knowing that they will need to be replaced soon enough anyway.
Children’s and Infants’ Clothing