Carlos Esteban Posadae This study analyzes the short‑ and long‑run determinants of household consumption in Colombia using monthly data from April 2003 to June 2022. We estimate Nonlinear Autoregressive Distributed Lag (NARDL) models to examine asymmetric long‑run responses of consumption to positive and negative changes in current income and interest rates. Consumption is measured using the retail trade index excluding motor vehicles and fuels. We find a cointegrating relationship linking consumption with current income (proxied by the Economic Situation Indicator, ESI), remittances, consumer credit, and the real interest rate on consumer credit. In the short run, consumer confidence and the age composition of the population also play a significant role in shaping consumption dynamics. The estimates reveal pronounced long‑run asymmetries: income increases are associated with a long‑run propensity to consume that is approximately 25 percent larger than that implied by income declines, while interest‑rate reductions elicit long‑run responses nearly 188 percent larger in absolute value than rate increases. These patterns are consistent with liquidity constraints, with the income‑driven asymmetry delivering a superior fit—by standard model‑comparison criteria—relative to the interest‑rate asymmetry. The negative association between population aging and consumption is difficult to reconcile with a benchmark life‑cycle model in the presence of capital‑market frictions. It also underscores the need to anticipate headwinds to aggregate demand as population aging proceeds.
The most recent
Approach
Using nonlinear cointegration techniques, this article examines the short- and long‑term determinants of household consumption between 2003 and 2022. As a proxy for consumption, it relies on the monthly retail trade indices excluding fuels and vehicles published by the National Administrative Department of Statistics (DANE in Spanish). The factors accounting for consumption dynamics include current income—measured here through the Economic Tracking Indicator (ISE)—interest rates, consumer credit, remittances, the consumer confidence index, and demographic indicators related to population aging.
Contribution
Despite representing nearly 75% of GDP, household consumption is, somewhat surprisingly, a relatively understudied variable in Colombia. This study contributes to broadening the understanding of household consumption expenditure. It analyzes whether consumption responds differently to positive versus negative changes in current income, and whether similar asymmetries appear in the responses to changes in interest rates and remittances. The article puts forward findings suggesting that such responses are asymmetric—particularly with respect to current income and interest rates—which may be consistent with the presence of liquidity constraints.
Increases in current income generate marginal propensities to consume that are about 25% higher than those associated with income declines, while reductions in interest rates induce long‑run responses nearly 188% larger than increases. These patterns are consistent with the presence of liquidity constraints.
Results
The estimates reveal important long‑run asymmetries: increases in current income generate marginal propensities to consume that are approximately 25% higher than those associated with income declines. Likewise, reductions in interest rates induce long‑run responses nearly 188% larger than those triggered by increases. Remittances and consumer credit also act as long‑run drivers of household consumption.
Consumer confidence and the age composition of the population play a significant role in shaping short‑run consumption dynamics. First, the sign and statistical significance of the coefficient associated with the consumer confidence index are consistent with the interpretation that higher confidence reflects lower perceived uncertainty regarding current and future interest rates, housing prices, credit availability, growth expectations, and other factors. This type of uncertainty is distinct from that related to income, which gives rise to precautionary savings. We do not find evidence of such behavior, as persistent increases in income lead to long‑run consumption responses close to 1, suggesting that consumption does not adjust downward in anticipation of possible future income declines. This is an issue that warrants further study.
Second, the fact that the estimates associated with population aging are negative and statistically significant is not consistent with the life‑cycle model, which assumes that consumption should remain smooth over the life cycle in a frictionless capital market. However, if the liquidity constraints highlighted above are taken seriously, this result should not be surprising, especially considering the decline in income at retirement in a country like Colombia, where persistent labor informality limits pension coverage to a small share of the population. This finding underscores the need to anticipate the effects on aggregate demand as population aging progresses.