Consumption and savings play a key role in the prevalence of chronic poverty. Yet the dynamics with respect to income and over time have been rarely explored due to a paucity of high-frequency data on a sufficient sample of households.
A panel error-correction model analysis of high-frequency income and consumption data from rural Uganda
We present a unique dataset and a behavioural model of sticky consumption that provides an explanation for a form of poverty trap: asymmetrically sticky consumption. We hypothesise that consumption is more sticky when it is too high (savings are being diminished) than when it is too low (savings are increasing).
We specify a novel latent-class panel error-correction model which accounts for aggregation bias within panel models of time series whilst providing improved efficiency for short-time panels. The model is estimated using weekly observations across 24 weeks on income, consumption, and savings from 480 coffee-producing households in eastern Uganda.
We qualitatively describe emergent latent cohorts in terms of savings dynamics and link cohort membership probabilistically to socio-demographic factors. Results support our hypothesis regarding asymmetrically sticky consumption for a large portion of our sample. In addition, we find 4 qualitatively differentiated cohorts basically characterised as ‘constant savers’, ‘episodic savers’, ‘sticky consumers’ and ‘inconsistent savers’. Sticky consumers and inconsistent savers in particular appear to be younger, have fewer assets and are more likely to engage in casual labouring.