'Investing' in care for old age? An examination of long-term care expenditure dynamics and its spillovers

Abstract

We study the dynamic drivers of expenditure on long-term care (LTC) programmes, and more specifically, the effects of labour market participation of traditional unpaid caregivers (women aged 40 and older) on LTC spending, alongside the spillover effects of a rise in LTC expenditure on health care expenditures (HCE) and the economy (per capita GDP). Our estimates draw from a panel of more than a decade worth of expenditure data from a sample of OECD countries. We use a panel vector auto-regressive (panel-VAR) system that considers the dynamics between the dependent variables. We find that LTC expenditure increases with the rise of the labour market participation of the traditional unpaid caregiver (women over 40 years of age), and that such expenditures rise exerts large spillover effects on health spending and the economy. We find that a 1% increase in female labour participation gives rise to a 1.48% increase in LTC expenditure and a 0.88% reduction in HCE. The effect of LTC spending over HCE is mainly driven by a reduction in inpatient and medicine expenditures, exhibiting large country heterogeneity. Finally, we document significant spillover effects of LTC expenditures on per capita GDP.

Supplementary information: The online version contains supplementary material available at 10.1007/s00181-022-02246-0.

Keywords: Care spillovers; Dynamic panel data; Female labour market participation; Health spending; Long-term care spending; Panel-VAR.

Fig. 1

Impulse response function of female labour participation. Figures show the orthogonalized impulse response functions (OIRF) along with 95% confidence intervals (“impulse variable” in logs; “response variable” in logs) based on 200 Monte Carlo simulations with 200 repetitions. In each figure, the horizon (5 periods) is set on the x-axis and the percentage change in the response variable is on the y-axis. Estimation of GMM panel-VAR for all countries and Bayesian panel-VAR for Northern and Southern countries. Step: time unit equivalent to one year

Fig. 2

Impulse response function of LTC expenditure over GDP pc. Figures show the orthogonalized impulse response functions (OIRF) along with 95% confidence intervals (“impulse variable” ⟶ “response variable”) based on 200 Monte Carlo simulations with 200 repetitions. In each figure, the horizon (5 periods) is set on the x-axis and the percentage change in the response variable is on the y-axis. Model 1 (FemPart, LTC, GDP pc) and Model 2 (LTC, HC, GDP pc). Estimation of GMM panel-VAR for both models with one lag and one to four lags in the endogenous instruments has been estimated. Step: time unit equivalent to one year

Fig. 3

Impulse response function of LTC expenditure over HC expenditure. Figures show the orthogonalized impulse response functions (OIRF) along with 95% confidence intervals (“impulse variable” in logs; “response variable” in logs) based on 200 Monte Carlo simulations with 200 repetitions. In each figure, the horizon (5 periods) is set on the x-axis and the percentage change in the response variable is on the y-axis. Estimation of GMM panel-VAR for all countries and Bayesian panel-VAR for Northern countries. Step: time unit equivalent to one year