Estimating private health expenditures within a dynamic consumption allocation model [ Article 2008030701 - 07/03/2008]

The HTML version of the article above usually does not contain all information held in the PDF version. For a full version (including charts and tables), please download the PDF version available in the box 'PDF & download' on the top right side.

In the health economics literature, it is customary to model total (private and public) aggregate health expenditures as a function of income, demographic ageing and a host of other variables. Private health expenditures are rarely modelled separately and if they are, the models are often based on individual data and limited to specific medical services such as physician visits. In this paper we specify a model of aggregate private health expenditures, embedded in a generalisation of Deaton & Muellbauer’s Almost Ideal Demand System (AIDS). The main advantage of specifying a complete demand system is that all household consumption decisions, including the ones about the use of health care services, are subject to the same budget restriction.

The data used to estimate the demand system consist of total expenditures, budget shares and price indices of four aggregate consumption categories over the period 1980-2005. They are the broad aggregates: Non-durables, Durables, Health Care and Rent. Following Deaton & Muellbauer’s model, the budget shares of these aggregates are explained by total consumption spending and relative prices. The basic model is extended by including two demographic variables expected to capture shifts in consumption patterns related to the changing age composition of the population. The dynamic adjustment of consumption allocation decisions to exogenous shocks are modelled using the familiar ‘error-correction mechanism’ specification.

The estimation results are plausible, although homogeneity and symmetry restrictions placed on the parameters are rejected. The implied own price elasticities (which are calculated using the values of the model variables in 2005) indicate inelastic demand for Health and Rent, and elastic demand for Durables and Non-durables. Income elasticities are below unity for Durables, Rent and Health care, and above unity for Non-durables. The elasticities with respect to the demographic variables are positive for Health care and Durables and negative for Rent and Non-durables.

The model is used to project the budget shares of the four aggregates over the period 2006-2050. While this is much longer than the simulation horizon typically used for the medium-term HERMES model (in which a similar consumption allocation model is included), it provides a check that the model is well-behaved. Moreover, it is a time period that is relevant to studying the effects of demographic ageing. In the base run scenario, which uses available demographic projections and autoregressive extrapolations of prices and income, health care spending is projected to reach 5.4% of the household budget in 2050 (up from 4% in 2005). This increase reflects the effect of an increasing share of people aged 75 and more in the population, tempered by the low income elasticity of health care demand. Indeed, the ‘pure’ demographic effect, obtained in a scenario where income and prices are kept constant, leads to projected health spending close to 6% of total household consumption.