Forecast of Public Health Insurance Revenues
Forecasting public health insurance revenues is essential not only for health insurance companies, but also for the Ministry of Finance, for example when setting medium-term expenditure frameworks of the state budget and state funds within the budgetary process. The revenue forecast is used by the health insurance companies, because they are obliged to prepare an annual health insurance plan according to Act No. 551/1991 Coll. and Act No. 280/1992 Coll. Then Act No. 23/2017 Coll. requires health insurance companies to create a medium-term outlook for revenues and expenditures for the next two years.
In connection with the Fiscal Responsibility Rules Act, the Ministry of Finance draws up a binding government document for the preparation of the draft state budget, the budgets of state funds and their medium-term outlook – the Budgetary Strategy of the General Government Sector of the Czech Republic. The document contains expenditure frameworks of the state budget and state funds or financial relations of the state budget to the special account of public health insurance. As part of establishing medium-term expenditure frameworks, the forecast of public health insurance revenues represents both a component of total general government revenues, from which total general government expenditure is derived, as well as one of the key variables for determining the structural balance of the general government sector components outside the state budget and state funds.
To forecast revenues from public health insurance, it is crucial to determine the amount of revenues from health insurance contributions, since it accounts for about 99% of the total revenues of health insurance companies. The other components of the revenues of health insurance companies can be considered negligible. The amount of the health insurance contribution is generally determined by the assessment base, the number of insured persons and the health insurance contribution rate (for all payer groups it is 13.5% of the relevant assessment base). Since Act No. 592/1992 Coll. defines different assessment bases for different insurance payers, the determination in this methodology of the public health insurance revenues is based on the specification of partial models for the forecast of insurance health insurance contributions from individual payers. The total public health insurance revenues are then determined by aggregating the partial forecasts of health insurance contributions of the relevant payers.
To formulate the partial models, use is made of quarterly time series of relevant variables for the years 2000–2018. In all time series, a logarithmic transformation is used to reduce the level difference and the difference in the variance of the individual series. Because of the non-stationarity of time series, the variables are expressed in the first differences. The forecast accuracy of the models is assessed for each specification based on the average root mean square error. According to the minimum value of this criterion, use was made of the estimate using the modified least squares method as the most appropriate estimate to forecast health insurance contributions from employees; for health insurance contributions paid by the state for the state insured persons, the regression model includes, as an independent variable, the assessment base and the number of unemployed registered with the Labour Office. In the case of self-employed persons, the error correction model appears to be the best of the tested models, and for persons without taxable incomes it is a simple process of first order moving averages without the logarithmic transformation of variables. As a reference model, a “naïve” first order autoregression process is then chosen for all payers without further independent variables.
In the period between 12 July and 17 August 2018, the material was subject to consultation with the Ministry of Health, the General Health Insurance Company and the Association of Health Insurance Companies. Their relevant comments were subsequently incorporated in the methodology. Commenting institutions expressed their agreement with the concept of partial forecasts according to individual payers as well as with selected model specifications, including selected independent variables. The methodology has been found to be useful for forecasting the revenues of the entire public health insurance system.