For business and public policy applications, state demographic forecasting requires a modeling approach that incorporates a wide variety of assumptions about birth rates, death rates, fertility patterns, and domestic and foreign migration reliance on a large multidimensional matrix modeling system with extensive input vectors aids in projection of aggregate population and its characteristics, and facilitates policy applications.
California's population is one of the most dynamic and diverse in the :world. It currently totals 35 million, continues to grow rapidly, is one of the most ethnically diverse in the world, accounts for the majority of United States foreign in-migration, experiences high rates of both domestic interstate population inflows and outflows, and encompasses a broad age mix. It has dozens of different foreign languages spoken in its public schools, borders on large foreign nations both to its north and south, experiences great intrastate mobility, and exhibits considerable geographic socio-economic variations. Given all this, forecasting California's demographics is extremely challenging.
The California Legislative Analyst's Office (LAO) has a special interest in accurate short-term and long-term demographic forecasting. As the Legislature's nonpartisan fiscal and budgetary review office, the LAO projects the course of California's trillion-dollar economy and the state government's $70plus billion expenditure and revenue bases. In addition to contributing to these projections, demographic data also are used in evaluating and funding various individual governmental programsincluding education, prisons, health and social services programs, and infrastructure. Because the state population's age, gender, ethnicity, and geographic distribution all are of interest both to public policymakers who oversee the state budget and businesses that make up California's economy, the LAO's demographic forecasts must be highly disaggregated.
THE LAO'S FORECASTING APPROACH
As shown in Figure 1, the LAO uses the following basic three-stage demographic modeling approach.
Stage One-The Baseline Forecasting Matrix. This involves compiling base-year decennial U.S. Census data for California by age, gender, ethnicity, and each of California's 58 individual counties. Age breakouts in one-year increments ranging from I to the over-100 category are combined with five ethnic categories (white, black, Hispanic, Asian, and "other") and the two genders to produce 1,000 individual demographic "cells" that reside in a three-dimensional matrix. A fourth dimension is then formed by decomposing the state-level matrix along county lines. Thus, for example, there is a "cell" for the number of Asian California females aged 35 living in Los Angeles County.
Stage Two-Making the Matrix Base Current. The individual demographic cell data are then made "current" by adjusting the base-year data values for applicable birth, death, and migration data, as well as for the movement of the population into different cells each year due to the normal human aging process. The birth and death data come from state agencies and are compiled by the California Department of Finance (DOF). Regarding domestic and foreign migration data, the DOF develops these from a variety of sources, including California drivers' license records and federal income tax returns (both of which contain change-of-address information), and information from the Immigration and Naturalization Service (INS).
Stage Three-Projecting the Matrix Outward. The final step is to project the individual demographic cell data over the forecasting time horizon. This involves a wide variety of considerations, including age/gender/ethnicity/county-specific assumptions regarding the levels and intertemporal time patterns of fertility and birth rates, mortality rates, both domestic and foreign population inflows and outflows, and intrastate demographic locational shifts. …