Since 1986, Woolhandler and Himmelstein, alone
or with others, have written a series of articles that
follow a simple template.
In them, the authors
measure the administrative costs of the U.S. and
Canadian health care systems, subtract the second
from the first, and note the difference. This issue of
contains another article in the series.
The authors report that the difference between the
United States and Canada in outlays for health care
administration seems to be increasing. Others have
provided alternative estimates of administrative
costs in the United States and elsewhere.
literature has been motivated, in part, by speculation
that the savings in administrative costs from
switching to a single-payer system without cost
sharing could pay for the added health care services
that would result under a national health insurance
In reviewing this literature, an economist is
struck by how hard it is to identify and estimate administrative
costs accurately at a single point in time
in a single nation, how doubly hard it is to compare
costs at a single point in time among nations, and
how triply hard it is to make meaningful international
comparisons of trends in administrative costs
over time. All estimates depend on assumptions
about which costs are purely administrative and
how much of the costs of multipurpose functions
should be allocated to administration. Accurate international
comparisons must also account for differences
among accounting conventions and institutional
arrangements. In addition, international
comparisons over time must deal with shifting exchange
rates and divergent trends in relative wages.
As a practical matter, the conditions for accurate
comparison have proven impossible to satisfy.
Against this background, three questions arise.
First, do administrative costs in the United States
exceed those in Canada by about as much as Woolhandler
and colleagues say? Second, would the difference
in administrative costs really pay for the added
services induced by universal coverage with no
cost sharing? Third and, I think, most important,
what is the significance of the answers to the first
two questions in terms of policy?