Regional differences in health care costs: What do moves within Switzerland reveal about supply and demand?

On average, annual per-capita costs in our dataset across all health care regions (Hospital Service Areas, or HSAs) are around CHF 2,400—yet the differences are substantial. HSAs are functional care areas (74 in Switzerland) defined based on patient flows, allowing regional analyses that follow actual care pathways. Particularly high spending is found in urban centers such as Geneva and Basel-Stadt, while central German-speaking regions tend to be lower; the ranking of regions over time is surprisingly stable.

When people move from one health care region to another, they don’t just change address—they also enter a different local care environment. We use exactly this change to measure how strongly health care use adjusts after a move to the spending level of the destination region: for example, if someone moves from a high-cost Lake Geneva region to the comparatively low-cost region of Uri, then under a 100% supply effect their health care costs would immediately drop to Uri’s lower level after the move. Under a 100% demand effect, the same person would not adjust their consumption at all and would remain at the high original level despite moving.

We use administrative data from CSS (2010–2022) and follow individuals who move once across HSAs and, after moving, actually receive a large share of their care in the destination region. Our empirical strategy follows the movers design of Finkelstein et al. (2016) and is implemented as an event study. In other words, we examine the years before and after a move and aggregate across all moves in our sample. The CSS data are rich in detail and allow precise tracking of individual trajectories around the move. However, they may not be representative of the entire resident population—especially if moving increases the likelihood of switching insurer. We therefore additionally replicate the key results using individual-level data from the Federal Office of Public Health (FOPH/BAG) covering the full population (2019–2023) to assess external validity.
 

For total spending, our analysis shows that the larger share of regional differences is driven by the demand side: around 60% can be explained by demand-side factors. These include potential differences in health status and morbidity, age structure, risk behavior, health literacy, and socioeconomic conditions, as well as preferences and expectations regarding medical care. The remaining 40% therefore reflects supply-side factors, such as the density and composition of providers, accessibility, or regional practice styles.

The supply effect varies substantially across types of services. Areas that are more strongly shaped by patient preferences and involve less physician discretion—such as prescription drugs (14%) or physiotherapy (33%)—are the least supply-driven. By contrast, services with greater leeway in diagnostics and treatment intensity are much more supply-driven, such as lab services (42%) or total GP costs (57%). Differences are also large within specialist care—for example between gynecology (33%) and surgery (67%). This fits with the idea that surgical services more often involve procedure-related decisions, whereas many gynecological services are more standardized. An especially revealing pattern emerges for general practitioners: the first contact (a GP visit) is comparatively less supply-driven, while follow-up services (and thus total costs) are much more so—consistent with the idea that providers mainly shape treatment intensity.
 

Among those under 45, the estimated supply component is much higher—68%—than among those over 45 (30%). This suggests that younger people adjust their utilization more strongly after a move to match the spending level of the new regional care environment—for example because they are less tied to chronic conditions and established treatment pathways, and therefore can respond more flexibly to local structures.

By gender, the overall effects are very similar, indicating that men and women react to regional differences to a comparable extent in aggregate. At the level of specific service types, however, some differences do emerge: for certain services—such as laboratory services—adjustment patterns differ noticeably.
 

In den BAG-Daten fällt die geschätzte Angebotskomponente der totalen Kosten niedriger aus (rund 22%). Der Unterschied erklärt sich vor allem dadurch, dass dort wichtige Einschränkungen (z.B. «nicht exponierte» Umzüge oder Umzüge aufgrund gesundheitlicher Veränderungen) nicht gleich umgesetzt werden können. Die Angebotswirkung wird dadurch mechanisch nach unten verzerrt. Interessanterweise verschwinden die Unterschiede weitgehend, sobald wir in den CSS-Daten dieselben (weniger restriktiven) Kriterien anwenden wie in den BAG-Daten: Ohne diese relevanten Einschränkungen liegt die geschätzte Angebotskomponente auch in den CSS-Daten bei nur rund 19% und damit sehr nahe am BAG-Wert. Es zeigt sich also, dass die CSS-Daten den Gesamtmarkt sehr gut abdecken und sogar von Vorteil sein können, aufgrund des Detailgehalts.

First, a large share of regional cost differences is linked to the demand side. Supply-side steering alone can therefore address only part of the variation.

Second, where supply effects are large (e.g., diagnostics, surgical services, treatment intensity), targeted levers are especially relevant—for instance through incentives, clinical guidelines, transparency, or organizational models that help structure treatment decisions.
 

Moves within Switzerland provide a rare, clear window into why health care costs vary so strongly across regions. Overall, differences are more demand-driven than supply-driven—but the balance shifts markedly by type of service and by age group. This kind of differentiation is crucial if efficiency-improving measures are to be targeted where they can have the greatest impact.