Poor mental health is a leading cause of disease burden worldwide, with an estimated one million people dying each year from suicide (Whiteford et al., 2013, Vigo et al., 2016, Wasserman, 2016). Despite the existence of effective treatment, such as cognitive behavioural therapy and pharmacotherapy (e.g., antidepressants, anxiolytics), many countries have documented high levels of unmet need and persistent regional variation in mental healthcare utilization (Moscone and Knapp, 2005, Cipriani et al., 2018, Maconick et al., 2021). The appropriate policy response to this variation depends on what causes it. Broadly speaking, this variation could be considered warranted if differences are due to patient need or preferences, but problematic if it is due to factors such as the availability of appropriate services or patients’ ability to pay. In particular, supply-driven variation could signal inefficiencies – if high utilization areas are characterized by excessive treatment yielding negligible marginal benefits – or inequity – if patients in low utilization areas are unable to access beneficial care. Considering a history of “chronic underfunding” and workforce shortages for mental healthcare across many countries, the latter may be more plausible (Krausz et al., 2019, WHO, 2021).

Empirical evidence on the population-level causes and consequences of variation in mental healthcare utilization is critical for the design of effective mental health policy. In this study, we investigate regional variation in mental healthcare utilization and mental health outcomes in Australia, a setting where the supply of mental health providers is limited and unevenly distributed (Phillips, 2013, Johar et al., 2017, Productivity Commission, 2020, Pulok et al., 2020). Specifically, our analysis focuses on government spending under Australia’s universal health insurance scheme, Medicare, which generally reimburses a fixed amount for eligible mental health services and prescriptions. There are large differences in mental healthcare utilizations across regions. For example, per capita spending on mental healthcare services in urban Central and Eastern Sydney is around nine times greater than rural Western Queensland. Aggregate data suggest that spending differences are correlated with health outcomes: suicide rates are much higher in rural Western Queensland (10.3 suicides per 100,000 population each year) compared to urban Sydney (3.5 suicides per 100,000 population each year).

Previous studies on the causes of regional variation in mental health and mental healthcare utilization have largely relied upon observable patient and regional level characteristics (Lê Cook et al., 2013, Awaworyi Churchill et al., 2019, Shiner et al., 2022). However, it is challenging with such an approach to distinguish demand- or supply- related factors. Patient demand for mental health treatment is largely unobservable and hindered by limited availability of population-level estimates on mental illness (Moscone et al., 2007). Risk factors that are often used as proxies for mental illness, such as area-level deprivation measures, may be influenced by supply and other place-based factors. At the same time, the greater availability of mental health providers in some regions may be a response to higher patient demand. Accounting for unobservable patient preferences is particularly important in the context of mental healthcare where mental health stigma and the perception of treatment efficacy can dramatically influence help seeking behaviour (Sickel et al., 2014).

To address these endogeneity concerns, we apply a ‘movers’ strategy that exploits patient migration across regions to identify the effects of patient and place-based factors (Finkelstein et al., 2016, Moura et al., 2019, Godøy and Huitfeldt, 2020, Salm and Wübker, 2020, Johansson and Svensson, 2022). The basic idea of this approach is that if regional differences in healthcare utilization are explained primarily by demand-side factors, individuals’ utilization will remain unchanged after they move, regardless of whether they relocate to a region with lower or higher average utilization. On the other hand, if regional differences are mainly driven by place-base factors, the utilization of patients who move will tend to adjust towards the average utilization of their destination region.

Our analysis is based on longitudinal administrative healthcare claims data linked to the Australian Census. We separately analyse the regional variation in the utilization of mental healthcare services and mental health prescriptions. These data provide comprehensive information on utilization and, importantly, allow us to observe when individuals move from one region to another.

To aid with interpretation of the main results, we test whether the estimated place effects are correlated with several region-level measures of supply: the number of psychiatrists and psychologists per capita, out-of-pocket costs, and mean wait times for mental health treatment. Finally, we explore whether larger place-based utilization predicts reductions in acute mental healthcare use or fewer suicides.

Our main results indicate quite different patterns for mental healthcare services and prescriptions. Place effects account for roughly 72% of the regional variation in utilization of mental healthcare services. In contrast, we find that place effects account for only 19% of the variation in the utilization of mental health prescriptions. These results are consistent with the fact that in Australia, most mental health medications are prescribed by GPs, whereas most mental health services are provided by specialty providers. There is also much less regional variation in the supply of GPs as compared to specialty mental health providers (Department of Health and Aged Care, 2023) and prescribing generally requires less health professional time. Thus, in Australia supply constraints are likely to be less relevant for prescription drugs than for mental health services. Furthermore, because GPs have limited financial incentives to over-prescribe, there is less reason to expect that supplier induced demand would produce regional variation in prescribing. We find that greater availability of mental healthcare providers and lower out-of-pocket costs are positively correlated with the estimated place effects, particularly for mental health services. We also find a positive relationship between waiting times and the place effects for mental health prescriptions. These patterns suggests that regional variation in mental healthcare utilization is largely driven by variation in supply and that where it is difficult to see mental health providers, patients are more likely to be treated with prescription drugs.

The welfare implications of these results depend on the extent to which greater supply translates to better health outcomes. We find suggestive evidence that higher place-based utilization is associated with improved mental health outcomes. A one standard deviation increase in place-based utilization for mental health services predicts a 10% reduction in mental health emergency department (ED) presentations, a 20% reduction in self-harm hospitalisations, and a 10% reduction in suicides. These associations suggests that the variation in utilization reflects inadequate supply across regions rather than inefficient high supply in high-utilization areas.

This paper contributes to a growing literature that seeks to determine the relative importance of demand and supply-side factors in explaining regional variation in health care utilization. The results from existing studies indicate that the importance of supply-side factors varies substantially across types of care and institutional context. For example, whereas Finkelstein et al. (2016) find that place factors explain 60% of the variation in all health care utilization among Medicare beneficiaries in the US, Salm and Wübker (2020) find that place effects account for only 9% of the variation in ambulatory care utilization in Germany. The only other concurrent study focusing on mental healthcare in this literature, which used US Medicare data, found that place factors explained 46% and 15% of the variation for mental healthcare services and prescription medicines, respectively (Ding, 2023). Our study also connects to the literatures on the determinants and consequences of scarce mental healthcare supply. In particular, our results align with previous evidence that selection of providers into regions can lead to large inequities in healthcare access (Gravelle and Sutton, 2001, Rosenthal et al., 2005, Isabel and Paula, 2010, Grobler et al., 2015, Swami and Scott, 2021) and that easing supply constraints can improve health outcomes (Ludwig et al., 2009, Okeke, 2023).

Our findings have important implications for mental health policy. Recent legislative inquiries in Australia have emphasized the need for “improved targeting” and more equitable access of mental healthcare (Productivity Commission, 2020, Pirkis et al., 2022). Australia has implemented various policies aimed at reducing regional variability in GP access, including financial incentives and compulsory service (Swami and Scott, 2021, Department of Health and Aged Care, 2022). These policies may have contributed to the improved accessibility of mental health prescriptions. However, there have been relatively few measures to increase the supply of specialty mental health providers in underserved areas. The identified relationships between place-based expenditure and mental health outcomes further suggests that, at a population level, universally increasing effective supply might be more beneficial than focusing on allocative inefficiencies across regions.

The paper is organised as follows. Section 2 describes the institutional setting and data. Section 3 presents the empirical strategies for estimating the relative importance of patient- and place-related factors and discusses our identifying assumptions. The main results, robustness checks, heterogeneity analyses, and potential mechanisms are presented and discussed in Section 4. Section 5 explores the association with place-based expenditure and mental health outcomes. Section 6 concludes.