Explaining variations in government health expenditure: evidence from Canada

Other possible confounders which are held constant by construction in our analysis include exchange rates, trade policy and associated access to technology, immigration policies, participation in military conflicts, the legal environment, and so forth. Other factors not identical but more similar across provinces than across countries include cultural norms, education opportunities and social support systems. All provincial systems also benefit from federal health transfer payments that fund approximately 20 to 25 percent of health spending.

Unlike the US where Medicaid and Medicare are federal entities, Canadian provincial systems each administer to all residents, including the elderly and low-income populations. Certain indigenous communities are the exception, some of which are federally supported.

Bohm et al., [15] details health system categories across OECD countries, including National Health Services, National Health Insurance (the case of Canada), Social Health Insurance, Private Systems and Statist Social Health Insurance systems.

After decades of increase, the OECD average health expenditure-to-GDP ratio flattened to around 8.8% from 2010 until the pandemic, when it rose again. In 2022 the ratio was 9.2 percent, down from its pandemic peak of 9.7 percent in 2021 [66].

Even in the US, one of the only nations without widespread public healthcare insurance, government funded Medicare and Medicaid in 2022 comprised nearly 40% of spending [20].

Authors calculations using CIHI data, 1975–2022. Average provincial government share of health spending slowly declined from around 76% prior to the 1990s to around 71% since the mid 1990s, rising to 74% again in 2020.

National health expenditure growth in the United States was projected to “stabilize” around 5.4 percent per year over the coming decade, increasing the health expenditure-to-GDP ratio to 20 percent [48]. The latest national health expenditure data for the United States shows that for now there has been a return to the pre-pandemic growth rate of 4.1 percent [42]. Estimates of health expenditure growth for Canada were 1.5% in 2022 and expected to grow 2.8 percent in 2023, following rates of 13.2% and 7.6% in 2020 and 2021 that departed from the pre-pandemic average of 4%.

A common specification explains per capita health care expenditures with per-capita national income, population age shares, the number of practicing physicians or hospital beds per capita, and public finance measures. We follow this convention. International aid is often considered in analyses of non-OECD nations.

Often, where healthcare is found to be income inelastic – a necessity – the argument can be made for more government involvement whereas if a luxury a free-market approach is likely to be preferred by some [22, 25].

Abdullah, Siddiqua, & Huque [1] and Murthy& Okunade [62] provide estimates for Asian and African countries, respectively, that suggest healthcare is a necessary good. Other cross-country estimates include Baltagi & Moscone [5] and Yetim et al. [81] for the OECD and Rodríguez and Nieves Valdés [78] for Latin American countries in comparison to the OECD.

The early literature on health spending determinants is global, and also includes Kleiman [50], Leu [56], Parkin, McGuire, and Yule [68], Gbesemete and Gerdtham [32], Gerdtham et al., [33] Hitiris and Posnett [44], Hitiris [45], Barros [9], Gerdtham et al., [34] and Gertham and Jonsson [35].

Using life expectancy as a proxy for technology, Murthy & Ketenci, [63]) examine US data for structural breaks.

The key data sources are as follows. GDP: Statistics Canada, Expenditure based, millions of dollars, 1961 to 1980: 3840015, 1981 to 2020: 36100222; 2021-2022 Estimate CIHI. Population and population by Age category: Statistics Canada CANSIM Tables 17100005 (1971-2022) and 1710060 (1968-1970). Physician numbers: CIHI, Supply-distribution-migration-physicians-ib-Canada-2020-datatables, CIHI https://​www.​cihi.​ca/​en/​physicians-in-canada. Provincial government health expenditures: CIHI, Supply-distribution-migration-physicians-ib-Canada-2020-datatables, CIHI https://​www.​cihi.​ca/​en/​physicians-in-canada and Statistics Canada (total provincial government health spending from 1965 to 1974 from Public Finance Historical Statistics, 68-512.); Federal government cash transfers from Finances of the Nation Database 1981 to 2022; Statistics Canada 65-88: D12842,: D12862,: D12882, D12902, D12922; d12942; D12962, D12982,: D13002; d12822. Annual unemployment rates are sourced from Statistics Canada table: for the years 1976 + and from Leacy [54] series D477-483 and D463-469 for prior years, where maritime and prairie provinces are grouped together.

Real GDP is deflated using the GDP deflator from Statistics Canada, with CPI interpolation for pre-1981. All other nominal variables are transformed to their real equivalents with base year 2021 using the annual average of CPI for all items available from Statistics Canada. Our results are robust to deflating health expenditure with the health care implicit price index from CIHI, which may help capture relatively higher inflation in the health sector. For the 1975 to 2023 period, the average All Items CPI inflation rate was 3.6 percent while that for the Government Current Expenditure Implicit Price Index was 3.9 percent, Health Care component of CPI was 3.7 percent, and the Health Care Implicit Price Index was 3.8 percent.

Additional benefits of log-transforming all variables include that it eases comparability across covariates (since estimates are unitless) and because our key variables are distributed normally only in log format within each panel.

Gross National Income is an alternative measure (see Parliamentary Budget Office of Ireland [69] for example). Provincial GNI data for Canada are not available for comparison.

Federal transfers to the provinces consist of the Canada Health Transfer, the Canada Social Transfer and Equalization. And in 2022-23 are estimated at about 83-billion dollars. See Canada, Department of Finance, Major Federal Transfers. https://​www.​canada.​ca/​en/​department-finance/​programs/​federal-transfers/​major-federal-transfers.​html. The Canada Health Transfer which in 2022-23 is estimated at 45 billion dollars (excluding COVID era top ups of 2 billion in 2022-23) is estimated to fund 20 to 25 percent of provincial government health spending. See Di Matteo [28].

Di Matteo, [28] delineates several eras in Canadian healthcare spending: an early boom (1957-1975) followed by stabilization (1976-1989), decline (1990-1996), recovery (1997-2009), renewed restraint (2010-2019) and the pandemic and immediate post pandemic era.

Standard unit root tests may be sensitive to structural breaks as illustrated by [74] and our data are suggestive of a structural break in the 1990s. Panel unit root tests that ignore breaks are available upon request.

Constant elasticity describes our data well; we estimated a moving average of the income elasticity of health spending relationship over the range of observations ± 2 SD from the mean, after partialling out other covariates. Predicted values fall within the confidence interval of an equivalent linear fit for almost the entire range pre-1997 and the entire post-1997 range.

Driscoll and Kraay [30] recommend their estimator for panel data with time dimension greater than 25 obs based on Monte-Carlo evidence.

We adopt the common choice for lag selection: \(nlags=floor\left[4{(T/100)}^{2/9}\right]\)

Own source revenue includes a “tax point transfer” that the federal government has directly made a part of provincial government revenues.

This replaced the 50/50 funding arrangement that had been in place since the introduction of hospital insurance in 1957 and Medicare in 1966. Various provinces join Medicare between 1966 and 1971. Regrettably, this break comes too early in our series for us to analyze its impacts.

Health expenditure data in particular is revised periodically and the pandemic period lends an additional dimension to such revisions.

This specification does not capture any dynamics in the income elasticity. We relax this assumption in a later section by allowing for different coefficients pre and post 1997.

See for example, Moir and Barua [59] and Canadian Institute for Health Information [16] for rankings showing the challenges facing Canadian when accessing health care.

See Memtsoudis et al., [60]. In 2021–21, 26 percent of male hip replacement patients were aged 55 to 64 and 12 percent were aged 18 to 54. For females, the numbers were 7 percent and 19 percent respectively. See Tables 4 and 5, CIHI [17] Hip and Knee Replacements in Canada: CJRR Annual Report, 2020–2021 — Updated Sept. 2022.

Bilgel and Tran [13] also find negative estimates in Canadian data using different estimators. Casas et al. [19] and others also include youth population shares (aged < 14), though these are not robustly significant. In a review of the literature on OECD countries, Martín Martín et al. [57] finds no consensus on age measures. Our results are generally robust to deviations in non-age covariates. We also consider the ratio of population aged 65 plus to population aged 15-44, finding a negative relationship with similar interpretation. Average relative provincial log-population shares for the unequal age groupings are available in Table 1.

We define left political parties in power to include New Democratic Party and Partis Québécois while Right include United Conservative party, Social Credit party, the Saskatchewan party, the Coalition Avenir Québec, Progressive Conservative party and the Unione Nationale. Omitted political party is centrist Liberal party.

Bellido et al. [12] find interplay between partisan relationships with health spending and recessions across the OCED while Potrafke [75] finds no partisan relationship in older data. Jacques et al. [46] also find that political leadership does not matter to certain lines of health care expenditure in Canada.

This decrease, in the context of our first-differences approach, may imply an important longer-term elasticity. We thank an anonymous referee for suggesting this exercise.

A manifestation of this result was evident recently as shortages of PPE, vaccines and other resources were evident at the onset of COVID-19 in most developed countries.

Allied professionals include dentists, chiropractors, optometrists, denturists, massage therapists, osteopaths, physiotherapists, podiatrists, psychologists, nurses and naturopaths.

This consolidated the separate health and social transfer cash grants of Established Program Financing into one cash grant but also reduced the cash total by about one third. See Di Matteo (2014, 2021). Federal health transfer funding only began to fully recover after 2004 with the transfer escalator of the Health Accord that saw federal health cash transfers grow 6 percent annually until 2017. It should be noted that in 2004 the CHST was broken up into the Canada Health Transfer (CHT) and a separate Canada Social Transfer (CST).

Murthy and Ketenci [63] find up to three breaks in US health care expenditure data spanning 1960–2012.

This test is designed for use in panel data. Our implementation is without fixed effects because our data are first differenced and with a HAC covariance estimator, finding a test statistic of \(W\left(\tau \right)=3.27\)

The estimate for the post 1975 period falls outside the 95% confidence interval of (0.16,0.28) for the period up to 1996.

See for example, Wierns et al., [79] and Ledford [55].

For example, for Ontario see Laberge et al., [53].