Methodology

We focus on the ratio of health spending to GDP, and decompose the growth rate of the ratio (that is, not the growth rate of health spending in nominal terms). The spending-to-GDP ratio (g_t) is expressed as

where G_t, Y_t, and N_t are health spending in nominal terms, nominal GDP, and the population year t, respectively. G_it and N_it are health spending and the population for age cohort i. $\frac{G_{bt}}{N_{bt}}$ is the benchmark per capita health spending i.e., per capita spending of a benchmark population group. This equation shows that g_t consists of:

• e_t: the benchmark per capita health spending, divided by GDP per capita;

• h_it: per capita health spending of cohort i, divided by the benchmark per capita health spending; and

• n_it: the population share of cohort i.

With this, the growth rate of g_t can be decomposed into three components: ⁴

where $w_{it}\equiv h_{it}{n}_{it}/\left(\mathrm{\Sigma }{h}_{it}{n}_{it}\right)$. Here a “hat” represents the percent change of a variable, i.e., ${\hat{a}}_{t+k}=(a_{t+k}-a_t)/a_t$. Thus, the spending ratio to GDP grows when:

• the share of old-age population increases (ageing);

• The benchmark per capita health spending rises faster than GDP per capita (excess cost growth); and/or

• per capita spending of some age cohorts grows faster than the benchmark per capita spending (spending profile shift).

Ageing—an increase in the population share of the elderly—raises the health spending to GDP ratio because an elderly person spends more on health than the young. More specifically, ageing brings up the population share (n_i) of the elderly, while reducing that of the young. This increases the ageing effect component $\left({\mathrm{\Sigma }}_{\text{i}}{\text{w}}_{\text{it}}{\hat{n}}_{i,t+k}\right)$ because the weight W_t is larger for the elderly cohort than for the young cohort.

Excess cost growth can result from various non-demographic factors, such as high income elasticity of health care demand; technological advances (which expand the boundary of medically possible treatments and improve the quality of services at a higher cost); the so-called Baumol effect (which refers to rising unit labor costs in the health sector reflecting its relatively low productivity growth); and health policies and institutions (Clements and others, 2012). Excess cost growth will be further decomposed into changes in three subcomponents: (i) the benchmark per capita spending in real terms (x_t); (ii) potential real GDP per capita (poty_t); and (iii) the output gap plus one (ygap_t).⁵ The idea is to separate the effect of the output gap on excess cost growth. Specifically,

By looking at spending profile shift, we can examine whether healthy ageing has taken place or not. In other words, if the health status of the elderly improves in tandem with higher life expectancy, their per-capita spending at a given age would be lower. If this is the case, health spending growth would slow down.

We decompose health spending growth during 1970–2011 into the three components. For HC, we examine spending data broken down for 18 age cohorts, with age 0–4 as the youngest age cohort and age 85 and over as the oldest. The breakdown of HC spending by age cohort is not available for years before 1998; for these years, the normalized per capita health spending for each age cohort (h_it) is assumed to be unchanged at 1998 leves. For LTC, we examine spending for 6 age cohorts (age 40–64, 65–69, 70–74, 75–80, 80–84, and 85 and over). For HC, the benchmark per capita health spending (G_bt/N_bt) is per-capita spending for the working age population (age 20–59). For LTC, it is per-capita spending of the cohort with age 65–69, because per-capita spending of younger age cohorts is very small.

Decomposition results

Table 1 summarizes the result of the decomposition analysis. It tabulates the average growth rate of each component for various sub-periods since 1970. We exclude observations for 2000–02 to take account of the structural break associated with the launch of public long-term care insurance. Data for spending profile shift is not available prior to 1998, for the reason noted above.

Table 1.

Japan: Decomposition of Changes in Health Spending, 1970–2011

(Annual percent change, period average)

Sources: Ministry of Health, Labor, and Welfare; IMF World Economic Outlook database; and IMF staff estimates.

^1/Excludes observations for 2000–02.

Table 1.

Japan: Decomposition of Changes in Health Spending, 1970–2011

(Annual percent change, period average)

				1990–2011 1/		1970–79	1980–89	1990–99	2003–11
				Mean	St. dev.	Mean	Mean	Mean	Mean
Ratio of health care and long-term care spending to GDP				3.1	2.8	4.4	−0.2	2.8	3.5
	Ageing effect			1.8	0.2	1.0	1.4	1.7	1.9
		Health care		1.6	0.2	1.0	1.4	1.7	1.5
		Long-term care		4.4	0.6	…	…	…	4.4
	Excess cost growth			0.9	2.9	3.3	−1.5	0.6	1.3
		Per capita spending of benchmark cohort (real)		1.8	2.1	7.1	2.3	1.6	2.0
			Health care	1.8	2.1	7.1	2.3	1.6	2.1
			Long-term care	1.6	4.8	…	…	…	1.6
		Potential GDP per capita		0.9	0.9	3.9	4.0	1.1	0.6
		GDP gap (plus one)		0.0	1.7	−0.2	−0.1	−0.1	0.1
	Spending profile shift			0.3	1.2	…	…	…	0.2
		Health care		0.4	1.2	…	…	…	0.3
			Age 0–19	0.1	0.2	…	…	…	0.2
			Age 20–59	0.0	0.0	…	…	…	0.0
			Age 60 and over	0.2	1.1	…	…	…	0.1
		Long-term care		0.2	1.2	…	…	…	0.2
Memorandum items:
	Ratio of health care spending to GDP			3.0	2.7	4.4	−0.2	2.8	3.1
	Ratio of long-term care spending to GDP			5.5	4.5	…	…	…	5.5

Sources: Ministry of Health, Labor, and Welfare; IMF World Economic Outlook database; and IMF staff estimates.

^1/Excludes observations for 2000–02.

View Table

Table 1.

Japan: Decomposition of Changes in Health Spending, 1970–2011

(Annual percent change, period average)

				1990–2011 1/		1970–79	1980–89	1990–99	2003–11
				Mean	St. dev.	Mean	Mean	Mean	Mean
Ratio of health care and long-term care spending to GDP				3.1	2.8	4.4	−0.2	2.8	3.5
	Ageing effect			1.8	0.2	1.0	1.4	1.7	1.9
		Health care		1.6	0.2	1.0	1.4	1.7	1.5
		Long-term care		4.4	0.6	…	…	…	4.4
	Excess cost growth			0.9	2.9	3.3	−1.5	0.6	1.3
		Per capita spending of benchmark cohort (real)		1.8	2.1	7.1	2.3	1.6	2.0
			Health care	1.8	2.1	7.1	2.3	1.6	2.1
			Long-term care	1.6	4.8	…	…	…	1.6
		Potential GDP per capita		0.9	0.9	3.9	4.0	1.1	0.6
		GDP gap (plus one)		0.0	1.7	−0.2	−0.1	−0.1	0.1
	Spending profile shift			0.3	1.2	…	…	…	0.2
		Health care		0.4	1.2	…	…	…	0.3
			Age 0–19	0.1	0.2	…	…	…	0.2
			Age 20–59	0.0	0.0	…	…	…	0.0
			Age 60 and over	0.2	1.1	…	…	…	0.1
		Long-term care		0.2	1.2	…	…	…	0.2
Memorandum items:
	Ratio of health care spending to GDP			3.0	2.7	4.4	−0.2	2.8	3.1
	Ratio of long-term care spending to GDP			5.5	4.5	…	…	…	5.5

Sources: Ministry of Health, Labor, and Welfare; IMF World Economic Outlook database; and IMF staff estimates.

^1/Excludes observations for 2000–02.

Key observations are fourfold. First, health spending growth became substantially positive during the last two decades. The increase in the health spending to GDP ratio from 4.4 percent in 1990 to 9.8 percent in 2011 is equivalent to an average annual growth rate of 3.1 percent. In contrast, the ratio had decreased during the 1980s at an annual rate of 0.2 percent; thus, health spending growth picked up discontinuously in the beginning of the 1990s, coinciding with the so-called “lost decades” for Japan. The growth rate of the spending ratio accelerated to 3.5 percent annually during 2003–11.

Second, about two-thirds of the spending ratio’s increase during the last two decades originated from the ageing component, which grew by 1.8 percent annually. The ageing effect accelerated from an annual growth rate of 1 percent during the 1970s to 1.9 percent during the last decade, reflecting the rapid pace of population ageing. The standard deviation of the ageing effect is significantly smaller than that of excess cost growth or spending profile shift, reflecting the fact that population growth does not vary much from year to year.

Third, at an annual average rate of 0.9 percent, excess cost growth also contributed significantly to the rapid health spending growth during the last two decades. An interesting observation is that the growth rate of the benchmark per capita spending has been relatively stable during the last three decades, with the average increase over each decade around 2 percent annually. In fact, the pickup in excess cost growth from minus 1.5 percent during the 1980s to 1.3 percent during 2003–11 reflects a decline in per-capita GDP growth during the lost decades (this is consistent with the right panel of Figure 3). The contribution of the output gap is negligible for all sub-periods; thus, the recent pickup in excess cost growth did not result from a widening of the negative output gap after the global financial crisis.⁶

Fourth, at an annual average growth rate of 0.3 percent, spending profile shift contributed positively to health spending growth from 1998 to present. Albeit small, the positive contribution from spending profile shift for HC for the elderly (age 60 and older) and for LTC means that healthy ageing has not taken place in Japan during this period.

Cross-country comparison of health spending decomposition

How do these results compare with other advanced economies? Figure 4 presents a similar decomposition of health spending growth in OECD countries over 1990–2010.⁷ Japan’s growth rate of the health spending to GDP ratio, at an annual rate of 3.1 percent, was the third highest among these countries (the median increase for OECD countries was 1.8 percent annually). This is a result of the significant ageing effect for Japan—the 1.8 percent annual growth rate of the ageing component was by far the highest among advanced economies (the median increase was only 0.5 percent). In addition, Japan’s excess cost growth of 1.3 percent annually (inclusive of spending profile shift) was comparable to its peers (the median increase was 1.1 percent annually).

View Full Size

OECD Countries: Decomposition of Changes in Health Spending ^1/

(Annual percent change, average for 1990–2010)

Citation: IMF Working Papers 2014, 142; 10.5089/9781498365277.001.A001

Sources: OECD; IMF, World Economic Outlook database; and IMF staff estimates.Note: The data for Japan is from Table 1 (public and private spending during 1990–2011 excluding 2000–02). For other countries, OECD data on public spending during 1990–2010 is used.^1/ Includes LTC spending.

• Download Figure
• Download figure as PowerPoint slide

The ageing contribution is substantial in Japan because per-capita health spending by the elderly (relative to that by the young) is much higher than in other countries. Figure 5 illustrates the profile of per-capita spending across age cohorts in OECD countries. In Japan, a person of age 85 years and more spends on health care about 7 times as much as a person of age 40–44 years, and on long-term care about 30 times as much as a person of age 65–69; these ratios are the highest among OECD countries. This pattern is also observed for cohorts of age 70 years and older. Because of this, population ageing amplifies health spending growth much more significantly in Japan than in other advanced economies.

View Full Size

OECD Countries: Health Spending by Age Cohort ^1/

Citation: IMF Working Papers 2014, 142; 10.5089/9781498365277.001.A001

Sources: Ministry of Health, Labor, and Welfare; De La Maisonneuve and Oliveira Martins (2013); and IMF staff estimates.^1/ The data for Japan is public and private spending for 1990–2011. The data for other countries is public spending.

• Download Figure
• Download figure as PowerPoint slide

Spending projection

We project the health spending to GDP ratio using the decomposition framework discussed above. Recall that the spending ratio consists of three components (g_t = e_t Σ_i h_it n_it):

(i)the population share by age cohort (n_it); (ii)per capita health spending by age cohort, divided by the benchmark per capita spending (h_it); and (iii)the benchmark per capita health spending divided by GDP per capita (e_t). For n_it, we rely on population projections published by the National Institute of Population and Social Security Research (IPSS), a popular source among researchers. The IPSS publishes population projections for 3 variants of fertility and mortality rates (high/medium/low), and we use the one under the medium fertility and mortality rates.⁸ For h_it, we assume no change throughout the projection period and use the average values for 2009–11 for each age cohort (later in this subsection, we consider the effect of healthy ageing). For e_t, we consider excess cost growth of 1 percent as a scenario that reflects the historical trend; this would be reasonable also from the cross-country perspective discussed in the previous subsection. In addition, we also consider excess cost growth of 0 percent to illustrate the impact of the ageing effect alone, and 2 percent as an indicative upper bound (Figure 4 shows that 5 out of 23 advanced countries had annual excess cost growth above 2 percent during 1990–2011). Projections are done separately for health care and long-term care, and then aggregated.

Our projection indicates that population aging and continuation of excess cost growth would substantially increase health spending (Figure 6). As shown by the scenario with zero excess cost growth, the ageing effect alone will bring up health spending from 9½ percent of GDP in 2010 to 13 percent of GDP in 2030. The scenario with annual excess cost growth of 1 percent shows that, if excess cost growth continues at the trend rate observed during the last two decades, health spending will balloon to 15½ percent of GDP in 2030, a 6 percentage point increase between 2010 and 2030. As a worst-case scenario, if annual excess cost growth is 2 percent, the ratio will double to 19 percent by 2030. If 1 percent annual excess cost growth continues through 2060, spending will skyrocket to more than a quarter of GDP. The uncertainty surrounding spending projection increases dramatically as the time horizon is extended as far as to 2060.

View Full Size

Japan: Projected Health Spending, 2010–60^1/

(In percent of GDP)

Citation: IMF Working Papers 2014, 142; 10.5089/9781498365277.001.A001

Sources: Ministry of Health, Labor, and Welfare; the National Institute of Population and Social Security Research; and IMF staff estimates.^1/ Includes LTC spending.

• Download Figure
• Download figure as PowerPoint slide

A statistical analysis indicates that the health spending to GDP ratio in 2030 likely falls between the two scenarios with no excess cost growth and with 2 percent annual excess cost growth. The left panel of Figure 7 illustrates a fan chart for the spending to GDP ratio through 2030 based on an assumption that annual excess cost growth follows a white noise process with the mean and the standard deviation observed during 1990–2011 (excluding observations for 2000–02; see Table 1) and the population projection under the medium variant of fertility and mortality rates. The fan chart shows that, in 2030, the lower bound of the spending ratio’s 90 percent confidence interval roughly coincides with the projection under no excess cost growth; and the upper bound of the confidence interval with the projection under 2 percent excess cost growth. Additionally, our spending projection does not seem sensitive to assumptions on fertility and mortality rates: the right panel of Figure 7 shows that the projected spending ratio in 2030 under the low-fertility-and-low-mortality case differs only by 1 percentage point from that under the high-fertility-and-high-mortality case.

View Full Size

Fan Chart of Health Spending to GDP Ratio and Sensitivity to Population Dynamics^1/

(In percent of GDP)

Citation: IMF Working Papers 2014, 142; 10.5089/9781498365277.001.A001

Sources: Ministry of Health, Labor, and Welfare; the National Institute of Population and Social Security Research; and IMF staff estimates.^1/ Includes LTC spending. In the fan chart (left panel), the yellow area shows the range between the 25 percentile and 75 percentile, the orange area between the 10 percentile and 90 percent, and the light blue area between the 5 percentile and 95 percentile.

• Download Figure
• Download figure as PowerPoint slide

Healthy ageing could slow down health spending growth. Between 2010 and 2030, Japanese life expectancy is expected to increase by 2.3 years, according to the population projection by the IPSS (using the medium variant mortality rate). If the health status of the elderly improves in tandem with higher life expectancy, their per-capita spending at a given age would be lower. To gauge the effect of such healthy ageing, we consider a case where the per-capita spending profile by age would shift concomitantly with the increase in life expectancy, as depicted in the left panel of Figure 8. The result presented in the right panel of Figure 8 shows that healthy ageing could trim about 1 percentage point of GDP from the projected increase in health spending over 2010–30.⁹

View Full Size

Effect of Healthy Ageing

Citation: IMF Working Papers 2014, 142; 10.5089/9781498365277.001.A001

Sources: Ministry of Health, Labor, and Welfare; the National Institute of Population and Social Security Research; and IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

Economic growth can also affect the health spending to GDP ratio in the long run. In the analysis above, we focused on excess cost growth, assuming a stable relationship between health spending and economic growth. We take this approach to avoid pinning down the long-run GDP growth rate in light of large forecasting errors. Nevertheless, the fact that health spending growth was much higher than economic growth in the last two decades (see Table 1 and the left panel of Figure 3) warrants discussion of an illustrative scenario where health spending is “decoupled” from economic growth. The latest government projection envisages a real GDP growth rate of about 2 percent per year for 2013–2022 under a scenario with growth-enhancing reforms (the “reform scenario”), while also presenting a pessimistic scenario with a real GDP growth rate of about 1 percent per year (the “pessimistic scenario”).¹⁰ Assuming the annual growth rate of the benchmark per capita health spending in real terms at 2 percent as observed during the last decades (see Table 1), and incorporating the average population growth rate of -0.5 percent per year during 2010–30 (under the mid-fertility-mid-mortality assumption), the GDP growth rate for the reform scenario would imply excess cost growth of -0.5 percent per year, which corresponds to an increase in health spending by 2.2 percentage points of GDP over 2010–30. Similarly, the pessimistic scenario would be equivalent to excess cost growth of 0.5 percent per year, which implies an increase of health spending by 4.7 percentage points of GDP over 2010–30.

Financing mix projection

Health spending is financed by premium contributions, patient copayments (out-of-pocket spending), and government transfers. In 2010, premium contributions accounted for 48 percent of health spending (48 percent for HC and 45 percent for LTC), and patient copayments for 13 percent (13 percent for HC and 10 percent for LTC). The reminder, 39 percent of health spending, is financed by government transfers (by both the central and local governments).

The working age population carries a heavier burden of premium contributions and copayments than the elderly. Figure 9 depicts premium contributions and copayments for HC spending by age cohort in 2010, as a percent of HC spending for each cohort.¹¹ Since Japan’s health insurance system is built on a social insurance principle, the working age population financially supports the elderly who are vulnerable to financial and health risks. Reflecting this, age-specific premium contributions for working age groups (aged between 20 and 55) exceed their HC spending. The ratio of contributions to HC spending declines with age as HC spending becomes disproportionately higher as individuals get older. With higher HC spending at older age, the ratio reaches even below 10 percent for those aged 75 years and older. Likewise, the ratio of copayments to spending is disproportionately higher for the young: the ratio for HC spending is above 20 percent for people of age 5–69 years and below 10 percent for age 70 years and more.

View Full Size

Japan: Premium Contributions and Patient Copayments for Health Care by Age Cohort, 2010

(In percent of health care spending for each age cohort)

Citation: IMF Working Papers 2014, 142; 10.5089/9781498365277.001.A001

Source: Ministry of Health, Labor, and Welfare.

• Download Figure
• Download figure as PowerPoint slide

Taking this profile into account, we project the long-term financing mix of health spending. We focus on the scenario with annual excess cost growth of 1 percent. To project contribution payments and copayments for HC spending, we apply the profile of age-specific premium contributions in percent of spending and copayment rates in 2010 (as depicted in Figure 9). For LTC spending, we project the financing mix based on the statutory splitting rule (45 percent by premium contributions, 10 percent by copayments, and 45 percent by government transfers). Government transfers are then computed as the residual (total HC and LTC spending minus the sum of premium contributions and copayments). Figure 10 shows the projected financing mix through 2030 in percent of total spending as well as in percent of GDP.

View Full Size

Japan: Projected Financing Mix of Health Spending, 2010–30^1/

Citation: IMF Working Papers 2014, 142; 10.5089/9781498365277.001.A001

Sources: Ministry of Health, Labor, and Welfare; the National Institute of Population and Social Security Research; and IMF staff estimates.^1/ Includes LTC spending.

• Download Figure
• Download figure as PowerPoint slide

Three observations are worth highlighting. First, the share of health spending financed by copayments is projected to decline marginally from 13 percent in 2010 to 12 percent in 2030. This reflects population ageing, that is, the rising share of the elderly whose effective copayment rate is lower. As a percent of GDP, copayments will increase by 0.6 percentage points over 2010–30 since assumed excess cost growth raises the spending ratio to GDP.

Second, premium contributions are projected to increase as a percent of GDP, from 4.4 percent in 2010 to 6.6 percent in 2030.¹² For HC, this occurs mainly because excess cost growth raises HC spending regardless of age cohort, thereby bringing up premium contributions relative to GDP. For LTC, premium contributions will increase in parallel with total spending owing to their fixed financing share (45 percent). This will result in a substantial increase in the contribution rate, i.e., premium contributions as a percent of labor earnings (for the working age population) or pension benefits (for the retirees).¹³ For HC, the increase in the contribution rate will be disproportionately higher for the working-age population, whose contributions will increase from 3¾ percent of labor income in 2010 to 12 percent in 2030 (Figure 11). Despite this, premium contributions as a percent of health spending are projected to decrease from 48 percent in 2010 to 43 percent in 2030. This is mainly because population ageing reduces the share of the working age population who contributes much more than retirees.

View Full Size

Japan: Premium Contributions for Health Care by Age Cohort, 2010–30

(In percent of earnings or pension benefits)

Citation: IMF Working Papers 2014, 142; 10.5089/9781498365277.001.A001

Sources: Ministry of Health, Labor, and Welfare; the National Institute of Population and Social Security Research; and IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

Third, government transfers need to increase because higher premium contributions and copayments fall short of the overall spending increase. Government transfers are projected to increase from 3.6 percent of GDP in 2010 to 7.1 percent of GDP in 2030. To put this figure in perspective, the “financing gap” of 3.5 percentage points of GDP by 2030 is equivalent to raising the consumption tax rate by 7 percentage points.

Comparison of spending projections with other studies

To conclude this section, we compare our projections with those made by the government and other researchers (Table 2). Key observations are as follows.

Table 2.

Japan: Comparison of Projected Increases in Health Spending, 2010–60

(In percentage points of GDP)

Sources: references cited; and IMF staff estimates.

^1/http://www.mhlw.go.jp/seisakunitsuite/bunya/hokabunya/shakaihoshou/dl/shouraisuikei.pdf. Spending projection is available only for 2015, 2020, and 2025. Projection for 2030 is calculated by authors with linear extrapolation.

^2/Uses OECD spending data based on SHA. Computed by the authors from Statistical Table 12a of IMF (2014), which shows an increase in Japan’s public health spending by 1.8 percentage points of GDP during 2014–30.

Table 2.

Japan: Comparison of Projected Increases in Health Spending, 2010–60

(In percentage points of GDP)

				2010-30			2010-60
			Definition of spending	Total	Ageing / demographic	Other	Total	Ageing / demographic	Other
Our projection
	Without healthy ageing
		Excess cost growth = 0%	HC + LTC, public + private	3.4	3.4	0.0	6.8	6.8	0.0
		Excess cost growth = 1%	HC + LTC, public + private	6.2	3.4	2.8	17.2	6.8	10.4
		Excess cost growth = 2%	HC + LTC, public + private	9.6	3.4	6.2	34.1	6.8	27.3
		Excess cost growth = 1%	HC + LTC, public only	5.6	3.1	2.5	15.6	6.2	9.4
	With healthy ageing
		Excess cost growth = 0%	HC + LTC, public + private	2.5	2.5	0.0	…	…	…
		Excess cost growth = 1%	HC + LTC, public + private	5.1	2.5	2.6	…	…	…
		Excess cost growth = 2%	HC + LTC, public + private	8.2	2.5	5.7	…	…	…
The government projection, March 2012 1/
	No reform scenario		HC + LTC, public only	4.2	…	…	…	…	…
	Reform scenario		HC + LTC, public only	5.1	…	…	…	…	…
Fiscal monitor (IMF, 2014) 2/			HC + LTC, public only	2.2	2.2	0.0	…	…	…
OECD (De La Maisonneuve and Oliveira Martins, 2013) 2/
	Cost pressure scenario		HC + LTC, public only	2.8	0.8	2.0	7.7	1.1	6.6
	Cost containment scenario		HC + LTC, public only	2.1	0.8	1.3	3.3	1.1	2.2
Ueda and others (2011)			HC + LTC, public + private	3.9	…	…	8.8	…	…
Ueda and others (2014)			HC + LTC, public + private	4.1	…	…	8.6	…	…
			HC + LTC, public only	3.7	…	…	7.9	…	…

Sources: references cited; and IMF staff estimates.

^1/http://www.mhlw.go.jp/seisakunitsuite/bunya/hokabunya/shakaihoshou/dl/shouraisuikei.pdf. Spending projection is available only for 2015, 2020, and 2025. Projection for 2030 is calculated by authors with linear extrapolation.

^2/Uses OECD spending data based on SHA. Computed by the authors from Statistical Table 12a of IMF (2014), which shows an increase in Japan’s public health spending by 1.8 percentage points of GDP during 2014–30.

View Table

Table 2.

Japan: Comparison of Projected Increases in Health Spending, 2010–60

(In percentage points of GDP)

				2010-30			2010-60
			Definition of spending	Total	Ageing / demographic	Other	Total	Ageing / demographic	Other
Our projection
	Without healthy ageing
		Excess cost growth = 0%	HC + LTC, public + private	3.4	3.4	0.0	6.8	6.8	0.0
		Excess cost growth = 1%	HC + LTC, public + private	6.2	3.4	2.8	17.2	6.8	10.4
		Excess cost growth = 2%	HC + LTC, public + private	9.6	3.4	6.2	34.1	6.8	27.3
		Excess cost growth = 1%	HC + LTC, public only	5.6	3.1	2.5	15.6	6.2	9.4
	With healthy ageing
		Excess cost growth = 0%	HC + LTC, public + private	2.5	2.5	0.0	…	…	…
		Excess cost growth = 1%	HC + LTC, public + private	5.1	2.5	2.6	…	…	…
		Excess cost growth = 2%	HC + LTC, public + private	8.2	2.5	5.7	…	…	…
The government projection, March 2012 1/
	No reform scenario		HC + LTC, public only	4.2	…	…	…	…	…
	Reform scenario		HC + LTC, public only	5.1	…	…	…	…	…
Fiscal monitor (IMF, 2014) 2/			HC + LTC, public only	2.2	2.2	0.0	…	…	…
OECD (De La Maisonneuve and Oliveira Martins, 2013) 2/
	Cost pressure scenario		HC + LTC, public only	2.8	0.8	2.0	7.7	1.1	6.6
	Cost containment scenario		HC + LTC, public only	2.1	0.8	1.3	3.3	1.1	2.2
Ueda and others (2011)			HC + LTC, public + private	3.9	…	…	8.8	…	…
Ueda and others (2014)			HC + LTC, public + private	4.1	…	…	8.6	…	…
			HC + LTC, public only	3.7	…	…	7.9	…	…

Sources: references cited; and IMF staff estimates.

^1/http://www.mhlw.go.jp/seisakunitsuite/bunya/hokabunya/shakaihoshou/dl/shouraisuikei.pdf. Spending projection is available only for 2015, 2020, and 2025. Projection for 2030 is calculated by authors with linear extrapolation.

^2/Uses OECD spending data based on SHA. Computed by the authors from Statistical Table 12a of IMF (2014), which shows an increase in Japan’s public health spending by 1.8 percentage points of GDP during 2014–30.

• The latest government projection envisages an increase of 4–5 percentage points of GDP for public health spending over the next two decades. Spending is larger in the reform scenario than in the no-reform scenario because the former incorporates reorganization of the health care system, which aims to reduce the average length of hospital stays by increasing the number of doctors and nurses. These projections are more pessimistic than our projection with no excess cost growth, but more optimistic than the one with 1 percent excess cost growth.

• The IMF Fiscal Monitor (IMF, 2014) projects public health spending to rise by about 2 percentage points of GDP over 2010–30. This is more optimistic than our projections mainly for two reasons. First, the Fiscal Monitor projection envisages zero excess cost growth for Japan, since it assumes convergence of the long-run excess cost growth toward the average during the last three decades (as discussed above, Japan’s average excess cost growth was about zero during this period). Second, the Fiscal Monitor projection envisages healthy ageing, resulting in a smaller ageing effect than our projections. Thus, the Fiscal Monitor projection is broadly in line with our scenario with healthy ageing and zero excess cost growth.

• The projection by the OECD (De La Maisonneuve and Oliveira Martins, 2013) envisages public health spending to rise by 2–3 percentage points of GDP for 2010–30 and 3–8 percentage points for 2010–60. Their projection is more optimistic than ours mainly because of a smaller ageing effect.

• The long-run projections by the Fiscal System Council (2014a) and Ueda et al. (2011, 2014) are more pessimistic than our projection with no excess cost growth, reflecting the increasing trend of utilizing in-house long-term care services and labor cost, but more optimistic than the one with 1 percent excess cost growth.¹⁴

Budget caps

Introducing budget caps for hospitals or health care providers can help contain health spending. Budget caps are a typical macro-level control, and refer to limits on overall health care spending for hospitals and practitioners. Case studies indicate that they were used to contain cost increases in many successful reform episodes (Clements and others, 2012). Budget caps, however, must be designed carefully. They are a blunt instrument to contain costs and can limit access to health care, as evidenced by longer waiting times for elective surgery in Canada, Sweden, and the United Kingdom during the period of expenditure consolidation (Clements and others, 2012). Health care providers in Japan are currently not subject to a budget cap. Reflecting this, the OECD indicator for the stringency of the budget constraint is 0 for Japan (the indicator ranges between 0 and 6 with the score of 6 representing the most stringent), while the OECD average is 2.0.

Introducing budget caps would be challenging in Japan. While strict budget caps tend to be implemented in countries where public health providers play a predominant role (such as Italy and Sweden), public hospitals account for only about 30 percent of total hospital beds in Japan. Nevertheless, health budget targets exist in countries where health services are mainly provided privately; in these counties, budget overshooting is possible, but it triggers cost-containment measures (Paris and others, 2010).¹⁶ Thus, there would be scope for Japan to draw from successful practices in other advanced countries.

Supply constraints

Supply constraints, another type of macro-level controls, aim to regulate the volume of health care supply including in areas such as workforce and medical facilities and equipments. In Japan, supply constraints are imposed on both workforce and facilities, but they are less effective for the latter (Table 3):

Table 3.

OECD Countries: Health Care System Characteristics, 2011 or Latest Year Available

Source: OECD.

Table 3.

OECD Countries: Health Care System Characteristics, 2011 or Latest Year Available

	Japan	OECD median	Japan’s ranking	No. of countries
Physicians per 1,000 population (head counts)	2.2	3.3	29th highest	34
Total hospital beds per 1,000 population	13.4	4.2	1st highest	34
Curative (acute) care beds per 1,000 population	8.0	3.0	1st highest	34
MRI units, total, per million population	46.9	10.7	1st highest	32
CT scanners, total, per million population	101.3	16.1	1st highest	32
Doctors consultations, number per capita	13.1	6.6	2nd highest	34
Average length of stay at hospitals, days	18.2	7.2	1st highest	34

Source: OECD.

View Table

Table 3.

OECD Countries: Health Care System Characteristics, 2011 or Latest Year Available

	Japan	OECD median	Japan’s ranking	No. of countries
Physicians per 1,000 population (head counts)	2.2	3.3	29th highest	34
Total hospital beds per 1,000 population	13.4	4.2	1st highest	34
Curative (acute) care beds per 1,000 population	8.0	3.0	1st highest	34
MRI units, total, per million population	46.9	10.7	1st highest	32
CT scanners, total, per million population	101.3	16.1	1st highest	32
Doctors consultations, number per capita	13.1	6.6	2nd highest	34
Average length of stay at hospitals, days	18.2	7.2	1st highest	34

Source: OECD.

• The government controls the health care workforce with enrollment quotas for medical schools.¹⁷ Reflecting this, the number of doctors per capita is lower in Japan than in other OECD countries.

• It also regulates the number of hospital beds by restricting installation of new units beyond benchmarks set at regional levels. Despite this, Japan has the highest number of beds per capita among OECD countries. Since hospitals are granted discretion over the purchase of medical equipments (Hashimoto and others, 2011), high-tech medical equipments are the most available in Japan on a per-capita basis.

Redesigning supply constraints, especially targeted at the number of beds per capita and aimed to rationalize the abundance of medical equipments, may help contain health spending growth in Japan. According to the OECD indicator that measures the degree of regulation of health workforce and equipment, Japan’s score is 1.5, lower than the average of 2.9 (the indicator ranges between 0 and 6 with the score of 6 representing the highest degree), suggesting potential room for additional supply constraints. The number of beds per capita would be an important target to contain per-capita health spending by the elderly, because these two indicators are positively correlated at prefecture levels. Japan could learn from the experience in Canada and Finland, where reduction of the number of beds helped contain health care spending. Nevertheless, supply constraints need to be carefully designed since they can be less effective when providers can circumvent or offset the effect of controls (Clements and others, 2012).

Price controls

Price controls, currently a primary lever for the Japanese government to control health spending, need to be designed carefully. The government determines prices of services covered by health and long-term care insurance, with prices adjusted every two years for health care and every three years for long-term care. The government restrained the price increase during 2003–08, and succeeded in reining in health spending especially in 2006 (Figure 12). Nevertheless, for HC, the price control measures in 2006 and 2008 were followed by a pickup in spending in the subsequent year. This seems consistent with a hypothesis that price controls are circumvented by service providers’ reactions to increase service volumes or direct patients to higher-cost services (Docteur and Oxley, 2003). Also, at times, price adjustments across HC services resulted in unintended consequences and exacerbated distortions in the health system (the 2013 National Council Report). In this context, price controls should be carefully designed and can be complemented by other options.

View Full Size

Japan: Price Index of HC and LTC Services and Per-capita Health Spending, 2003–11

(In annual percentage change)

Citation: IMF Working Papers 2014, 142; 10.5089/9781498365277.001.A001

Sources: Ministry of Health, Labor, and Welfare; and IMF staff estimates.1/ Converted to real terms using the GDP deflator.

• Download Figure
• Download figure as PowerPoint slide

Reforming provider payment arrangements

A micro-level reform of provider payment arrangements could help contain health spending by correcting incentive problems. Specifically, shifting from a fee-for-service arrangement (providers are paid according to the volume of services provided to patients) to a Diagnostic-Related Group (DRG) arrangement (providers are paid a fixed amount based on the diagnostics of a patient, regardless of the volume of services provided) would reduce supplier-induced health care demand and hence help contain spending growth.

With a fee-for-service arrangement used for compensating doctors and limited use of DRG-type arrangements for hospitals, Japan’s system tends to encourage supplier-induced demand (Joumard and others, 2010). According to the OECD indicator to measure the strength of providers’ incentives to increase the volume of care, Japan’s score is 5.7, much higher than the OECD average of 3.1. This may explain, at least partially, the fact that doctor consultations per capita and the average length of hospital stay in Japan are very high compared with other advanced economies (Table 3). Moreover, the positive correlation between the average length of hospital stay and the number of beds per capita at prefecture levels (Figure 13, left panel) may reflect hospitals’ desire to minimize unoccupied beds. Given that the average length of hospital stay is positively associated with per-capita health spending of the elderly at prefecture levels (Figure 13, right panel),¹⁸ mitigating the incentive problem could be effective in reining in health spending growth.

View Full Size

Japan: Per-capita HC Spending by the Elderly, Their Length of Hospital Stay, and the Number of Beds Per-capita, by Prefecture ^1/

Citation: IMF Working Papers 2014, 142; 10.5089/9781498365277.001.A001

Source: Ministry of Health, Labor, and Welfare.^1/ Scatter plots for 47 prefectures.

• Download Figure
• Download figure as PowerPoint slide

Introducing gatekeeping and improving public management

As a micro-level reform, introducing gatekeeping arrangements could help contain spending without undermining health outcomes in Japan. Gatekeeping requires patients to register with a general practitioner and/or to receive referral from a general practitioner to access specialist care (Joumard and others, 2010). It can save healthcare costs by preventing unnecessary treatments. While 20 out of 29 OECD countries use gatekeeping arrangements at least to some extent, there is no such arrangement in Japan as patients are guaranteed “free access”: they can choose hospitals and health care providers regardless of locations and specializations, and are not required to obtain referrals from general practitioners to receive secondary- or tertiary-level care.¹⁹ The need for gatekeeping is emphasized in the 2013 National Council Report.

More broadly, there is room to improve public management and coordination of health care resources. The high share of private hospitals and providers, relatively lax supply controls, and the limited effectiveness of price controls seem to have led to suboptimal allocations of health care resources. For example, the share of hospitals that can offer highly acute care was about 40 percent in 2010, much higher than the range considered as ideal by the government (under 20 percent); a large number of long-term care beds still remain in hospitals, not in long-term care facilities as was intended when the long-term care insurance was launched (OECD, 2009); and a supply shortage of in-home care induces provision of long-term care at hospitals. Also, supply and demand mismatches exist across doctor specialties and regions (Shibuya and others, 2011; the 2013 National Council Report). While medium-term planning and target setting by prefecture governments started in 2006, it has not led to a dramatic decline in the average length of hospital stay, and the mismatches still persist. As proposed by the 2013 National Council Report and Shibuya and others (2011), a possible option is to strengthen the regulatory power of prefecture governments to monitor and coordinate resource allocations.

Other options

Policies to promote the use of generic medicines can save health care cost. Japan’s utilization of generic medicines was 24 percent in 2009, while it was 89 percent in the United States, 75 percent in Germany, and 71 percent in the United Kingdom (Sheppard, 2010). Further incentivizing and advertizing the use of generic drugs should be pursued.

Prevention of lifestyle related diseases can promote healthy ageing. Major risk factors for non-communicable diseases, such as high blood pressure, smoking, and high concentrations of blood glucose, can be further reduced for the Japanese (Ikeda and others, 2011). The ongoing public campaign for prevention of lifestyle related diseases can be supplemented by greater use of monetary incentives, such as higher tobacco taxes (OECD, 2009).

Raising patient copayments

Raising copayment rates would contain public health spending while reducing excess demand for health care and long-term care. Compared with other OECD countries, the copayments in Japan are relatively low both in levels and as percent of total spending (Figure 14), indicating potential room to increase copayment rates. Research indicates that higher copayment rates are also expected to discourage unnecessary demand, and thus help to contain total health spending.²¹ Increasing copayments, however, should be designed carefully since it can weaken the effectiveness of public health insurance to protect against health risks and limit health care access by the poor (OECD, 2009). An option to address this is to further differentiate copayment rates based on income levels.

View Full Size

OECD Countries: Out of Pocket Payments, 2011 or Latest Year Available ^1/

Citation: IMF Working Papers 2014, 142; 10.5089/9781498365277.001.A001

Source: OECD.^1/ Including for LTC.

• Download Figure
• Download figure as PowerPoint slide

Raising premium contributions

Raising premium contributions have benefits and costs. As a benefit, the link between contributions and income can be tightened to ameliorate negative effects of health care reforms on income inequality. There is scope to strengthen the link in Japan; for example, ceilings on individual premium contributions can be increased, as proposed by the 2013 National Council Report. As a major disadvantage, higher premium contributions could increase the labor tax wedge (the sum of personal income tax and employee plus employer social security contributions), thereby reducing labor supply and demand. This can be mitigated by introducing a progressive contribution (or payroll tax) schedule (i.e., the contribution rate rises as income rises), which can reduce tax wedges for low-skilled workers where distortions are the largest (IMF, 2012).

Premium contributions can also be increased by reducing inequality in contribution payments. Among others, the current system exempts dependent spouses of employees covered by employment-based HC insurance programs (Kyokai Kempo, Kumiai Kempo, and Mutual Aid Associations; see Box 1) from paying health care insurance premiums if their annual earnings are below 1.3 million yen (approximately US$13,000), while dependent spouses of self-employed workers (covered by the NHIP) are required to pay contributions. This also creates work disincentives, providing the exempted spouses with a strong incentive to work part-time and keep their labor earnings below the threshold.²² More generally, inequality in the contribution rates across health care insurers should be reduced, given that the average contribution rate for the NHIP is three times as high as that for employment-based Kumiai Kempo insurance programs (Ikegami and others, 2011).