Introduction

On 1 January 1995, a National Health Insurance Law was enacted in Israel. The Law defined a uniform package of services to be provided by the four sickness funds operating nationally in Israel. It envisioned a ‘managed competition’ scheme, where the sickness funds compete on the quality of the services as well as on some ‘extras’, to attract consumers. Unlike the Dutch or the German systems, Israelis do not pay contributions directly to their sickness funds, so competition on price is not relevant. Under the law, health care is universal, and no (explicit) risk selection is allowed. Furthermore, open enrollment with an annual switching option is specified. Since 1998 the sickness funds are allowed to offer voluntary community-rated supplementary insurance policies to their members (only), which cover acute care services not covered by the package of benefits (such as alternative medicine, more physical therapy sessions, medicines not in the package, choice of surgeon etc.). These policies have become a major service on which sickness funds compete.

The incomes of the sickness funds consist of governmental transfers, insurees’ copayments, and the supplemental insurance premiums. The governmental transfers include the age-based risk-adjusted capitation payments, complemented by a five severe conditions-specific risk sharing (Aids, Thalasemia major, Gaucher, Hemophilia, and End-stage renal disease). The Israeli risk-adjustment scheme is incomplete, and thus might provide perverse incentives to the sickness funds, for two reasons (for further details on the Israeli system see Shmueli, Chernichovsky, and Zmora, Reference Shmueli, Chernichovsky and Zmora2003). First, it is based on age alone, so that sicker (healthier) than average persons in each age group constitute a predictable loss (profit). Second, the capitation rates are calculated based on relative quantities of inpatient care and ambulatory care, weighted by their relative shares in the total sickness funds' expenditures. The omission of other types of care (e.g. medicines) introduces biases in the capitation rates.

As envisioned by the managed competition scheme, it was expected that some level of consumer mobility would be observed. Consumer mobility is traditionally taken to reflect viable and active competition, which might assure good quality of care and good quality of service. During the first three years of the law, about 4% of the total population switched sickness funds annually (switching does not include newborns, deaths, and first registration of new immigrants). This figure reflected two factors: first, the sickness funds began aggressive marketing campaigns to sign up members (e.g. in malls or bus stations). Second, the law cancelled previously established effective restrictions on switching. Before 1995, two sickness funds exercised risk selection, and a third sickness fund was the only option available to certain parts of the population because of socio-political affiliation or geographical location. All those who had wished to switch but had not been able to do so previously, were able to fulfill their wish during the first years after 1995. Since 1 January 1998, the intention to switch sickness fund must be registered in a Post Authority office, and the annual rate of switchers has been quite fixed at 1%.

The reasons for the low switching rates in Israel (and in Belgium and the Netherlands as well) are discussed, in a five country comparative context, by Laske-Aldershof et al. (Reference Laske-Aldershof, Schut, Beck, Greß, Shmueli and Van de Voorde2005). The conclusions are that, abstracting from reasons commonly associated with inertia, status quo bias, and reservations about changes, the limited choice options facing the Israeli population reduces the potential net benefits of switching, leading to low mobility. The limited choice setting results from the apparent similarity across the four sickness funds, both in the basic and the supplementary insurance domains.

All four Israeli sickness funds operate as variations of managed care organizations, characterized by contracted providers, the use of clinical and administrative guidelines, and performance and outcomes reviews. The biggest sickness fund (HSC), with a 56% market share in 2000, operates like a staff-model HMO, owns eight general hospitals, and most of the primary and secondary care is provided by salaried physicians in the fund’s clinics. The small funds (market shares 23% (MHS), 11% (MSF), and 10% (LSF)) have no hospitals, and contract selectively with hospitals, with independent physicians and institutions for acute care. Many independent physicians and most hospitals contract with more than one sickness fund. This results in different sets of attributes of sickness funds.

An additional possible barrier to comsumer mobility among Israeli sickness funds, and a known way for implicit risk-selection, has been supplemental insurance regulations. The supplemental insurance policies offered by the sickness funds are community-rated with no rejection of applicants. However, a ‘waiting period’ of 3–9 months, depending on the particular service, is imposed on new buyers. Switching sickness fund bears, consequently, transaction costs in the supplemental insurance market as well, which further reduce the net potential benefits of mobility.

The present paper focuses on the question of who are those who are more likely to switch sickness fund, and to discuss the implications for the Israeli system.

The importance of the question: ‘who switches?’

Some may argue that an annual rate of 1% of switchers is too low to justify a concern about who are the switchers. We believe that the issue is important, and perhaps even more so just because of the low rate of switching, for several reasons. First, because of the highly skewed distributions of health, medical expenditure, and income, and the strong correlations among them, a 1% of the population who switches sickness funds might concentrate much larger shares of (ill) health, income or medical care costs. Second, as mentioned above, competition among the sickness funds can be effective if consumers search and compare options. In order for public policy to spread information and to enhance switching efficiently, the identity of the (non)switchers must be known. Third, switching behavior might affect the quality of care offered by the sickness funds. If, for instance, older and sicker persons are less inclined to switch, sickness funds may find it unprofitable to invest in high quality care for those patients, and to direct their efforts in order to compete for persons who are more likely to switch. Finally, switching behavior is important for risk adjustment. Risk adjustment aims to make the sickness funds indifferent with respect to the insurees’ level of risk and expected costs and to the quality of care of different services. Thus, with reference to the example just given, risk adjustment should search to neutralize such distorting incentives.

Generally, three types of switches among sickness funds can be identified. The first type consists of ‘technical switching’. These switches are the result of other, mostly exogenous, decisions. For example, one may switch sickness fund because of moving to another city or region. Technical switches are expected to concentrate in young ages, when several key lifetime decisions are made. The second type of switches reflects choices made by individuals based on their preferences vis-à-vis the different attributes of the optional sickness funds. These are the classical consumer choice cases. The attributes may include price (as mentioned above, not relevant in the Israeli system), waiting times for appointments, accessibility of specialist services, administrative complexity, proximity of providers, and other determinants of the quality of services. Different bundles of attributes may appeal to different personal characteristics such as level of education, income, family composition, or age. Most of the studies have focused on these switches; including the estimation of price elasticities (Schut et al., Reference Schut, Gres and Wasem2003; Stombom et al., 2002; Laske-Aldershof and Schut, Reference Laske-Aldershof and Schut2002). These switches reflect consumers’ free choice and serve as a competitive threat to the sickness funds. Earlier studies generally found that younger age groups, more educated and richer persons were more likely to switch sickness funds/health plans. We reiterate that price sensitivity has been an important factor in all those studies, and that the present study offers an opportunity to examine the above effects when price is not relevant.

The third type of switches includes those related directly to health state and to the level of expected medical expenditure. These switches might be initiated by the individual, searching for, for example, better quality or quantity of services and care, a reason which appeals more to sick individuals than to the young and healthy. This is a demand-side factor, which is a variant of adverse selection. They might be, however, the result of the sickness funds’ strategy, namely (implicit) risk selection activities to induce sick individuals to leave, facing an incomplete risk-adjustment scheme. Earlier research shows that healthier persons are more likely to switch, probably reflecting the correlation of good health with second-type switchers' higher education, better economic status, and younger age.

From societal perspective, and from risk-adjustment point of view, the third type of switches is the most interesting one. However, it is the most difficult to analyze, since information on health status and expected and actual expenditure is needed. Either as an adverse selection or risk selection act, such switches reduce welfare, since they distort the incentives for high quality care for sick persons, and might cause financial difficulties to certain plans.

Data and methods

We use information from the National Insurance Institute’s Central Health Registry, covering the entire Israeli population. The Registry contains selected personal characteristics of residents by their sickness fund, including an identification of the switchers. Children under the age of 18 are registered with the sickness fund of either their mother or their father, and switch with them.

For technical reasons, we were forced to conduct two separate analyses. In the first, using the information for the years 1999–2000, we focused on the bivariate relationships between each of the following personal characteristics and switching behavior: age, gender, labor income, employment status, and whether the individual receives income support, disability, or unemployment benefits (we note that the receipt of the above benefits stops at age 60 for women and 65 for men, and they are replaced by old-age benefits).

With respect to each of the above personal traits, in each sex–age group in the sickness fund of origin, we calculated the ratio between of the means of the personal trait under discussion (e.g., the percent receiving disability benefit) among the movers and among the stayers. The totals were calculated as weighted averages of the group-specific rates. We tested statistically these ratios, by testing the equality of the means/proportions between movers and stayers. A t-test was used to test the equality of mean incomes, and chi-square tests were used to test the equality of the proportions receiving income support, disability, or unemployment benefits.

The second analysis was a multivariate logit analysis. The population was the entire Israeli population above the age 21 as of 1 May 2006. The dependent variable was defined as 1 if the person had switched sickness funds between 1 January 2005 and 30 April 2006, and 0 otherwise. The explanatory variables included the characteristics mentioned above: age (7 groups: 18–24, 25–34, 35–44, 45–54, 55–64, 65–74 and 75+); sex (1 = women); the interactions age–sex; average relative labor-income in 2004 (5 groups: no income, 0–0.5 the national mean wage, 0.5–1 the national mean wage, 1–2 the national mean wage, and more than twice the mean wage); marital status (1 = married); recipient of disability benefit (1 = yes); recipient of unemployment benefit (1 = yes); recipient of income support (1 = yes). In addition, we included the following four characteristics: sickness fund (for switchers – the sickness fund of origin) (4 sickness funds: HSC, MHS, MSF, and LSF); number of children under 18 (6 groups: no children, 1–2, 3, 4, 5–7, 8+ children); whether the person has changed residence during the period 1 January 2005–30 April 2006 (1 = yes); and whether the person has changed marital status during the same period (1 = yes). The two latter variables are meant to control, at least roughly, for the ‘technical movers’, namely persons who have switched sickness funds as a result of an exogenous change such as moving to another residence or changing their marital status.

Results

Bivariate analysis: 1999–2000

Table 1 presents the percent switching sickness fund in the years 1999 and 2000 by age and gender (the far right column presents the size of the age groups in the population in November 1999). The overall rate is 1%, which has been stable for the period since 1998. The mobility rates are quite similar among men and women in most age groups. Particularly high rates are found in age groups 15–34 and 0–4. These switches are probably technical, of young singles and couples with small children. The propensity to move declines steadily with advanced age. Among the 75+ elderly, only 0.3% switched sickness fund in 2000. Although the rates differ across the four sickness funds, that pattern is seen in all of them.

Table 1

Percent leaving their sickness funds by age and gender

Additional general findings (not presented in Table 1) indicate that, while the mean age of the insured population was 32 in 2000, the mean age of movers was 26. It is interesting to note that in the ‘older’ two sickness funds the mean age of leavers was higher than that of joiners, while in the other two the opposite was found. This finding implies that in the long run, switching sickness fund is going to change the age structure of the four sickness funds, rendering it more uniform.

Table 2 presents the mean labor-income of switchers relative to the stayers by age–sex in their sickness fund of origin in 1999 and in 2000. Among salaried persons, both men and women, switchers have about 20% lower wages than the stayers. With advanced age, the switchers have a relatively lower wage than in younger age groups. For example, in 1999, in the 35–44 age group, the mean wage among movers was 95% of the mean among stayers, while in the 65–74 age group the proportion was only 60%. A similar pattern was found in all sickness funds. In most of the cells, the ratio of the movers’ to the stayers’ wages was significantly smaller than one. Only in the 15–24 age group, and among men in particular, movers are richer than stayers. A similar picture appears among the self-employed (because of small cells we combined the upper age groups into 45+).

Table 2

The ratio of mean income of movers over that of stayers

Notes:

* = The ratio is different than 1 at p = 0.01.

# = The hypothesis that the ratio equals 1 cannot be rejected at p = 0.01.

The findings that the switchers were poorer than average, and that the gap increased with age, are confirmed by Table 3. In 1999, the rate of income support recipients among switchers was 1.4 times the rate among stayers (the rate of recipients in the population was 2.3%). In 2000, the ratio was 1.19 (and the rate in the population was 2.48%). This ratio increased with age in both years and among men and women. For example, it was 1.56 in 1999 in the 25–34 age group, and reached 2.01 in the 55–65 age group. All ratios are significantly larger than 1.

Table 3

The ratio of the percent of persons with income support among movers over that among stayers

Notes:

* = The ratio is greater than 1 at p = 0.01 one-side chi-square test.

Table 4 shows that the relative rate of unemployment benefit recipients among movers was around 1 (and not significantly different from 1) in 1999 (the population rate was 1.52%), and well (significantly) above 1 in all cells except two in the year 2000 (the population rate was 1.34%). For the entire population, for example, the ratio was 1.27.

Table 4

The ratio of the percent of persons with unemployment benefit among movers over that among stayers

Notes:

* = The ratio is greater than 1 at p = 0.01 one-side chi-square test.

♯ = The hypothesis that the ratio equals 1 cannot be rejected at p = 0.01.

As the final examination of the first analysis, Table 5 presents the relationship between switching and the receipt of disability benefit. Overall, the rate of recipients of disability benefit among movers was 0.73 and 0.63 the rate among stayers, in 1999 and 2000 respectively (the rates in the population were 2.3% and 2.31% respectively). In other words, disabled individuals were less likely to switch. However, the ratio increased somewhat with age in both years and among men and women. In particular, it was larger than 1 in 1999 in the age groups 35–54. In one sickness fund, the 1999 overall ratio was 1.46, and in all age groups the ratio was higher than 1, reaching 2.76 in the 45–54 age group. In another sickness fund, the ratio was greater than 1 for ages above 25. The 2000 figures were similar.

Table 5

The ratio of the percent of persons with disability benefit among movers over that among stayers

Notes:

* = The ratio is different from 1 at p =0.01.

# = The hypothesis that the ratio equals 1 cannot be rejected at p =0.01.

Multivariate analysis: 2005–2006

Table 6 presents the means and the odds-ratios of the likelihood to switch sickness funds by the personal characteristics included in the logistic analysis. Total population aged 21+ on 1 May 2006 amounted to 4,312,074 individuals.

Table 6

Means and odds-ratios in the logistic analysis of switching sickness funds¹

Notes:

1 During the period 1/05–4/06.

The likelihood to leave one's sickness fund varies across sickness funds. Members of the LSF have the highest likelihood to switch, followed by the MSF (the reference), HSC and the MHS has the lowest rate of leavers. Persons who, during the period 1/05–4/06, made a change in their marital status or place of residence are 37% and 58% respectively more likely to be (probably technical) switchers during the same period.

Controlling for these variables, among both men and women, advanced age is related to lower propensity to switch sickness funds. Those aged 65+ are about 80% less likely than persons aged 21–24 to switch. The age effect is continuous – without a particular threshold – over all age groups, both among men and women. However, the age effects are different between men and women.

Women aged 25–54 are less likely than men in the same age group to switch. In ages 55+, men and women are equally likely to switch. At ages 21–24, however, women are more likely to switch than men. This might be a technical difference related to the different military experience of men and women. Soldiers (typically ages 18–20 or 21) receive their medical care through the military and are not insured by the sickness funds. Upon discharge, they register with the sickness fund they choose, being considered as switchers if their choice is different than their parents' sickness fund. A higher proportion of women than men do not serve in the army, and women serve a shorter period (2 vs. 3 years). Consequently, in that age group, a higher rate of women need to register with a (civilian) sickness fund, having a greater likelihood to switch (from their parents' sickness fund of choice).

Married persons (who got married before 2005) are 45% more likely than non-married to switch sickness funds during the period 1/05–4/06. Part of these moves is (delayed) technical, another part might be related to switches following pregnancy and/or giving birth.

As we saw in the bivariate analysis, switching sickness funds is an inferior good. Persons with labor-income higher than twice the average wage are 38% less likely to switch than persons with income of up to half the average wage. The labor-income gradient is continuous over the four relative income groups. Similarly, those receiving unemployment benefit or income support are 12% and 44% respectively more likely to switch. However, the recept of disability benefit does not affect the probability to switch.

Finally, persons with more children aged up to 18 are more likely to switch. For example, persons with eight children or more are three times more likely than childless persons to switch. The gradient is continuous over the entire range of the number of children.

Discussion

The findings from both the bivariate and the multivariate analyses for both 1999–2000 and 2005–2006 are very similar, and indicate that the 1% of the population who switch sickness fund in Israel yearly, is far from being a random draw from the population with respect to several important characteristics.

The tendency to switch drops continuously with advanced age. A proportion of the moves among young persons and couples is probably technical. However, the same tendency exists even after controlling for changes in marital status or relocation. ‘Non-technical’ moves among young persons/couples might reflect comsumer choices and may have originated from a search for better quality or availability of pediatric or gynecological care. On the other hand, switching costs in terms of switching providers and retrieving administrative information are higher among older persons who use medical services more intensively than the young. Lower tendency of elderly persons to switch sickness funds was found in Germany, the US, and the Netherlands as well (Schut et al., Reference Shmueli, Chernichovsky and Zmora2003; Stombom et al., 2002; Laske-Aldershof and Schut, Reference Laske-Aldershof and Schut2002).

The age-gradient is different, however, between men and women. In particular, women during their fertility age are less likely to switch. It is well known that, in this age group, women use medical services more intensely than men, and their reluctance to switch might stem, as discussed above for elderly persons, from higher costs related to switching providers.

Poor persons (with low relative labor-incomes, and/or receiving income support or unemployment benefit) are over-represented among the switchers. Furthermore, that over-representation increases with age, at least among wage earners. The higher tendency to switch among families with a large number of children might be seen as another indicator of that conclusion, since large numbers of children and poverty are correlated. This means that switching sickness funds is largely an inferior good. This conclusion is confirmed by the multivariate analysis. This is inconsistent with other international studies’ results, suggesting that switching, when one of the factors is the premium level, is a normal good.

The poor persons’ switches are hardly expected to be technical moves. They are also not a result of higher price-elasticity, since there is no price competition on the premium in Israel. Copayments, which are set by the individual sickness funds, are approximately equal across sickness funds as well, so that a search for lower copayments can hardly be a reason to move (in addition, recipients of income support and disability benefit were exempted from paying copayments).

One possible demand-side explanation for the inferiority of switching is the lower ownership rate of supplemental insurance among poor persons (Achdut et al., 2005). As was mentioned earlier, newcomers who buy supplemental insurance must bear a waiting period, which is expected to discourage moves. The lower ownership rate thus reduces the transaction costs of moving for poor persons.

Further insight into the dynamics of switching might be gained by considering possible supply-side and risk-selection strategies. Children represent a predictable profit under the present capitation formula, due to the over-paying rate for that age group. Zmora (Reference Zmora2004) estimated that the rate is 90% above the mean medical expenditure in ages 1–4, and 60% in ages 5–14. Young couples with (many) young children are thus a target for selective marketing, and the sickness funds try to attract them. Naturally, such a strategy should have been adopted by all sickness funds, so that no switching should be observed at equilibrium, but obviously certain funds succeeded in that more than others. Many children, according to this argument, are no longer an indicator of poverty, but represent predicted profits by the sickness funds.

Two subpopulations are typically characterized by large families: Arabs and religious (mainly orthodox) Jews. These two populations are also over-represented among the poor (NII, 2005). While the two subpopulations cannot be individually identified in the NII health registry, locations with a majority of each can be.

Arab residents, with, on average, bigger families, worse health, and lower incomes have been historically concentrated with the HSC sickness fund, as a result of risk selection by the other funds prior to 1995. When premiums, paid directly to the sickness funds, are proportional to family income, Arab families are clearly a predictable loss. The HSC, with an ideological mission to serve the entire Israeli population, accepted all applicants. With the introduction of the national health system in 1995 with its open enrollment principle, the services of the other sickness funds became more accessible, and the over-paying risk-adjusted rate for children made young Arab families particularly attractive. And, indeed, the switching rate in Arab locations in 2005 was 1.4%, compared to 1% switchers in Jewish locations.

Orthodox Jews are another population, which has become attractive to the sickness funds because of high fertility. And, indeed, in typically young orthodox families' locations, the switching rate in 2005 was 4%, compared to the overall 1%.

Elderly persons, on the other hand, are predictable losses under the current capitation rates (Zmora, Reference Zmora2004). The low switching rates among the elderly are thus explained by both the (demand-side) higher switching costs, and the (supply-side) non-attractiveness to the sickness funds.

Another type of explanation for the higher propensity to move among poor persons lies in the correlation between income and health. It is well known that in Israel, as in other countries, poor persons are sicker than rich persons, and use more health services (The Israel National Institute, 2003). According to Zmora (Reference Zmora2004), persons with up to eight years of education are twice as likely to be hospitalized as persons with 13 or more years. Unemployed persons consume 10% more medical care than employed, controlling for age. The medical expenditure on persons receiving disability benefit is twice the expenditure of persons who do not receive such benefits. Furthermore, sick persons are predictable losses under the incomplete Israeli risk-adjustment scheme, so that under-serving is not an illogical strategy of the sickness funds. The age-adjusted average medical cost for persons with cancer, heart disease, diabetes (on insulin) and asthma level 3 is 60% higher than the relevant age-adjusted average capitation rate.

The over-representation of the poor of all ages among switchers may thus reflect a higher tendency to move among sick individuals. These might be health-related moves of the third type mentioned above. While the literature concludes that the likelihood to move is lower among sick persons because of the (demand-side) high transaction costs of switching, originating from their greater involvement with both the administrators and providers of care (e.g. Strombom et al., Reference Strombom, Buchmueller and Feldstein2002), sick persons are expected to be more sensitive to the quality, availability or the quantity of the care they receive. They are also more knowledgeable on the quality of services because of their greater use of services. If sick persons feel they are being under-served, they might switch to another sickness fund, where they hope to find more and better care (although all sickness funds are expected to adopt similar risk-selection strategies). Our data are insufficient to relate directly to the effect of health on switching, but the above argument is supported by Berg et al. (Reference Berg, Rosen and Morgenstein2002), who documented the higher propensity to switch sickness funds of income-maintenance recipients for 1995–1997. When asked about the reason for their move, the main stated reason was dissatisfaction with the health services provided by the sickness fund.

While disability had no effect on switching in the multivariate analysis, the bivariate analysis provided an illustration of the working of the possible risk-selection strategy discussed above: the sickness fund with the highest (consistently across all age groups and in both years) over-representation of disabled persons among the switchers, contracts independent physicians to provide primary and secondary care. These independent physicians largely run solo-practices in their own homes or in rented apartment-clinics, with no special accessibility arrangements for disabled persons. Disabled persons are thus forced to switch providers – and often sickness fund – in order to gain physical accessibility to clinics. Because disabled persons, of all ages, are – under the Israeli risk adjustment scheme – predictable losses, that particular sickness fund profits from the implicit risk selection.

Conclusions

Switching sickness funds is rare in Israel. About 1% of the population switch sickness fund annually. Such a low mobility might indicate that competition is not effective in providing incentives for good quality of care and services. Public policy should remove barriers to move, and assure the dissemination of relevant information. An important barrier to move is the terms of supplementary health insurance offered by the sickness funds.

The main observation of the present study is that, unlike in other systems, in Israel, poor persons are more likely to switch. While several demand-side explanations were offered, one cannot rule out the possibility that because of an incomplete age-based risk-adjustment system, the poor – who are sicker than average in all age groups – are predictable losses, and their higher switching rates result from implicit risk-selection. On the other hand, young children are predictable profits, and thus favorably selected (attracted) by the sickness funds, inducing again high switching among them and their young parents. The higher switching rates among Arabs and orthodox Jews are consistent with that argument. Public policy should improve the risk-adjustment system, in order to provide incentives to the sickness funds to differentiate themselves as providing good quality care to all, including those sicker than average.