Pricing behavior in long term care markets: evidence from provider-level data for home help services

See e.g. the Forbes news article on 09/26/’17 about the main findings of the Genworth 2017 Cost of Care Survey: https://​www.​forbes.​com/​sites/​nextavenue/​2017/​09/​26/​the-staggering-prices-of-long-term-care-2017/​#78d6c68c2ee2. One of the noteworthy outcomes of this survey is that in 2017 the monthly cost for a private room in a U.S. nursing room rose by a yearly 5.5% and by almost 50% since 2004. The yearly increase of the cost for home health care was even higher and has been mainly attributed to changes in price drivers. In comparison, the U.S. Consumer Price Index rose by only 1.4% in 2017. Martin et al. (2011) show that for health care expenditures in general, price increases accounted for 60% of the growth in U.S. health care spending in 2009. Finally, Anderson et al. (2003) argue that the variability in health care spending among OECD countries is mainly due to differences in prices for health care goods and services.

See e.g. Bartlett and Phillips (1996) and Glendinning and Moran (2009).

Gaynor and Town (2012), discussing recent trends in health care markets in general, show that the upward trend in prices for health care services is associated with an increase in the concentration on the supply side of the market.

The impact of market concentration on prices has also been studied for hospital care. From the results found by Melnick and Keeler (2007) for the Californian hospital market, one obtains that a reduction of the degree of market concentration by 10%—using the sample average as the benchmark—yields a price decrease of 3.6%. Gaynor and Vogt (2000) present an overview of the pre-2000 studies on this issue. All of these studies consider a hypothetical “standard merger case” in which there is a market initially consisting of five identical hospitals and where two of these hospitals merge. Most of these studies find that such an increase of market concentration by 40% are associated with higher prices, ranging from 2 to 17%. A few studies however obtain a small negative relation between market concentration and hospital prices.

See e.g. Bresnahan (1989) for a thorough discussion on these issues.

The empirical evidence on the effects of competition on quality is rather mixed: while Lin (2015), based on a counterfactual analysis, concludes that entry deregulation does not improve the quality of care in U.S. nursing homes, Jung and Polsky (2014) find for the U.S. home health care market a nonlinear relationship between competition and various quality measures, indicating that too intensive competition hurts market outcomes. Taking a somewhat broader perspective, recent studies by Cooper et al. (2011) and Bloom et al. (2015) find that stronger competitive pressures lead to improvements in quality in the English hospital market.

See e.g. Glaeser and Shleifer (2001) and Besley and Ghatak (2005) for theoretical explanations for why the incentives for non-profit organizations divert from the incentives for for-profit companies.

Another paper on this issue, studying spillover effects between the different ownership types, is Grabowski and Hirth (2003). It shows for the U.S. case that a higher market share of non-profit nursing homes has a positive impact on the quality provided by competing for-profit homes, a result in line with the idea that the non-profit status acts as a quality signal thereby alleviating asymmetric information problems throughout the entire industry.

This percentage holds for municipalities in the estimation sample and for municipalities that are not included in the estimation sample; see Table 10 in the Appendix.

The market for privately financed home help in the Netherlands is half the size of the market for publicly financed home help (see Putman et al., 2016).

In 2007, there were 443 municipalities in the Netherlands with an average size of about 40.000 citizens. The 32 EMEA-regions consisted of about half a million people on average.

The size of the grant is based on population characteristics, except for the first year of decentralization. In this year the grant was based on expenditures by health care purchasing agencies within municipality boundaries in 2005.

In addition to opening up local markets to competition, most municipalities also replaced assessments for more expensive advanced home help by assessments for the cheaper basic form and charged user fees in order to curtail expenditures.

By contrast, under the EMEA health care purchasing agencies typically guaranteed incumbents a minimal turnover equal to 90% of turnover in the previous year.

What is interesting to note here is that the municipality typically selected more than one provider, which is rather uncommon for other goods and services purchased by the public sector via scoring or price-only auctions (see also Asker & Cantillon, 2008).

We have estimated our main specification for suppliers active in municipalities using regulated price auctions and, as expected, did not reject the null hypothesis of a non-relationship between market share and price. Results are available upon request.

We only observe prices contracted by municipalities for in-kind delivery. Municipalities also have the option to offer cash benefit to users, whom in turn can purchase home help services on their own (possibly from providers which have not been contracted by the municipality). On average, the cash benefit equals about 75% of the in-kind expenditures a patient is eligible for (Botter, 2010). Thus providing a cash benefit is cheaper for municipalities, although they cannot control the quality of suppliers who are hired using cash benefits. This refrains them from using cash benefits on a large scale (VNG, 2011). A small minority (0.14 promille of the population) uses a cash benefit to buy home help services. The use of cash benefits does not vary with the incentive to economize on expenditures (Kattenberg & Vermeulen, 2017).

See e.g. Kessler and McClellan (2000) for a discussion on this issue and a method to deal with this.

We have tested whether large municipalities or municipalities in a collaborative procurement possess buyer power, i.e. are able to purchase services at a lower price than small municipalities do. We do not find evidence for this in our data. See van Eijkel et al. (2017).

We know this from studying several procurement documents and from conversations with local policy makers.

The specification assumes the marginal effect of market share on the logarithm of price to be linear. As a robustness check, we have also added a second and third order term to our main specification. These higher order terms turn out to be insignificant in our estimates. We therefore conclude that the relationship between market share and (the logarithm of) price is best explained by a linear relationship with constant marginal effects and we only discuss the results for such a specification throughout the paper. Estimation results for the nonlinear specifications are available upon request.

The method introduced by Kessler and McClellan (2000) allows the researcher to compute provider-specific (predicted) Herfindahl indices, based on a logit model that uses observables –mainly travelling distance—to predict patient flows in hospital markets. Given that in our paper travelling distance does not play a role in the market under analysis, we abstain from using this method here.

From conversations with people working in the field we know that in some cases it is the local policy maker who picks a provider on behalf of the user, for example when the user is indifferent or cannot make a sound judgement him or herself.

We assign the value zero to providers that were not active in the municipality at that time.

Using an instrument that is fully exogenous to the market process is a common way to tackle this type of endogeneity bias. For instance, in order to measure the effect of competition on hospital quality in the U.K. Propper et al. (2015) use heterogeneity in political pressure among regions as a source of exogenous variation in hospital entry and exit, which is controlled by the central government.

Bergman et al. (2012) argue that with respect to past performance as a selection criterion, the EU guidelines are actually too strict and that more liberal rules, i.e. more freedom for the buyer to select the winner(s), could improve market performance.

This pattern also holds for advanced home help.

www.​aanbestedingskal​ender.​nl.

A few observations in our procurement data are inconsistent with our price data, i.e. a municipality reporting that it employed a regulated price auction while we do observe price variation among suppliers in the municipality. In those cases, we overrule the procurement data and assume that the municipality applied a scoring or price-only auction.

In a few cases the legal form is a mixture of both, for instance a private entity being a subsidiary of a foundation. In these cases, we assume that the provider is for-profit as long as the subsidiary has a for-profit status as we expect pricing decisions to be taken at the subsidiary level.

Our data set contains 295 unique suppliers and 152 unique municipalities. Note that only those municipalities are included for which we have price data stemming from a scoring or price-only auction.

The summary statistics for advanced home help are very similar to the ones for basic home help.

The reduced entry by small firms due to high cost of procurement participation is one of the main reasons for the existence of the California Small Business Preference program. This program grants preferential treatment to, amongst others, small firms in public procurements in California. See e.g. Krasnokutskaya and Seim (2011).

Alpha helps are formally not employed by the home care organization but by the client. In this type of arrangement, the provider acts as an intermediary between the home help and the user. The home help is not required to have specific qualifications to become an alpha help and is not financially protected in the event of illness or disability (see also Arts et al., 1998).

This particular market configuration results in a HHI of 5450, which is close to the average HHI in our data sample.

Note that the expected difference between the two suppliers equals approximately 1%—the point estimate of 0.033 multiplied by the difference in markets shares of 30%—which boils down to an absolute price difference of about 22 eurocents.

Now, the expected price difference between the two incumbents becomes 0.8% (0.033*25). Given that the price of the second largest provider does not change (as its market share does not change), the price of the market leader drops to 22.03 (1.008 * 21.86). Likewise, the expected price difference between the second largest provider and both entrants equals 1.1% (0.033*32.5) boiling down to a price for the entrant of 21.62 (21.86/1.011). Multiplying these price with the corresponding market shares yields an average market price of €21.95.

The average fixed effect for for-profit suppliers does not significantly differ from that for non-profit suppliers. We regard this as evidence that there are no systematic price differences between for-profit and non-profit suppliers, apart from the effect of market share or the year and market in which the supplier operates.

We do not report results based on pooling observations for basic and advanced help, because this approach assumes that the year, municipality and supplier fixed effects are identical for basic and advanced home help, which is a very strong assumption. Still, pooled results are very similar in size to the IV estimates reported in Tables 3, 4 and 5. Results are available upon request.

See Chung et al., 2016.