Sticky costs and regulation in private healthcare providers

Custos assimétricos e regulação em operadoras de planos de saúde

Porporato and Werbin (2012) emphasize that traditional cost accounting is based on the assumption that costs vary proportionally according to the level of activities. Recent empirical research has shown that cost variation depends on the magnitude of changes in the level of activities and the direction of these changes (ascending or descending) (Porporato & Werbin, 2012; Shust & Weiss, 2014). These costs behavior contradicts the traditional model, and they are called sticky costs.

The seminal article by Anderson, Banker, and Janakiraman (2003) established the empirical foundation for analyzing sticky costs. However, the authors focused only on selling, general, and administrative (SG&A) expenses as cost proxies. Subsequent studies have observed sticky costs based on other proxies (Pamplona, Fiirst, Silva, & Zonatto, 2016; Grejo, Abbas, Camacho, & Junqueira, 2019). Also, Anderson, Banker, and Janakiraman (2003) limited their research to industrial companies, whereas subsequent literature expanded cost behavior analyses, examining other sectors (Subramaniam & Weidenmier, 2003; Holzhacker, Krishnan, & Mahlendorf, 2015). Notwithstanding, an aspect that remains little addressed in the literature is the role of sectoral regulation on sticky costs, even though regulation significantly influences this phenomenon, as highlighted by Cook, Kieschnick, and Moussawi (2019).

Among the studies that examined this cost behavior from a regulation viewpoint are Porporato and Werbin (2012) and Holzhacker et al. (2015). For these authors, organizations subject to price regulation by external agencies tend to reduce sticky costs to mitigate operational risks. In Brazil, this is the case of private healthcare providers (PHP), which are closely regulated by the National Supplementary Health Agency (NSHA) (Varella & Ceschin, 2014). Among the NSHA classification of PHP, those classified as medical cooperatives and group medicine stand out. NSHA data (ANS, 2020) indicate that these two classes of PHP, together, are responsible for serving more than 35.6 million beneficiaries and sum 542 organizations operating in the country. A preliminary study by Avelar, Jordão, Boina, Santos, and Ferreira (2019a) suggested relationships between PHP sticky costs and the NSHA. However, the authors focused exclusively on one class of providers, failing to highlight the role of regulated prices - which seems relevant in this context. Thus, there is a theoretical gap on the effects of price regulation on the Brazilian PHP sticky cost, a relevant object of research given the essential role of these organizations in the national health system.

This study sought to answer the following research question: how does the National Supplementary Health Agency’s price regulation affect the private healthcare providers’ sticky costs? Therefore, we analyzed the effects of price regulation on the costs’ asymmetric behavior in Brazilian PHP (medical cooperatives and group medicine) between 2010 and 2018. The research adopted the following specific objectives, (a) to identify sticky costs in the two classes of PHP, (b) to observe accounting variables that influence such costs, and (c) to verify the effect of the NSHA’s regulation variables on the cost behavior of the PHP analyzed.

Studies that focus on Brazilian PHP sticky costs are relevant for many reasons. First, it is important to study this cost behavior in emerging economies, which present quite different characteristics compared to developed countries (where the phenomenon was initially reported) (Zonatto, Magro, Sant’Anna, & Padilha, 2018; Stimolo & Porporato, 2019). Second, PHP have a crucial role in the Brazilian national health system. The country’s supplementary health system has more than 73 million beneficiaries and almost 1,200 operating PHP, the main ones classified as medical cooperatives or group medicine (ANS, 2020). Third, the providers classified in these two categories have several characteristics in common. However, at the same time, these characteristics distinguish them from other PHP registered in the NSHA, which hinders comparability among all the PHP operating in the country (Resolução de Diretoria Colegiada n. 39, 2000; Guimarães & Nossa, 2010; Avelar, Souza, Amaral, & Reyes, 2019b). Fourth, studies focusing on the role of price regulation in the organizations’ sticky costs are important, particularly those in the health area (Holzhacker et al., 2019). Finally, this study expands the analysis developed by Avelar et al. (2019a) when considering the most relevant modalities of PHP in the country, focusing on the role of price regulation as a determinant of sticky costs, and gathering data from a period where the country faced different macroeconomic realities (growth, recession, and low growth) - which is a relevant variable for the analysis of sticky costs (Reis & Borgert, 2019).

The research counted on financial and operating data of PHP, obtained from the NSHA. Data were treated and analyzed using panel data regression. Costs of services were considered inadequate as a proxy for analyzing sticky costs in the proposed models. However, SG&A expenses, as well as total costs, were proxies that had significant relationships in all estimated models. Among the regulation variables, the PHP dependency ratio was considered significant for both classes of PHP.

Models involving providers classified as medical cooperatives presented more significant variables for sticky costs than group medicine. Findings contributing to enrich the literature are (i) the importance of analyzing the organizations’ legal nature when examining sticky costs; (ii) different regulation variables have been proposed to understand this phenomenon, demonstrating its relevance in highly regulated sectors; and (iii) the relevance of employing different proxies when measuring sticky costs, which is a procedure often ignored in the literature.

The first step in developing this study consisted of a search for articles on sticky costs in the databases of the Brazilian Coordination for the Improvement of Higher Education Personnel (Capes) - Science Direct and Scopus. We selected articles according to relevance following the databases criteria and identified those offering empirical, theoretical, and methodological contributions to the topic. In addition, databases gathering articles produced in Latin America were consulted to consolidate a regional character. We selected articles from Scielo, Redalyc, and Google Scholar, observing the studies’ relevance and contribution. Thus, articles that contributed to the development of the theme of sticky cost were selected in both developed and emerging economies, with a focus on Latin America and Brazil.

The population examined was Brazilian private healthcare providers (PHP), and the sample comprised the PHP classified by the National Supplementary Health Agency (NSHA) as medical cooperatives and group medicine. The study used secondary data collected from the NSHA website, referring to 2010-2018 (ANS, 2006). The year 2010 was chosen because it was the year Brazil adhered to the international accounting standards (Gelbcke, Santos, Iudícibus, & Martins, 2018). The option to limit the sample to PHP classified as medical cooperatives and group medicine was due to their characteristics, as explained in subsection 2.2. Table 2 shows the number of PHP in the sample (per class and year).

Table 2
Number of PHP in the sample, per year, per classification

Source: Elaborated by the authors

Financial data was obtained from the providers’ annual financial statements, and operational data was gathered from the NSHA’s website by request supported on the Brazilian freedom of information law (Law 12527/2011).

The data were analyzed using panel data regression, widely used in studies on sticky costs since the seminal study by Anderson et al. (2003). Three models were estimated for each class of PHP, and three different cost proxies were used, following the works by Pamplona et al. (2016), Zonatto et al. (2018), Avelar et al. (2019a), Reis and Borgert (2019): costs of rendered services, selling, general, and administrative (SG&A) expenses, and total cost.

The first estimated model was based on the original study by Anderson et al. (2003), aiming to identify the providers’ sticky costs (Equation 1). For all models presented below, (a) cost means the different proxies of costs mentioned above, (b) REV is the providers’ net sales revenue, and (c) Red is the multiplier reduction, which is a dummy variable (1 is applied when there is a decrease in revenues between the periods, and 0 otherwise). Also, α, β, i, t, and μ represent, respectively, the intercept, the coefficients, the organization, the period, and the error.

The models were estimated to verify the influence of accounting variables on the sticky costs, based on the studies by Richartz and Borgert (2016) and Reis and Borgert (2019). Not all variables cited by those authors were used, either because they are variables exogenous to the providers or because they demand information from the capital market (unavailable to the vast majority of PHP). Equation 2 presents the model.

Where:

AVAI is the available asset;

FIX is the fixed asset (property and equipment);

TPC is the third party capital used; and

TAS is the total asset.

Finally, models related to NSHA regulation on the PHP’s sticky costs were estimated. Based on legislation on healthcare providers and the study by Avelar et al. (2019a), three variables with the potential to influence costs with asymmetric behavior were selected: the proportion of old plans in the providers’ portfolios, proportion of collective plans in their portfolios, and the providers’ dependency ratio. The possibility of a PHP owning a hospital was also considered. Chart 1 shows the operationalization of the variables related to regulation. Equation 3 presents the model.

Chart 1
Operationalization of variables related to regulation

Source: Elaborated by the authors

Tests were used to select the best-estimated model: (a) Chow test, to choose between the fixed effects model or pooled ordinary least squares; (b) Breusch-Pagan Lagrange multiplier test, to choose between the random effects model or pooled ordinary least squares; and (c) Hausman test to choose between the random effects or fixed effects model (Gujarati & Porter, 2011; Fávero, 2015). Other tests were applied after estimating the models to assess the compliance with assumptions and adequacy criteria: F/Wald test, variance inflation factor, Wooldridge test, and modified Wald test (Gujarati & Porter, 2011; Fávero, 2015). Figure 1 summarizes the methodological procedures.

Figure 1
Synthesis of the methodological procedures
Source: Elaborated by the authors

The results obtained when applying the original model by Anderson et al. (2003) (Equation 1) are described in Table 3. There was no significance observed for any variable in any of the providers’ classes when adopting the dependent variable cost of rendered services. On the other hand, the variables related to revenue were significant for the models of SG&A expenses and total costs in both classes of PHP. Table 3 also shows that sticky costs only occur when the variable SG&A expenses were used as a cost proxy in the case of medical cooperatives. Sticky costs were not evident for group medicine providers.

Table 4 shows the results obtained for each PHP classes based on the model with accounting variables (Equation 2). Again, no significant relationship of their independent variables was found when the proxy was the costs of rendered services. On the other hand, two variables revealed sticky costs - earnings available and fixed assets - since they presented significant relationships in both remaining models estimated for the medical cooperatives. The asymmetric behavior of costs, when observing the total costs variable, can be considered sticky. In contrast, anti-sticky behavior was observed for the dependent variable SG&A expenses. PHP classified as group medicine did not show sticky behavior in this model.

Finally, Table 5 presents the results for the models that considered the regulation variables (Equation 3). None of the variables in the models that used the dependent variable cost of rendered services were considered significant. As observed in the previous models, when using the variables SG&A expenses and total costs as proxies, significant results were obtained regarding revenues and, in some cases, there is evidence of sticky costs.

Considering the dependent variable total costs, the estimated model for medical cooperatives revealed sticky cost - with a contribution from the variable dependency ratio. On the other hand, when considering the variable SG&A expenses as proxy costs in the providers classified as group medicine, the variable dependency ratio demonstrated an anti-sticky asymmetric behavior. It should be noted that the other regulation variables were not considered significant in the estimated models. From the results, it was possible to confirm hypothesis 3, that the highest dependency ratio in the providers’ portfolios increases the occurrence of sticky costs. However, it was not possible to confirm hypotheses 1 and 2, which assumed the relationship between the sticky costs and the variables proportion of old plans and proportion of collective plans in the providers’ portfolio. Chart 2 summarizes the results regarding the hypotheses.

Table 3
Results - Original model

Source: Elaborated by the authors Note: *Significance of less than 10%; ** significance of less than 5%; *** significance of less than 1%

Table 4
Results - Accounting variables

Source: Elaborated by the authors Note: *Significance of less than 10%; ** significance of less than 5%; *** significance of less than 1%; ^a values multiplied by the dummy of revenue reduction.

Table 5
Results - Regulation variables

Source: Elaborated by the authors Note: *Significance of less than 10%; ** significance of less than 5%; *** significance of less than 1%; ^a values multiplied by the dummy of revenue reduction.

Chart 2
Hypotheses

Source: Elaborated by the authors

The models using the cost of rendered services as costs proxy did not present any significant variables, which is a result that contradicts the expectations based on the model by Anderson et al. (2003) and the studies by Avelar et al. (2019a), Holzhacker et al. (2014), and Richartz and Borgert (2016). Therefore, this proxy was not suitable to analyze sticky costs in the models applied to Brazilian private healthcare providers (PHP). This result confirms Shust and Weiss’ (2014) findings on how the choice of proxies can influence the measurement of sticky costs through different statistical models. It is also aligned with findings by Zonatto et al. (2018), who showed the significance of models that used this same proxy in Brazilian companies.

On the other hand, the fact that the variables related to revenues are significant in the models where selling, general, and administrative (SG&A) expenses and total costs were used as cost proxies in both classes of PHP suggests a positive relationship between these proxies and revenue, which is aligned with findings of Porporato and Werbin (2012). Sticky costs were only observed in medical cooperatives, confirming the results of Avelar et al. (2019a). In this case, it appears that even when operating in the same sector, organizations with different legal nature may or may not present sticky costs.

The model with accounting variables estimated for medical cooperatives identified the contribution of the organization’s earnings available and fixed assets to sticky costs. This result confirms that obtained in Grejo et al. (2019) and Richartz and Borgert (2016) and, partially, the health sector findings by Avelar et al. (2019a) and Holzhacker et al. (2014). There was anti-sticky cost behavior when the cost proxy was the total cost, contrary to what was observed in Avelar et al. (2019a).

Finally, the results regarding the regulation variable providers’ dependency ratio showed different sticky costs depending on the classification of the PHP and the cost proxy used, demonstrating the complexity of this asymmetry. This phenomenon seems to be sensitive to the legal nature of organizations, which is a variable that is not usually present in national or international studies. Usually, such studies focus on the sector (cost structure proxy) as a way to obtain evidence of sticky cost, as observed in the studies by Carmo and Xavier (2016), Santos et al. (2017), Stimolo and Porporato (2019), and Zhang et al. (2019). However, this research demonstrated that sticky cost could differ significantly among organizations in the same sector; in this case, Brazilian supplementary healthcare. Thus, the use of different models to analyze organizations of distinct legal nature can capture such heterogeneity.

The research results show the unsuitability of costs of rendered services as a proxy to analyze sticky costs in the proposed models. Other proxies such as SG&A expenses and total costs demonstrated significant relationships in all estimated models, as expected based on the literature. In these cases, significant results were found regarding the level of revenue (a proxy for the level of activity). Also, when estimating the original model by Anderson et al. (2003), sticky costs were found only for medical cooperatives. When other accounting variables were inserted in the models, some of them presented significant relationships, indicating their effects on medical cooperatives' sticky costs. As for PHP classified as group medicine, none of these variables were significant.

For the regulation variables, providers’ dependency ratio was considered significant in both classes of PHP. The effect on sticky costs was expected, considering that this measure indicates a lower margin for price management (and consequently lower revenue). However, the hypotheses related to greater flexibility in price management - considering the variables related to the proportion of old and collective plans in the providers’ portfolios and their relationship with sticky costs - were rejected. The smaller number of variables contributing to the PHP’s sticky costs compared to companies in other sectors can be related to the healthcare providers’ lean structure in the face of an uncertain environment, with limited management mechanisms in a regulated context, as highlighted by Holzhacker et al. (2015).

The findings obtained in this study contribute to improving the work of different stakeholders in the sector of supplementary health. First, the National Supplementary Health Agency has evidence that regulation may affect the sticky cost of the PHP in different ways. Several variables are involved in the phenomenon of costs asymmetric behavior, and regulation must consider the organizations’ sticky costs to avoid measures that can harm their economic-financial sustainability. From the perspective of the PHP managers, the awareness that different variables influence sticky costs enables them to make informed decisions and reduce the risks associated with such asymmetry. Finally, society as a whole can benefit from these findings since they lead to better management and regulation in the supplementary health sector, which serves tens of millions of Brazilians (ANS, 2020).

This study innovated by noting the importance of the organizations’ legal nature, even within a single sector, when exploring the phenomenon of sticky costs, which is something previous studies did not consider. Different regulation variables were also proposed to understand the asymmetric behavior of costs, exploring their influence in highly regulated sectors. The study particularly revealed the different effects of a regulation variable related to prices on the different classes of PHP, demonstrating the relevance of the variable for decision-making. Finally, the importance of using different proxies to measure sticky costs was highlighted, a condition often ignored in research. Despite the innovation and advances, the study also presented limitations. First, the sample was relatively small in terms of available data, and because only two classes of PHP were investigated. Second, the limitation of available data led to a reduced number of regulation variables.

Future research could test the same or other regulation variables linked to prices in PHP classified as self-management, philanthropic, dental cooperative, or group dentistry, or in other time horizons, overcoming the limitations of this endeavor. Finally, further research is needed to capture the point of view of regulators and managers and better understand the effects of their decisions on sticky costs.