The impact of Portuguese generics drug legislation on the drug market: a structural break approach

1 The impact of Portuguese generics drug legislation on the drug market: a structural break approach by Nelson António Ferreira Duarte Master in Economics Dissertation Supervised by Álvaro Almeida, PhD 2014 i Biographical Note Nelson António Ferreira Duarte was born in Porto, Portugal on April 12, 1990. In 2008, he concluded his high school education and started his undergraduate degree in Departamento de Economia, Gestão e Engenharia Industrial (DEGEI), Universidade de Aveiro. He graduated in 2012 and in the same year started his Master’s Degree in Economics in Faculdade de Economia do Porto, Universidade do Porto. ii Acknowledgements I am extremely grateful to Professor Álvaro Almeida, who was the supervisor of my dissertation, for his advices, availability, insights and criticism. I would like to thank Faculdade de Economia do Porto, as well, for allowing me to pursue my objectives as a student and offering me the possibility to contribute to science. I also like to thank INFARMED for making available the data requested by e-mail conversations. I am thankful to my family, especially to my mother and sisters, for their support in these last two years. I am also grateful to my colleagues in FEP, especially João, Rui, Manuel and Marisa, for the hours of talks, fun and insights in these last two years. Finally, I am also extremely grateful to Diana, more than anyone, for her unconditional support, advice, insights and criticism in the last 4 and half years. iii Abstract The Economic and Financial Adjustment Program Portugal signed with international organizations included several recommendations whose objective was to reduce healthcare expenditures. These recommendations aimed primarily at stimulating generic drugs consumption, thus reducing the public expenditure on healthcare (by reimbursement and reference pricing). In this study, first we test for the presence of structural breaks on relative prices; then, we use a time series approach to determine the exact month on which a structural break in generic drugs consumption occurred, using monthly data on generic drugs market shares and relative prices from January 2007 to December 2013. We identify one structural break on generic drugs consumption that coincides in time with administrative reductions on relative prices. This result highlights the importance of relative prices on the demand for generic drugs. JEL Codes: I18; K32; L16 Keywords: Generic drugs consumption; Portugal; Policy Measurement; Structural Change iv Resumo O Programa de Ajustamento Económico e Financeiro, que Portugal assinou com organizações internacionais, incluía diversas recomendações, cujo objetivo era reduzir os gastos nos cuidados de saúde. Estas recomendações visaram principalmente estimular consumo de medicamentos genéricos, reduzindo assim os gastos públicos em cuidados de saúde (por comparticipação e preço de referência). Neste estudo, primeiro testamos a presença de quebras estruturais de preços relativos; de seguida, usamos uma abordagem de séries de temporais para determinar o mês exato em que ocorreu uma quebra estrutural no consumo de medicamentos genéricos, usando dados mensais sobre as quotas de mercado de medicamentos genéricos e dos preços relativos entre Janeiro de 2007 e Dezembro de 2013. Identificamos uma quebra estrutural no consumo de medicamentos genéricos, que coincide no tempo com reduções administrativas dos preços relativos. Este resultado destaca a importância dos preços relativos da demanda por medicamentos genéricos. Códigos JEL: I18; K32; L16 Palavras-Chave: Consumo de Genéricos; Portugal; Avaliação de Políticas; Quebra Estrutural v Table of Contents Chapter 1 Introduction ........................................................................................... 1 Chapter 2 Literature Review ............................................................................... 4 2.1 Determinants of the demand for generic drugs ........................................................ 4 2.1.1 Price ........................................................................................................................ 4 2.1.2 Reference Pricing System ....................................................................................... 7 2.1.3 Brand Loyalty, Physician/Patient/Pharmacy role on prescription .......................... 8 2.2 Studies on Portugal regarding generic drugs ......................................................... 12 Chapter 3 Legislation review ............................................................................. 13 Chapter 4 Methodological Overview ............................................................ 18 4.1 Data............................................................................................................................. 18 4.2 Methodology .............................................................................................................. 19 Chapter 5 Results and Discussions................................................................. 22 5.1 Relative prices ............................................................................................................ 22 5.2 Generic drugs market Shares .................................................................................. 26 Chapter 6 Conclusion ............................................................................................. 31 References ...................................................................................................................... 34 vi Relevant Legislation ................................................................................................ 36 Appendices .................................................................................................................... 38 Appendix A – Extended output for the estimation of relative prices using the BreakLS method ........................................................................................................................... 38 Appendix B – Results of the Breusch-Godfrey Serial Correlation LM Test in the relative prices estimation ............................................................................................... 39 Appendix CExtended output for the BreakLS estimation of the demand for generic drugs, using HAC estimators to correct serial correlation ............................................ 40 Appendix D – Description of the breakpoint specification used in the BreakLS estimation ...................................................................................................................... 41 Appendix E – Breusch-Godfrey Serial Correlation LM Test extended output results for the BreakLS estimation of the demand for generic drugs ....................................... 42 vii List of Tables Table Page 1 Summarized results of the presence of structural breaks in relative prices, using the BreakLS method 23 2 Summarized output of the estimation of the demand of generic drugs, using Least Squares with Breaks method 26 3 Breusch-Godfrey Serial Correlation LM Test summary results 28 List of Figures Figure Page 1 Generic Drugs Market Shares Evolution 20 2 Graphical representation of structural breaks on relative prices 24 3 Consumption of generic drugs, using the BreakLS method 28 1 Chapter 1 Introduction Health, according to the World Health Organization, is a “state of complete physical, mental and social well-being and not merely the absence of disease or infirmity”, and is a fundamental right both on The Universal Declaration of Human Rights (on its article 25) and on Constitution of Portugal (on its article 64 of the first chapter). However, the preservation of such rights underlies the creation of Health Systems, which implies costs and relocation of resources to the health sector of such economy. In Portugal, the burden of public and private expenditure on Health corresponds, in 2010, to 10.2% of Portuguese GDP1, and had been rising since 2006. This subject deserves particular attention, since Portugal is, from 2011, under an external financial aid program. 1 This data was collected from OECD ‘s website 8 period, the authors showed that “(…) RP reduces brand-name and generic drug prices and increases generic market shares”, thus confirming the previous two hypotheses. The same type of study is followed by Podnar et al. (2007). The authors studied the impact of the introduction of RPS (in 2003) on generic and brand name drugs market share, in the Slovenian market. Their results showed that “(…) originator manufacturers tend to lose market share because of reference pricing.”. Simoens et al. (2005) studied the effects of the introduction of RPS in Belgium on generic drugs market share, between the years 1998-2003. The authors used data from IMS Health Belgium to trace general trends and data from Ifstat to analyze price evolutions of certain drugs. They found that “The market share held by generic drugs increased following implementation of the RP scheme”. They also found that “the low market share of generic drugs in Belgium principally drives from lack of incentives for physicians to prescribe generic drugs”. This last result is particularly important because there are some studies in the literature that consider physician/patient role on prescription and brand loyalty as a relevant determinant of brand name and generic drugs market share. 2.1.3 Brand Loyalty, Physician/Patient/Pharmacy role on prescription Hurwitz and Caves (1988) studied how promotion, marketing and information can influence consumers on buying certain type of drugs (brand name versus generic drugs). After they provided a general description of the brand name and generic drug market worked, they analysed health professional’s decision-making process on drug prescription. They found that “The physician has no substantial economic incentive to choose the lowerpriced product, and doctors tend to be ignorant of specific drug prices. The brand name, simpler than the generic name, was learned when the drug was introduced and is easier to remember, and habit accordingly plays a strong role in the physician's prescription practice”. They also found that physicians have less faith in quality control on generic drugs manufacturers than of brand name drugs manufacturers. Using a sample consisting of 150 drugs that were available as generics and had, at some point, a patent held by originating manufacturers, the authors construct a model to study the impact of sales- 9 promotion on both generic and brand name drugs market shares. Their results showed that “the trademark holders' sales-promotion outlays do preserve their shares against incursion by generic entrants”, showing that trademarks have an important role on the evolution of generic drugs market share. Hellerstein (1998) studied the importance of physicians in the process of prescribing a brand name or a generic drug. The author argues that physicians, when playing as an agent for their patients, face some informational costs when prescribing medicines, such as collecting information on the availability and efficacy of generic drugs and price differentials between generic and brand name drugs, among others. It is costly, for the physician, to gather information about generic drugs, because its manufacturers do “(…) very little advertising, while information about new trade-name drugs is widely disseminated formally through advertising and the published results of drug efficacy studies. It may therefore be much more costly to a physician to learn about the introduction of new generic drugs”. Using a sample of 38.384 patient’s information, collected via a survey from 1.223 office-based physicians over the course of the year 1989 in USA, the author intended to establish a pattern in physicians’ drug prescription. Their results showed that “the central result is that the physician is an important agent in the prescription decision”. The same type of study is followed by Coscelli (2000). The author studied the contribution of physician and patient habits to persistence of market shares in prescription drugs market. Unlike Hellerstein (1998), the effects of both patient and physician, on prescribing behavior, can be isolated. This allows a better understanding of dynamics of prescription than aggregated data. Using a probit specification and a panel dataset provided by Italian National Health Institute, the author tested the null hypothesis of whether physicians and/or patients are indifferent between “different brands of the same molecule, as we would expect given their therapeutic equivalence [i.e. generic and brand name drugs]”. After rejecting this hypothesis, the author tried to isolate effects responsible for product differentiation. The results showed “significant evidence of doctor and patient `habit', which imply that in molecular sub markets in which brands are not allowed to compete on the basis of price, 10 habit persistence at the microlevel can translate into sticky and persistent market shares at the aggregate level”. She also found that “It is the state dependence at the micro level for the patients, and the habit persistence for the doctors, which lead to the persistence in market shares that we observe at the `macro' level”. Brekke et al. (2013) studied the pharmacies role in promoting generic substitution and thus stimulating competition between generic and brand name drugs. The authors first constructed a theoretical model where pharmacies can persuade patients with a prescription for a brand name drug to buy its generic counterpart. With this model, they showed that pharmacies incentives to substitution are determined by relative margins and relative copayments of drugs. This result meant that “a larger difference in margins between generics and brand name drugs increases the generics market share”, because it is more profitable (for the pharmacy) to promote the first. The authors also found that, through this model, the effect was stronger where drugs are subjected to reference pricing. To test the theoretical model, mentioned above, the authors collected data from two different databases (the Prescription and Wholesale database) of Norwegian Institute of Public Health. The prescription database contains information about prescription bound sales at pharmacy level from 2004 onwards, average prices and volumes per month over the 2004-2007 period. The Wholesale database contains information about producer prices “per product per wholesaler per month”. Using a fixed effect model regression and IV regression to control for endogenity issues, the authors found that “pharmacy margins on branded versus generic drugs have a strong association with generic market share”. However, the authors stressed that “that the empirical analysis has not established causality, and that [their] empirical findings can be interpreted as correlations only.”. Liu et al. (2009) studied the hypothesis of whether financial incentives had an important role on physicians prescription habits (between brand-name and generic drugs) when they prescribe and dispense drugs. Using a data set containing detailed records of personal health care services “(…) including outpatient visits, hospital admissions, and prescription drugs” from 200.000 diabetes patients, the authors used a probit estimation to test the probability of a physician to prescribe a generic drug. Their results showed “(…) that the 11 profit margin between the reimbursement and the acquisition price has a significant effect on a physician’s prescription making decision”. Iizuka (2007) studied the physician-patient relation in drug prescription in Japan. Using panel-data from 40 hypertension medicines, such as physician’s mark-up, retail price, market share and product characteristics (representing a data sample of 258 observations), the author found that “(…) physicians’ decisions are influenced by the markup they obtain.”, which “(…) suggests the existence of the agency problem in this market”. The author also found that physicians “(…) prefer to dispense drugs that cost less to the patient, ceteris paribus. Estimated parameter values indicate that physicians are willing to give up one dollar of their profit in order to reduce the copayment of non-elderly patients by 28 cents. This implies that, although physicians do take advantage of markup, they care more about patient welfare than their own profits from markup.” Rischatsch et al. (2013) analyzed the role of physicians and patients financial incentives, when choosing between generic and brand name drugs, in Switzerland. Considering physicians as an imperfect agent for patients (since in some Swiss cantons, physicians “(…) are allowed to dispense drugs to their patients on their own account”) and using data provided by “(…) a major Swiss health insurer” representing 15% of Swiss population , the authors constructed a random-effects logit model where they tested some hypothesis, namely the likelihood of generic prescription compared to brand name drugs prescription due to higher income contribution; and generic drugs being more often prescribed to lower income individuals due to their marginal utility of income; amongst many others. Their results showed that “(…) financial incentives, agency towards the patient, and agency towards insurers are all found to markedly influence generic substitution” and “generics are prescribed more often to patients with high copayments or low incomes.”. 12 2.2 Studies on Portugal regarding generic drugs Mendonca (2011) reviewed the main policies intended to the generic drugs market in Portugal, since 2000. The author provides an overview of the key measures taken addressed to the generic drugs market since 2000 (such as the implementation of the RPS in 2002, price differentials between generic and brand drugs, among others) and provided, as well, an overview of new measures being prepared (at that time) to increase the prescription of generic drugs. The author argued that the measures taken in 2010 (changes in RPS) “(…) were important steps in order to create substantial financial incentives for patients to buy generics and also contributed heavily to reduce the expenditure of the NHS [National Health System] with ambulatory medicines.”. Barros and Nunes (2010) studied the impact of several policy measures on total pharmaceutical spending. The authors use data provided from public sources and checked with INFARMED between January 1998 and August 2008. The dataset contained information about payment made by NHS, information about total sales in value of pharmaceutical products and information about the number of boxes sold in the Portuguese market. Using a endogenous structural break approach, the authors found that “(…) a transitory slowdown in NHS pharmaceutical expenditure growth in the first half of 2003, coinciding with the start of the reference pricing system. However, this slowdown was relatively short lived. Before the end of the year, growth of pharmaceutical expenditure had returned to the historical path.” They also found that “Government determined price decreases have only a level effect. The underlying dynamics do not change.”, 13 Chapter 3 Legislation review Governments and health authorities in particular, use legislation as their main instrument to implement policy and regulation to achieve the results proposed by themselves, or, in this case in particular, together with other authorities. Since those determinants of generic drugs consumption, discussed in the previous chapter, reveal what impacts the most on individuals and households process of choice, health authorities exploit those determinants as their object of legislation to achieve a better and faster impact on the drug market. As we will demonstrate next, by reviewing the main legal documents implemented by the Portuguese authorities, legislation and legislators use those determinants mentioned earlier as a mean to stimulate the generic drugs consumption and competition in the drug market, in Portugal. 14 Few studies present a review of legislation regarding medicines in Portugal. Mendonca (2011) and Barros and Nunes (2010), like we had seen in the last chapter, provided a good review on the results of the measures implemented by Portuguese legislators until 2010. A more detailed review on medicine policy was provided by Barros and Nunes (2011), in which the authors reviewed the previous ten years on medicine policy in Portugal, using different methodologies to determine the effects of all sorts of legislation created regarding medicines. Since there are several studies that review the legislation on medicines in Portugal until 2010, we will focus our legislation review on the main legislation implemented after 2010. We will review a few documents that came into effect in 2010 only because we consider that documents are important to explain some legislation that were published subsequently. Law number 62/2011 from December 12 created a regime of disputes composition between generic and brand name drugs, when industrial and property rights are potentially being violated and altering reimbursement regime for drug prices. The main alterations to LawDecree number 176/2006 from August 30, regarding industrial and property rights, were that the request for market introduction of certain drugs could not be based on intellectual property rights and that the market introduction authorization could not be revoked or suspend on terms of intellectual property rights. However, if the originating drug owners dispute the market introduction of a generic drug on arbitral tribunal, the latter cannot carry on its commercial activity (generic drugs commerce) due to intellectual and industrial property rights of the originating firm. Inserted in set of measures, some of them imposed by the Memorandum of Understanding signed with International Monetary Fund and the European Central Bank, the Portuguese Government revised the drug policy with the Law-Decree number 112/2011 from November 29. This Law-Decree sets the maximum price for the first generic drug entering the market (a minimum 50% discount on the Stockist’s price and a minimum 25% for the retailer’s price, from the reference drug present in the market) and revised the reference 15 price system, altering the countries that serve as reference3. Regarding the revision of the reference price system, the countries that serve as comparison are Spain, Italy and Slovenia. This Law-Decree was, however, changed many times along the years. The first time was by Law-Decree number 152/2012 from July 12. This document changed/revised some articles of the previous Law-Decree. The most relevant changes were: as assigning the regulatory powers regarding price revision exclusively to the Government member assigned with the Health affairs; changing the public entity responsible for the price authorization (INFARMED); and setting the originating drug’s price as the average of the last two previous years. The second change was made by Law-Decree number 34/2013 from February 27. This document intended to revise some articles in order reduce the burden on health expenses through National Health System (NHS). To do so, the Portuguese Government determined that the three countries used as comparison for price setting are determined annually 4, and the retail price for parallel imported drugs, to be introduced in the Portuguese drug market, should be 5% lower than the retail price for the same medicine and its equivalents. Complementary to Law number 112/2011 are Portaria number 3/2012 and Portaria number 4/2012, both from January 2. Portaria number 3/2012 defines the criteria, deadlines and proceedings for exceptional price revision, predicted on Law number 112/2011. Portaria number 4/2012 set the rules and deadlines for price formation and revision. Portaria number 4/2012, however, was altered two times since it came to effect. The first was by Portaria 335-A/2013 from November 15, altering the deadlines for price transition after the date the new price is approved. The second time that Portaria number 4/2012 was altered was by Portaria number 367/2013. This document suspended the application of the article that obliges for annual price revision, since the generic drugs average prices are lower than the maximum prices would be set after such price revision. Therefore, the 3 According to this Law-Decree, the retailer’s price is composed by the stockist’s price plus the wholesaler’s margin, the retailer’s margin and taxes. These margins are regulated by the Portuguese Government, as we can see in Section IV of this document. 4 The criteria used were GDP per capita similar to Portuguese’s GDP per capita (comparable by purchase power parity) or a lower price for each drug in particular. 16 legislator considered that “(…) there is no justification at present to proceed with its review in 2014, if the same average level do not suffer significant changes.”. The Law number 11/2012 from March 8 sets new rules on drug prescription and dispense. This document revises the rules on drug prescription, namely obligating physicians to declare the INN, among many other things, of the drug that they are prescribing; the receipt for such prescription must be filled electronically5. On the other hand, this document also regulates the pharmacy’s way to proceed. Pharmacies must inform patients, in the act of dispense of drugs, of the existence of generic drugs (if they exist), and the patient has the right to choose between a generic or a brand-name drug, even if the drug prescribed was, in fact, a brand-name drug. Complementary to the previous document, comes Portaria number 137-A/2012. This document establishes the legal regime to which the rules of prescription and dispense of medicines obey, created by the Law number 11/2012, and define new rules about to information provision to patients. These new rules obligate pharmacies to inform which are the cheapest drugs available, for that specific prescription and leave the patients the right to choose between the drugs available. Related to the Law number 11/2012 and Portaria number 137-A/2012 is the Portaria number 340/2012 from October 25. This document creates Pharmaceutical and Therapeutic Comissions to evaluate and control the pattern of drug prescription and dispense. It also regulates the mechanisms of such evaluation in order to Health authorities have some degree of evidence about the effectiveness of prior legislation. The Law-Decree number 48-A/2010 from May 13 showed the intents for a comprehensive review of the reimbursement system for medicines, with special focus on some regimes, in order to gain equity. The main objectives of this document is to improve access to medicines, to make the reimbursement system more efficient and sustainable and, finally, to promote the generic medicines substitution. To achieve those objectives, this Law5 Due to Information Technologies (IT), the physicians have information regarding cheapest drugs (brandname or generic) available on the market, which are set by the Government member responsible for the Health affairs. 17 Decree sets some measures, namely: 100% reimbursement rate, for the cheapest five medicines for each homogeneous group, to patients who are in the special regime; shortening the deadline for a decision on reimbursement in the price of a generic drug; new rules on generic drugs pricing, for each homogeneous group, when they have, at least 5% market share on generic drugs market. It also contains other amendments, “(…) in particular fixing marketing margins to existing levels in 2005, without this entailing any modifications to approved retail prices or impact on costs for the National System of Health.”. The previous document has been altered many times during the years. The Law-Decree number 106-A/2010 made the first alteration. This document’s objective was to promote fairer rules to medicines access and to combat fraud and abuse on reimbursement from NHS. To achieve those goals, the legislator reviewed, among other things, the rules for reimbursement approval. The second alteration came by Law number 62/2011, which had already been covered. The third amendment was by Law-Decree number 103/2013 from July 26. This document reviews the process reference prices approval for each medicine’s homogeneous group and reviews the mechanisms of homogeneous group’s formation, in order to promote savings from substitution to generic drugs. As we can see by this chapter, all the documents directly intervened in the drug market through different determinants of generic drugs consumption suggested by the literature, such as price, Reference Pricing Schemes and Brand Loyalty and the Physicians and Pharmacies role on drug prescription, to ultimately promote generic substitution, whether by imposing to mention in the prescription that a generic drug is available for that particular originating drug or just by administratively reducing prices of drugs. One final remark to this chapter is that it makes clear the bond between literature and Government regulation and policy, since the information and knowledge created by the scientific community is useful to Government authorities, and ultimately ends to being useful to patients too. 24 -.08 -.04 .00 .04 .08 .12 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2007 2008 2009 2010 2011 2012 2013 Residual Actual Fitted Figure 2Graphical representation of structural breaks on relative prices These results can be consequence of legislation published by the time the structural break occurred. The structural break in October 2008 can be consequence of the implementation of Portaria number 1016-A/2008 of September 8, which came into effect on October of the same year. This legal document reduced the maximum retail selling prices for generic drugs by 30%, for generics with retail prices above 5€. The structural break in June 2010 can be the outcome of the implementation of two legal documents, namely Law-Decree number 48-A/2010 of May 13 (but the date it came into effect was in June 1st) and Portaria number 312-A/2010 of June 11 (but the date It came into effect was July 1st). The Law-Decree number 48-A/2010 reviews some regimes on reference pricing and reimbursement rates for generic medicines; and Portaria number312A/2010 sets new rules on drug pricing, its annual review and its alteration. Additionally, this document set new rules on retail pricing for the new generic drugs to be subject to reimbursements. 25 The impact of these two documents cannot be identified separately, because they coincide in time. Although the last legal document has entered into effect after the month identified structural breaks, its effects may have been triggered in June, once it laid down the rules to be applied in the following month. The individual effect of these documents cannot be tested separately to know which of them has the dominant effect, because their effects coincide in time. So, we only can identify their aggregate impact, which lead to a structural break in June 2010. Finally, the structural break in April 2012 can be consequence of two documents, namely Law number 11/2012, which dates March 2012 and Portaria number 137-A/2012, which dates in May 2012, and sets the legal regime described by the previous document. These two documents intended to encourage the use of generic drugs in Portugal, in detriment of brand-name drugs, by requiring the prescription of drugs by International Non-Proprietary Name (INN). The ultimate goal is improve the efficiency of National Health System, since legislators state that generic drugs “(…) may play a pivotal role in promoting rationality and sustainability of the National Health Service (NHS), as well as generating significant savings for citizens.”. Despite these two documents did not regulate prices of neither generic nor brand-name drugs, it is clear that they lead to a structural break in relative prices of drugs, hence producing an indirect effect on relative prices. This result can be explained with the fact that these two documents may have led to price competition between generic and brandname drugs on the drug market, thus leading to a structural break. However, these two documents have not been published or came into effect in April 2012. One reason to that is that retailers may have anticipated the publication of Portaria number 137-A/2012 after Law number 11/2012 (published in March), and may have triggered an increase in generic drugs prices, by generic drugs manufacturers, in April (one month before the legal regime for the latter document was created) of the same year, which led to an increase of relative prices. Please note that this increase can be seen in Figure 2, when analyzing the relative prices during 2013. 26 5.2 Generic drugs market Shares In this section, we studied the impact of relative prices on the demand of generic medicines. Table 2 shows the summarized results of the estimation using the BreakLS method. Method BreakLS Specifications HAC standard errors & covariance (Quadratic-Spectral kernel, NeweyWest fixed bandwidth); Allow heterogeneous error distributions across breaks Break Date 2008M11 Prob(F-statistic) 0 R2 0,9802 Variable Coefficient Pvalue Before Breakpoint (number of observations: 22) C 0,3346 0,0104 Relative_prices -0,1314 0,0775 Trend -0,0001 0,8794 After Breakpoint (number of observations: 62) C 0,1238 0 Relative_prices -0,0267 0,0198 Trend 0,0021 0 Table 2 – Summarized output of the estimation of the demand of generic drugs, using Least Squares with Breaks method The number of observations included in the sample was 84, which is the total number of months between January 2007 and December 2013, and shows that the relative prices are responsible for explaining 98.02% of the variation of the generic drugs market shares. It should be noted that the variable relative_prices is statistically significant only after the identified break, since its p-value<0.05. 27 However, this model has some specifications for a more accurate estimation of the breakpoint dates, using a “sequential test in all subsets” method 9 and to prevent for serial correlation, namely the HAC (Newey-West) coefficient covariance matrix, as described by Newey and West (1986) and often used to correct the effects of correlation, and allowing errors distributions to differ across breaks. These specifications can be seen in the top part of the output presented in Table 2. This method provided us a quicker and more efficient estimation of the dates where the structural break occurred, since it automatically conducted a Multiple Breakpoint Test described on Bai and Perron (1998), which presents a sequential application of breakpoint tests. A detailed description of how this test works is provided by Eviews 8 User’s Guide (2013)10. Additionally, and since the object of this study is to test for a structural break, the output present the different coefficients for each partition of the sample on the date the break occurred, thus allowing us to interpret the different impacts of relative prices on the demand for generic drugs for each partition of the series. On Appendix C is represented an extended visualization of the BreakLS estimation output and on Appendix D is presented an extended breakpoint specification analysis provided by Eviews at regarding this specific estimation and containing information about the test conducted to determine the breakpoint dates. Additionally, in Figure 3 is presented a graphical representation of this estimation. Regarding the sign of the coefficients of relative prices, determined by the estimation present in Table 2, it is possible to verify that in all partitions of the sample, both are negative. Please note that the coefficient of the Trend is positive in the second partition.T his result shows not only that the impact of relative prices on the demand for generic drugs has changed after the structural break, but also that an increase in relative prices leads to a decrease on the demand for generic drugs in Portugal, as expected. 9 This method uses a Bai test of breaks in all recursively determined partitions. 10 Eviews user’s guide is provided by EViews after the installation of the software. 28 -.04 -.03 -.02 -.01 .00 .01 .02 .05 .10 .15 .20 .25 .30 2007 2008 2009 2010 2011 2012 2013 Residual Actual Fitted Figure 3 - Consumption of generic drugs, using the BreakLS method The next step is to test the presence of autocorrelation in the model and thus validating the method used in this study. The presence of autocorrelation invalidates all of the statistical hypothesis tests done under the OLS method. Using the built-in test in EViews 8, we conduct a Breusch-Godfrey Serial correlation LM test, to test the presence of autocorrelation, as described in Godfrey (1978) and Breusch (1978) and the summary results are showed on Table 3. Breusch-Godfrey Serial Correlation LM Test: F-statistic 2.180.572 Prob. F(2,74) 0.1200 Obs*Rsquared 4.558.622 Prob. ChiSquare(2) 0.1024 Table 3 – Breusch-Godfrey Serial Correlation LM Test summary results 29 As Table 3 shows, there is no evidence of the presence of autocorrelation, since the Prob.Chi-Square is 0.1024, which leads us to not reject the null hypothesis of no serial correlation, thus validating the all statistical hypothesis of this estimation. An extended output, containing more information regarding this test can be found on Appendix E. The estimation presented in Table 2 showed that there was only one structural break in the series, namely in November 2008. The structural break that occurred in November 2008 can be consequence of the legislation pack composed by Law-Decree number 184/2008 and by Portaria number 1016-A/2008 both published in September of that same year. These two legal documents complement each other, since the first established that generic drugs prices can be subject to exceptional revision at any point in time and that revision must be published by Portaria, and the second legal documents sets the first generic drugs price revision authorized by the previous document. Please note that Portaria number 1016-A/2008 imposed reductions of 30% on generic drugs prices above 5€. However, the reasons why the structural break occurred in different months may be due to the possibility of postponing the consumption of medicines, by the patients, thus leading to a structural change in the demand of generic drugs. Just like a previous structural break identified on our relative prices analysis, the individual effect of these documents cannot be tested separately to know which of them has the dominant effect, because their effects coincide in time. So, we only can identify their aggregate impact, which leads to a structural break in November 2008. Our results are consistent with the results of Aronsson et al. (2001), since relative prices have a significant impact on generic drugs market shares and, therefore, in the market shares of the brand name drugs, since the increase in relative prices resulted in the decrease in generic drugs market share, in all partitions. Additionally, our results are also consistent with the results found by Dalen et al. (2006), since that the introduction of a retail price cap in drugs prices helped to increase the generic drugs market shares. 30 In sum, these results present an interesting insight on the impact of legislation and its dynamics. First, on our relative prices analysis, we found that three structural breaks occurred during the entire period covered. The first two were motivated by legislation that regulates prices directly, thus reducing the relative prices. The last structural break was unexpected since it was a result of a legal document that targeted the demand for generic drugs and did not intend to regulate prices. Additionally, following the conclusions of Hellerstein (1998), the April 2012 structural break result reinforce the importance of physicians and pharmacies on the process of prescribing and dispense of medicines, since it is clear that the new rules on such matters led to a structural break after their implementation. Finally, our main model studies the relationship between generic drugs market shares and relative prices in our generic drugs market share approach. Our results showed only one structural break and that structural break was motivated by a legal pack intend to reduce generic drugs prices. This structural break highlights the importance of relative prices on the demand for generic drugs. However, this model was not able to indicate any legal document, intended to stimulate directly the demand for generic drugs, capable to create a structural break in the consumption of generics. Despite this, it does not mean that this legislation has not produced effects the demand for generic drugs, since it may have contributed to the growth trend of generic drugs market shares during the entire period covered. 31 Chapter 6 Conclusion This study was set out to stress the effectiveness of the legislation regarding generic medicines substitution, imposed the Economic and Financial Adjustment Program Portugal signed with international organizations. More precisely, our main objective was to determine which legal documents were more effective on creating a structural break in generic drugs consumption in Portugal. The results in our relative prices structural break approach indicate three structural breaks on relative prices: in November 2008, in June 2010 and in April 2012. The breaks coincide in time with legislation regarding price reductions and new rules on prescription and dispense of medicines, by physicians and pharmacies, respectively. It was expected, since most part of the legislation published concerns relative prices, that there were structural breaks motivated by legislation concerning reductions in relative prices. However, the result in the April 2012 structural break, were unexpected, since it dealt with prescription 32 and dispense rules, thus not concerning relative prices directly. It seems that these two legal documents triggered a raise in generic drugs price (that ultimately resulted in relative prices increases), by generic drugs manufacturers. Finally, our approach on generic drugs consumption showed that the structural break in generic drugs consumption occurred in November 2008 and coincides in time with a legal package that intended to reduce generic drugs prices. However, this model was not able to indicate any legislation intended to stimulate directly the demand for generic drugs capable to create a structural break in the consumption of generics. Despite this, it does not mean that this legislation has not produced effects the demand for generic drugs, since it may have contributed to the growth trend of generic drugs market shares during the entire period covered. This result allows us to see that the legislation pack intended to stimulate the consumption of generic drugs, imposed by the Economic and Financial Adjustment Program, was not successful and Portuguese authorities lack policy tools to stimulate generic consumption. On the other hand, the positive evolution of the generic drugs market shares has been important to competition on drug market, since it can drive prices down from both generic and brand-name drugs, lightening the household’s expenses on medicines. This study focused on aggregated data published on a monthly basis by INFARMED in its website. Taking that into consideration, future research may go through the same kind of study we conducted, but on a micro-level, selecting the most and/or least sold generic and its brand-name counterpart drugs, and see if that same legislation triggered some structural change in their consumption. Another subject that could be analyzed in further research is to study the impact on the top most imported drugs after the implementation of Law-Decree number 182/2009, and subsequent revisions, to see if there was an increase of importation of such drugs and its impact on prices and market shares. Finally, this study can also be replicated, for further research, but this time analyzing the hospital drug market. 33 This study has offered some insights on the effectiveness of legislation regarding generic drugs consumption, and was conducted using data provided by INFARMED. So, has a direct result of this methodology, it has encountered some limitations, namely the lack of variables .It is so because some data were not available in the entire period covered, namely the Public expenses on the National Health System (data only available since January 2012 onwards), average reimbursement rates (only available for the years 2010 and 2011) and the number of generic drugs available in the market (data only available for the years 2010, 2011 and 2012). It would be interesting to study the impact of these variables on the generic drugs market shares and, therefore, on generic drugs consumption. Although the relative prices can explain almost perfectly the variations of generic drugs market shares, we believe that adding more variables such average reference price and reimbursement rates and physicians prescription patterns, just as Coscelli (2000), can help to provide more insights on physicians and pharmacies role on the demand of generic drugs, in Portugal. Another issue that we faced was that the data of some variables mentioned was not available for the entire period covered, then, it could not be included in the model and we could not stress out the impact that those variable may have on generic drugs consumption. In the face of such limitations, we presented a study that sought to analyze the impact of various types of legislation and found that prices are the only responsible for creating the structural breaks in generic drugs consumption in Portugal. This result enforces the idea, identified by Mendonca (2011), that generic drugs prices have important implications on generic substitution policy not only from potential savings of households and the Portuguese Government, but also on Health authorities on changing public perception on generic drugs. 40 Appendix C – Extended output for the BreakLS estimation of the demand for generic drugs, using HAC estimators to correct serial correlation Dependent Variable: MRKT_SHARE_VOL Method: Least Squares with Breaks Sample: 2007M01 2013M12 Included observations: 84 Break type: Bai tests of breaks in all recursively determined partitions Break selection: Trimming 0.15, Max. breaks 5, Sig. level 0.05 Breaks: 2008M11 HAC standard errors & covariance (Quadratic-Spectral kernel, Newey-West fixed bandwidth) Allow heterogeneous error distributions across breaks Variable Coefficie nt Std. Error t-Statistic Prob. 2007M01 - 2008M10 -- 22 obs C 0.334698 0.127438 2.626359 0.0104 Relative_prices -0.131496 0.073513 -1.788734 0.0775 @TREND -0.000178 0.001169 -0.152268 0.8794 2008M11 - 2013M12 -- 62 obs C 0.123816 0.017617 7.028057 0.0000 Relative_prices -0.026767 0.011252 -2.378976 0.0198 @TREND 0.002175 0.000163 13.33926 0.0000 R-squared 0.980218 Mean dependent var 0.191498 Adjusted R-squared 0.978950 S.D. dependent var 0.056811 S.E. of regression 0.008243 Akaike info criterion -6.690258 Sum squared resid 0.005299 Schwarz criterion -6.516628 Log likelihood 286.9908 Hannan-Quinn criter. -6.620460 F-statistic 772.9949 Durbin-Watson stat 1.970378 Prob(F-statistic) 0.000000 41 Appendix D – Description of the breakpoint specification used in the BreakLS estimation Breakpoint Specification Description of the breakpoint specification used in estimation Summary Estimated number of breaks: 1 Method: Bai tests of breaks in all recursively determined partitions Maximum number of breaks: 5 Breaks: 2008M11 Current breakpoint calculations: Multiple breakpoint tests Bai tests of breaks in all recursively determined partitions Date: 08/28/14 Time: 02:28 Sample: 2007M01 2013M12 Included observations: 84 Breakpoint variables: C RELATIVE_PRICES @TREND Break test options: Trimming 0.15, Max. breaks 5, Sig. level 0.05 Test statistics employ HAC covariances (Quadratic-Spectral kernel, Newey-West fixed bandwidth) Allow heterogeneous error distributions across breaks Sequential F-statistic determined breaks: 1 Scaled Break Test Break F-statistic F-statistic 0 vs. 1 * 2008M11 19.53062 58.59187 1 vs. 2 --- --- --- 1 vs. 2 2006M12 4.578280 13.73484 * Significant at the 0.05 level, Bai-Perron (Econometric Journal, 2003) critical value 13.98. Break dates: Sequential Repartition 1 2008M11 2008M11 42 Appendix E – Breusch-Godfrey Serial Correlation LM Test extended output results for the BreakLS estimation of the demand for generic drugs Breusch-Godfrey Serial Correlation LM Test: F-statistic 2.180572 Prob. F(2,76) 0.1200 Obs*R-squared 4.558622 Prob. Chi-Square(2) 0.1024 Test Equation: Dependent Variable: RESID Method: Least Squares Sample: 2007M01 2013M12 Included observations: 84 Presample missing value lagged residuals set to zero. Variable Coefficient Std. Error t-Statistic Prob. @BEFORE("2008M11") -0.017025 0.052440 -0.324665 0.7463 @BEFORE("2008M11")*RELATIVE_PRIC ES 0.009714 0.030532 0.318154 0.7512 @BEFORE("2008M11")*@TREND 0.000175 0.000466 0.375679 0.7082 @AFTER("2008M11") -0.006172 0.022144 -0.278706 0.7812 @AFTER("2008M11")*RELATIVE_PRICE S 0.003686 0.014490 0.254359 0.7999 @AFTER("2008M11")*@TREND 5.73E-05 0.000192 0.298462 0.7662 RESID(-1) -0.001941 0.116092 -0.016717 0.9867 RESID(-2) -0.242819 0.116280 -2.088232 0.0401 R-squared 0.054269 Mean dependent var -2.09E-18 Adjusted R-squared -0.032837 S.D. dependent var 0.007990 S.E. of regression 0.008121 Akaike info criterion -6.698436 Sum squared resid 0.005012 Schwarz criterion -6.466930 Log likelihood 289.3343 Hannan-Quinn criter. -6.605372 Durbin-Watson stat 1.991777