Price strategies and economic uncertainty: Case-study for the Argentinian pharmaceutical sector using machine learning

Gutiérrez, Emiliano; Virdis, Juan Marcelo; Meller, Leandro; Domínguez, Diego Leandro Article Price strategies and economic uncertainty: Case-study for the Argentinian pharmaceutical sector using machine learning Revista de Métodos Cuantitativos para la Economía y la Empresa Provided in Cooperation with: Universidad Pablo de Olavide, Sevilla Suggested Citation: Gutiérrez, Emiliano; Virdis, Juan Marcelo; Meller, Leandro; Domínguez, Diego Leandro (2024) : Price strategies and economic uncertainty: Case-study for the Argentinian pharmaceutical sector using machine learning, Revista de Métodos Cuantitativos para la Economía y la Empresa, ISSN 1886-516X, Universidad Pablo de Olavide, Sevilla, Vol. 38, pp. 1-16, https://doi.org/10.46661/rev.metodoscuant.econ.empresa.6407 This Version is available at: https://hdl.handle.net/10419/314469 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-sa/4.0/ ARTÍCULOS 1 Cómo citar: Meller, L., Virdis, J., Gutiérrez, E., & Domínguez, D. Estrategias de precios e incertidumbre económica: Un estudio de caso para la industria farmacéutica argentina usando aprendizaje automatizado. Revista De Métodos Cuantitativos Para La Economía Y La Empresa, 38. https://doi.org/10.46661/rev.metodoscuant.econ.empresa.6407 Universidad Pablo de Olavide (España) Revista de Métodos Cuantitativos para la Economía y la Empresa número 38, 2024 ISSN: 1886-516X DOI: 10.46661/revmetodoscuanteconempresa.6407 Sección: Artículos Recibido: 23-12-2021 Aceptado: 05-04-2024 Publicado: 03-12-2024 Páginas: 1-16 Estrategias de precios e incertidumbre económica: Un caso de estudio aplicado al sector farmacéutico argentino usando aprendizaje automatizado Price strategies and economic uncertainty: case-study for the Argentinian pharmaceutical sector using machine learning Emiliano Gutiérrez Universidad Nacional del Centro de la Provincia de Buenos Aires (Argentina) https://orcid.org/0000-0002-6424-996X emiliano.gutierre[email protected] Juan Marcelo Virdis Universidad Nacional del Sur (Argentina) https://orcid.org/ 0000-0001-7118-9259 [email protected] Leandro Meller Instituto de Investigaciones Económicas y Sociales del Sur (Argentina) https://orcid.org/ 0000-0003-1246-4198 [email protected] Diego Leandro Domínguez Universidad Nacional del Sur (Argentina) dieguini[email protected] RESUMEN Introducción: En agosto de 2019, un resultado inesperado en las elecciones presidenciales generó una variación en el tipo de cambio y la inflación esperados. El Revista de Métodos Cuantitativos para la Economía y la Empresa N. 38, 2024 – ISSN: 1886-516X – DOI: 10.46661/revmetodoscuanteconempresa.6407 – [Págs. 1-16] Estrategias de precios e incertidumbre económica: Un caso de estudio aplicado al sector farmacéutico argentino usando aprendizaje automatizado Emiliano Gutiérrez, Juan Marcelo Virdis, Leandro Meller, Diego Leandro Domínguez ARTÍCULOS 2 objetivo de este estudio es analizar la relación entre la participación de mercado y la decisión de incrementar los precios en la industria farmacéutica en Argentina. Métodos: Se obtuvieron datos semanales en línea sobre las variaciones de los precios de algunos medicamentos mediante técnicas de web scrapping, y luego se aplicaron algoritmos de clasificación (Random Forests, Gradient Boosting Machine y regresión logística). Resultados: Los resultados fueron dispares. Se encontró que la participación de mercado es importante de acuerdo a los métodos basados en árboles (Random Forests y Gradient Boosting Machine). Sin embargo, en la regresión logística, dicha variable no era significativa. Conclusiones: La volatilidad en el tipo de cambio que siguió al resultado de la elección causó varios cambios en los precios esperados, y la estructura del mercado farmacéutico influyó sobre las reacciones de precios resultantes. Los laboratorios que tenían una mayor participación de mercado incrementaron sus precios primero. PALABRAS CLAVE Mercado farmacéutico; incertidumbre; aprendizaje automatizado. ABSTRACT Introduction: In August 2019 an unexpected presidential election result caused a change in expected exchange and inflation rates. The objective of this study is to analyze the relation between market share and the decision of increasing prices in the pharmaceutical industry in Argentina. Methods: Online weekly data on variations of some medicine’s prices were obtained using web scraping, and then classification algorithms (Random Forests, Gradient Boosting Machine and logistic regression) were applied. Results: The results were mixed: market share was found to have high importance in tree-based methods. (Random Forests and Gradient Boosting Machine). However, in logistic regression, this variable wasn’t significant. Conclusions: Exchange rate volatility after the election result caused several changes on price expectations, and pharmaceutical market structure influenced the resulting price reactions. Laboratories which owned a higher market share rose their prices first. Revista de Métodos Cuantitativos para la Economía y la Empresa N. 38, 2024 – ISSN: 1886-516X – DOI: 10.46661/revmetodoscuanteconempresa.6407 – [Págs. 1-16] Estrategias de precios e incertidumbre económica: Un caso de estudio aplicado al sector farmacéutico argentino usando aprendizaje automatizado Emiliano Gutiérrez, Juan Marcelo Virdis, Leandro Meller, Diego Leandro Domínguez ARTÍCULOS 3 KEYWORDS Pharmaceutical market; uncertainty; machine learning. JEL classification: C89, I10. MSC2010: 62P20, 91B26. INTRODUCTION The relationship between market share and drug prices has been well-documented in prior literature (Lu and Comanor, 1998). Companies wielding greater market power possess the capability to set higher prices, both during the initial launch of a new drug and in the ensuing years. In high inflationary environments, companies may find it necessary to raise prices in order to sustain profits against the backdrop of escalating costs. Considering the variability in market shares among companies, and thus their differing levels of market power, it becomes a subject of inquiry whether all companies can uniformly adjust prices, or if those with greater market power can sustain or augment their profits. In this paper we examine the relation between market share and the ability of increasing prices after an exogenous macroeconomic disruption caused by an unpredicted outcome in the presidential election which took place on August 11th, 2019. After the election, the exchange rate changed dramatically, and the Argentinian Peso (ARS) lost 25% of its value in four days. Also, the expected exchange rate for the end of the year increased significantly from 50 to 66,70 ARS (Central Bank of Argentine Republic [BCRA for its acronym in Spanish], 2019). It is worth noting that there is evidence of increase in inflation rate after devaluation in Argentina (Barberis, 2021; Castiglione, 2017; Otero et al., 2005). These changes in currency prices and expectations might trigger price strategies among economic agents. In this regard, the pharmaceutical sector represents a significant case, as the costs of medicines could pose a barrier to the realization of the right to health, which is a key objective of public policy (Gutman and Larvarello, 2011; Perehudoff et al., 2019). This study aims to determine the influence of exchange rate increase and pharmaceutical sector concentration on prices after the election result on August 11th 2019. The rest of the paper is organized as follows. The second section presents previous theoretical and empirical studies about the relation between uncertainty, exchange rate changes and industrial concentration, and price adjustments. The third section describes methodology and data. The fourth section shows the results of this investigation. The fifth section presents a discussion of the results. Finally, the sixth section contains concluding comments. BACKGROUND Economic policy as a source of uncertainty is currently a relevant topic. This is supported by the rapid growth in literature studying the relation between uncertainty of economic policy and the behavior of several microeconomic and macroeconomic variables (Al-Thaqeb and Algharabali, 2019). A large number of existing studies have examined how Economic Policy Uncertainty Index (EPU) (Baker et al., 2016) interrelate with commodity prices (Wang et al., 2015), or consumer price index (Jones and Olson, 2013; Balcilar et al., 2016a). This approach would be appropriate in order to achieve this paper objective. However, no EPU is available in Argentina. Previous empirical research highlights that EPU and exchange rate volatility are positively correlated (Krol, 2014; Balcilar et al., 2016b; Dai et al, 2017; Olanipekun et. al, 2019; Chen et. al, 2020). Similarly, but using more frequent data, Bartshch (2019) finds the same positive correlation. Also, Kurusawa (2016) finds similar evidence, but points out that, in the long term, this correlation is able to change its sign. Revista de Métodos Cuantitativos para la Economía y la Empresa N. 38, 2024 – ISSN: 1886-516X – DOI: 10.46661/revmetodoscuanteconempresa.6407 – [Págs. 1-16] Estrategias de precios e incertidumbre económica: Un caso de estudio aplicado al sector farmacéutico argentino usando aprendizaje automatizado Emiliano Gutiérrez, Juan Marcelo Virdis, Leandro Meller, Diego Leandro Domínguez ARTÍCULOS 4 The influence of the market structure on prices has been recognized by several authors. According to Herguera (1994), the relevance of market structure began to rise in the first half of the 80s. In this period, the American dollar was gaining value but the current account and import prices were unaffected. This situation motivated Mann’s (1986) and Dornbusch’s (1987) investigations, which represented seminal contributions that inspired a large number of studies. These studies developed partial equilibrium models in order to explain why prices in different sectors (even different firms) changed in different magnitudes as a response to exchange rate variations. To this end, Kirman and Phlips (1996) developed a model that shows the relevance of market share to the impact of exchange rate variations on consumer prices. Several empirical investigations also suggest that price variation, as a response to exchange rate changes, is different between sectors (Bhattacharya et. al, 2008; Mallick and Marques, 2010; He et. al, 2015; Thorbecke and Kato, 2018). However, only some of them point out the market characteristics as the cause of these differences (Bhattacharya et. al, 2008; Mallick and Marques, 2010; He et. al, 2015; Thorbecke and Kato, 2018). Pharmaceutical sector in Argentina The pharmaceutical sector in Argentina can be classified as an imperfectly competitive market, characterized by production and consumption complexities. First, the demand is inelastic due to difficulties of avoiding medicine purchasing when needed to reestablish or maintain a healthy status. Second, on the supply side, marketing strategies are implemented by firms, affecting the consumer power of decision, and providing minimum information about their products (Bramuglia et. al., 2015). Third, even though all laboratories represent a low market share (Urbitzondo et al., 2013), the two most important pharmaceutical business chambers own 79% of consumer sales. Similarly, a high concentration is able to be found if the market is divided by therapeutic class (3rd level classification of ATC). In 2019, 4.5% of these divided markets were monopolies, while nearly a 25% of them were characterized by a Herfindahl-Hirschman Index (HHI) higher than 7500 (Statement N° 113 National Commission of Competition Defense [CNDC for its acronym in Spanish], 2019). Even though a part of the supply is offered by the public sector, most of governmental laboratories are not authorized by the National Administration of Medicines, Food and Medical Technology (ANMAT for its acronym in Spanish) to sell their products to the public (Santos and Thomas, 2018; Bramuglia et al., 2012). Also, these laboratories are located in Buenos Aires, Santa Fé and Córdoba, which represents a non-uniform territorial distribution (Santos and Thomas, 2018). This sector interdependence is another relevant factor, mainly, due to its relation to the health care market. This market’s distortions are able to impact on the pharmaceutical one (Alomar et al., 2005). Similarly, medicines are an important component of many households' expenditures which can represent a regressive income factor (Apella, 2006; Perticara, 2008). DATA AND METHODS We obtained medicine prices by web scraping, which is an automatized computational technique used to extract high frequency online information at a lower cost than traditional price surveys. This procedure was applied to the web catalog of a retail pharmacy chain called Farmacity and a total of 29,034 observations were retrieved, one of each week between 12th August and 3rd September. 46% of these observations were classified as price increases, which represents a balanced sample. On the other hand, market participation data, understood as the producing laboratory’s share of retail sales, were obtained from a technical report done by the CNDC. These data are presented in Graph 1. Revista de Métodos Cuantitativos para la Economía y la Empresa N. 38, 2024 – ISSN: 1886-516X – DOI: 10.46661/revmetodoscuanteconempresa.6407 – [Págs. 1-16] Estrategias de precios e incertidumbre económica: Un caso de estudio aplicado al sector farmacéutico argentino usando aprendizaje automatizado Emiliano Gutiérrez, Juan Marcelo Virdis, Leandro Meller, Diego Leandro Domínguez ARTÍCULOS 5 Graph 1. Retail sales share by laboratory Source: Own elaboration based on a technical report of pharmaceutical market structure (Statement N° 113 CNDC, 2019). A dependent variable ( ) was defined as a dummy equal to one if a medicine price, measured in ARS, was higher than the previous week. Otherwise, was set to zero. As independent variables we used the market share of the medicine manufacturing laboratory ( and the difference between the exchange rate in the previous week and the one before . The variable represents the observed exchange rate variation. The variable that represents the exchange rate’s evolution is established as the difference of the maximum values ( ) of the two previous weeks, as exposed in the following equation The exchange rate was obtained from the sale price of the American Dollar, as published by BCRA. Graph 2 depicts the daily exchange rates during the period of analysis. A sudden increase is observed after August 11, 2019, the day of the presidential election. Graph 2. Daily exchange rate (29 July 2019 - 5 September 2019) Source: BCRA (2024). Revista de Métodos Cuantitativos para la Economía y la Empresa N. 38, 2024 – ISSN: 1886-516X – DOI: 10.46661/revmetodoscuanteconempresa.6407 – [Págs. 1-16] Estrategias de precios e incertidumbre económica: Un caso de estudio aplicado al sector farmacéutico argentino usando aprendizaje automatizado Emiliano Gutiérrez, Juan Marcelo Virdis, Leandro Meller, Diego Leandro Domínguez ARTÍCULOS 6 We find it important to distinguish between pharmaceutical national laboratories and foreign counterparts. A dichotomous variable (lab_arg) takes a value of 1 if the laboratory producing the medicine is Argentine, and 0 if it is a foreign laboratory. This information was extracted from the Industrial Chamber of Argentine Pharmaceutical Laboratories webpage (CILFA, acronym in Spanish). The choice of whether to increase the price or not may be significantly influenced by the drug associated with each medicine. For this reason, we introduce principal drugs as dichotomous variables. We consider only the 15 most frequently used drugs in our sample. Additionally, the weeks following the election results serve as an important signal for understanding how the price increase adjusts over time. It is expected that detecting a price increase becomes less likely in the subsequent weeks after the event. Therefore, an ordinal variable (t) is included in the estimations. To fulfill the objectives of this investigation, we conducted estimations using logistic regression, Random Forests, and Gradient Boosting Machine. The dataset was split into a training set (90%) and a test set (10%). Therefore, the training set was utilized for parameterizing the models, and the performance of each model was evaluated using the test set. All estimations were implemented using the R programming language (R Core Team, 2022). With the exception of logistic regression, specific packages in R were employed to execute the algorithms. The Logistic Regression is useful when the dependent variable is binary. A Bernoulli distribution of the predicted event is assumed, what in this paper is the increase of a medicine price ( . The regression’s parameters correspond to the independent variables and the regression is modeled as follows where is the independent variable vector. If the denominator is simplified and the exponents are deleted, the previous equation is expressed as follows The previous expression indicates that the logistic model can be stated as the quotient of event probabilities (odds-ratio). Also, it shows how the coefficients directly impact on this ratio (Hair et al., 2014). Breiman (2001) has suggested the utilization of classification and regression trees (CART) in order to generate a classifier algorithm by applying the mean of a large enough number of interrelated trees. This algorithm, called Random Forest (RF), assumes that multiple tree creation reduces the variability in comparison to any individual tree, and minimizes the final model overfitting. This corresponds to an assembly methodology, because the final model is composed of hundreds (or thousands) of predictive ones, independently created (Hastie, 2009). The first step consists of subsampling a part of the observations by bootstrapping (Typically, 2/3 of the training data). Every generated tree will be composed by a number of predictors (James et al., 2013). This procedure is performed times, generating a different classification tree every time. As a result, the prediction of a higher price is decided by votes, which means that the medicine final classification will depend on how it was classified most of the times: In this research, the implementation of RF algorithm is through the packages caret (Kuhn, 2008) and Ranger (Liaw & Wiener, 2002) from R. Revista de Métodos Cuantitativos para la Economía y la Empresa N. 38, 2024 – ISSN: 1886-516X – DOI: 10.46661/revmetodoscuanteconempresa.6407 – [Págs. 1-16] Estrategias de precios e incertidumbre económica: Un caso de estudio aplicado al sector farmacéutico argentino usando aprendizaje automatizado Emiliano Gutiérrez, Juan Marcelo Virdis, Leandro Meller, Diego Leandro Domínguez ARTÍCULOS 7 In contrast to RF, in which every tree is independent to the previous ones, Friedman (2001) has suggested the construction of an algorithm by a sequential tree building, which allows using previous tree’s information. This methodology is named Gradient Boosting Machine (GBM), and it minimizes a loss function that measures the difference between the real and estimated value. Consequently, the classifying method proceeds recursively times in order to adjust the parameter using previous models (F). Therefore, the last iteration of the model is able to be established as The value is called the learning rate and, in comparison to RF, the algorithm is indifferent about the variables employed in every tree, but it considers the ramifications in each one . The implementation of GBM applied was with the gbm package (Greenwell et al.,2022) An inherent challenge in tree-based models such as Gradient Boosting Machine (GBM) and Random Forests (RF) is the loss of interpretability. Unlike linear models like logistic regression, where coefficients indicate the contribution of each variable to the outcome, "black-box" models like RF and GBM lack this transparency. Addressing this issue, an intriguing question arises: how can we evaluate the impact of features in such "black-box" models? A valuable approach to enhance interpretability in machine learning is the utilization of model-agnostic methods. The model-agnostic approach refrains from imposing assumptions about the intrinsic characteristics of the model (Molnar, 2022). While various methods exist, we specifically selected the permutation feature importance method to assess the interpretability of our tree-based models. This approach enhances explicability without relying on assumptions about the internal structure of the model. In the feature importance procedure, the conventional approach involves running a model with all variables in the original data . Subsequently, the variable of interest in the vector with the total independent variables is permuted, and the model is rerun. The resulting matrix disrupts the linkage between and the variable of interest in predicting . The metric used to measure the loss of feature importance due to permutation is the cross entropy (CE), which quantifies incorrect predictions. This metric is defined as: Here is the probability that a price was higher than the previous week, and a low value of CE indicates a well-fit model. High values of CE suggest that the variable is crucial to the prediction, as its absence adversely affects the model's performance. In contrast, the package iml (Molnar et al., 2018) provides an interesting application of Feature Permutation Importance. RESULTS The logistic regression estimation results, as presented in Table 1, indicate a significant relationship between a price increase and variations in the exchange rate. Additionally, it reveals a statistically significant association between the decision to raise prices and whether the producer is a national laboratory. The week following the elections plays a crucial role in the strategic decision to anticipate and adjust the current prices. As expected, a greater number of weeks after the elections negatively and significantly impacts the likelihood of a price increase. Revista de Métodos Cuantitativos para la Economía y la Empresa N. 38, 2024 – ISSN: 1886-516X – DOI: 10.46661/revmetodoscuanteconempresa.6407 – [Págs. 1-16] Estrategias de precios e incertidumbre económica: Un caso de estudio aplicado al sector farmacéutico argentino usando aprendizaje automatizado Emiliano Gutiérrez, Juan Marcelo Virdis, Leandro Meller, Diego Leandro Domínguez ARTÍCULOS 8 Table 1. Logistic regression Coefficient Standard error t-value p-value (Intercept) -0,676 0,055 -12,288 0,000* t -0,145 0,022 -6,704 0,000* conc -0,946 0,732 -1,292 0,196 v_usd 7,444 0,121 61,548 0,000* lab_arg 0,313 0,036 8,633 0,000* carvedilol -0,065 0,130 -0,500 0,617 metformina -0,035 0,132 -0,267 0,789 pregabalina 0,252 0,136 1,847 0,065 rosuvastatina 0,203 0,142 1,434 0,152 quetiapina 0,232 0,157 1,476 0,140 alprazolam 0,170 0,154 1,107 0,268 enalapril 0,533 0,153 3,480 0,001* ibuprofeno 0,461 0,155 2,984 0,003* diclofenac -0,225 0,158 -1,422 0,155 amoxicilina 0,133 0,159 0,836 0,403 atorvastatín 0,518 0,162 3,190 0,001* sildenafil -0,085 0,164 -0,521 0,603 clonazepam 0,236 0,167 1,416 0,157 risperidona -0,152 0,177 -0,859 0,390 claritromicina 0,026 0,175 0,146 0,884 N = 28807 AIC = 27872 *'=: significant (Confidence Interval of 95%) Source: own elaboration. Another noteworthy finding is the lack of significance regarding the market share. According to the logistic regression estimation, there seems to be no linear relationship between the decision to increase the selling price and the percentage of sales held by the laboratory across the brand. The confusion matrix of the logistic regression (Table 2) illustrates the model's performance on the training set. The logistic regression accuracy stands at 59.94%, with better performance in predicting non-increased prices (60.36%) than in predicting price increases (56.52%). Revista de Métodos Cuantitativos para la Economía y la Empresa N. 38, 2024 – ISSN: 1886-516X – DOI: 10.46661/revmetodoscuanteconempresa.6407 – [Págs. 1-16] Estrategias de precios e incertidumbre económica: Un caso de estudio aplicado al sector farmacéutico argentino usando aprendizaje automatizado Emiliano Gutiérrez, Juan Marcelo Virdis, Leandro Meller, Diego Leandro Domínguez ARTÍCULOS 15 Gutman, G. & Lavarello, P. (2011). Formas de organización de las empresas biotechnológicas en el sector farmacéutico argentino. Desarrollo Económico, 51 (201), 81-105. https://www.jstor.org/stable/23612337 Hair, J. F., Black, W. C., Babin, B. J., & Anderson, R. E. (2014). Multivariate data analysis: Pearson new international edition. Essex: Pearson Education Limited. Hastie, T., Tibshirani, R., and Friedman, J. (2009). The elements of statistical learning: Data mining, inference, and prediction. Springer Science & Business Media. He, B., Zhu, H., Chen, D., & Shi, Y. (2015). On pass-through of RMB exchange rate to prices of different industries. Procedia Computer Science, 55, 886-895. https://doi.org/10.1016/j.procs.2015.07.146 Herguera, I. (1994). Industry price adjustment to exchange rate fluctuations in oligopoly: An empirical study of the pass-through relationship determinants in the Spanish automobile industry, 1981-1991 (PhD Thesis). European University Institute, Florence. https://op.europa.eu/es/publication-detail/- /publication/213f69a9-c5b3-4219-bf1d-a5195f9ab738 James, G., Witten, D., Hastie, T., & Tibshirani, R. (2013). An introduction to statistical learning (Vol. 112, p. 18). New York: springer. Jones, P. M., & Olson, E. (2013). The time-varying correlation between uncertainty, output and inflation: Evidence from a DCC-GARCH model. Economics Letters, 118, 33-37. https://doi.org/10.1016/j.econlet.2012.09.012 Kirman, A., & Phlips, L. (1996). Exchange rate pass-through and market structure. Journal of Economics, 64 (2), 129-154. https://doi.org/10.1007/BF01250111 Krol, R. (2014). Economic policy uncertainty and exchange rate volatility. International Finance, 17 (2), 241255. https://doi.org/10.1111/infi.12049 Kurasawa, K. (2016). Policy uncertainty and foreign exchange rates: The DCC-GARCH model of the US/Japanese foreign exchange rate. International Journal of Economic Sciences, 5 (4), 1-19. https://doi.org/10.52950/ES.2016.5.4.001 Kuhn, M. (2008). Building Predictive Models in R Using the caret Package. Journal of Statistical Software, 28(5), 1–26. https://doi.org/10.18637/jss.v028.i05Liaw, A., & Wiener, M. (2002). Classification and Regression by randomForest. The R Journal, 2/3, 18-22. https://journal.r-project.org/articles/RN-2002-022/RN-2002-022.pdf Lu, Z. J., & Comanor, W. S. (1998). Strategic Pricing of New Pharmaceuticals. The Review of Economics and Statistics, 80(1), 108–118. Mallick, S., & Marques, H. (2010). Data frequency and exchange rate pass-through: Evidence from India’s exports. International Review of Economics and Finance, 19, 13-22. https://doi.org/10.1016/j.iref.2009.02.007 Mann, C. L. (1986). Prices, profit margins, and exchange rates. Federal Reserve Bulletin, (Jun), 366-379. https://fraser.stlouisfed.org/files/docs/publications/FRB/pages/1985-1989/31910_1985-1989.pdf Mayo Clinic. (2023). Drugs and Supplements. Atorvastatin (Oral Route). https://www.mayoclinic.org/diseases-conditions/dementia/symptoms-causes/syc-20352013 Medications Committee of the Spanish Association of Pediatrics (2015). Pediamécum. Edición 2015. https://www.aeped.es/comite-medicamentos/pediamecum/enalapril. Molnar, C. (2022). Interpretable Machine Learning: A Guide for Making Black Box Models Explainable (2nd ed.). https://christophm.github.io/interpretable-ml-book/ Revista de Métodos Cuantitativos para la Economía y la Empresa N. 38, 2024 – ISSN: 1886-516X – DOI: 10.46661/revmetodoscuanteconempresa.6407 – [Págs. 1-16] Estrategias de precios e incertidumbre económica: Un caso de estudio aplicado al sector farmacéutico argentino usando aprendizaje automatizado Emiliano Gutiérrez, Juan Marcelo Virdis, Leandro Meller, Diego Leandro Domínguez ARTÍCULOS 16 Molnar, C., Casalicchio, G., & Bischl, B. (2018). iml: An R package for Interpretable Machine Learning. The Journal of Open Source Software, 3 (26), 786. https://doi.org/10.21105/joss.00786 Olanipekun, I. O., Olasehinde-Williams, G., & Güngör, H. (2019). Impact of economic policy uncertainty on exchange market pressure. SAGE Open, 9(3), 1-13. https://doi.org/10.1177%2F2158244019876275 Otero, G. A., Cadelli, M. E., Carbajal, R., y Cerimedo, F. (2005). Explorando los determinantes del traspaso de la devaluación a precios: Una explicación del éxito devaluatorio argentino de 2002 (documento de trabajo). Grupo de Investigación Económica (GIE) del Ministerio de Economía de la Provincia de Buenos Aires. https://www.ec.gba.gov.ar/prensa/Archivos/Julio2005.pdf Perticara, M. (2008). Incidencia de los gastos de bolsillo en salud en siete países latinoamericanos. Santos, G., and Thomas, H. (2018). Producción pública de medicamentos. Ciencia, Tecnología y Política, 1(1), 007007. https://www.cepal.org/es/publicaciones/6146-incidencia-gastos-bolsillo-salud-siete-paiseslatinoamericanos Perehudoff, S. K., Alexandrov, N. V., & Hogerzeil, H. V. (2019). The right to health as the basis for universal health coverage: A cross-national analysis of national medicines policies of 71 countries. PLOS ONE, 14(6), e0215577. https://doi.org/10.1371/journal.pone.0215577 R Core Team (2022). R: A language and environment for statistica lcomputing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/ Santos, G., y Thomas, H. (2018). Producción pública de medicamentos: desafíos para una política estratégica en materia de salud, Ciencia, Tecnología y Política, 1 (1), 1-7. https://doi.org/10.24215/26183188e007 Statement N° 113/2019. National Commission of Competition Defense, Buenos Aires, Argentina, 23 de December 2019. https://www.argentina.gob.ar/sites/default/files/im._4_-_medicamentos__disposicion_y_anexo.pdf Thorbecke, W., & Kato, A. (2018). Exchange rates and the Swiss economy. Journal of Policy Modeling, 40 (6), 1182-1199. https://doi.org/10.1016/j.jpolmod.2018.07.002 Urbiztondo, S., Cont, W., & Panadeiros, M. (2013). La competencia en el segmento upstream de la industria farmacéutica argentina. Documento de trabajo, 121. http://www.fiel.org/publicaciones/Documentos/DOC_TRAB_1396372329842.pdf Wang, Y., Zhang, B., Diao, X., & Wu, C. (2015). Commodity price changes and the predictability of economic policy uncertainty. Economics Letters, 127, 39-42. https://doi.org/10.1016/j.econlet.2014.12.030