1. Introduction

Various effects of tobacco use have been studied, mainly related to health status and economic costs. However, neoclassical consumer theory incorporates rationality into its foundations, which assumes that utility is maximized and validates the choice of an optimal basket of goods, whatever it contains. Although it is possible to consider externalities in these models, the rationality assumption and the effectiveness of intra-familiar compensations are questionable. The full consumer sovereignty is also arguable since the humans has social needs. These are not physiological or biological but must be satisfied because innate sociability marks us (Fernández, 2005). This allows smoking to become a transmissible habit. Moreover, there is coercive use caused by addiction to nicotine, with evidence of attempts by individuals to eliminate their consumption.

At the individual level, tobacco consumption is one of the leading causes of death while being preventable and is the only legal product that can cause death when used as its manufacturers indicate (U.S. National Cancer Institute & World Health Organization, 2016). However, it generates external costs. The negative externalities involve damage to health, costs for social security systems financed by taxes, material damages, and loss of human lives. Part of the externalities are considered internal if the smoker’s relatives obtain some compensation, but others remain uncompensated. For example, diseases and consumption decisions affect the capacity to acquire other goods and to invest in physical and human capital. In this sense, an externality occurs through the distribution of household income. Since resources are limited, tobacco consumption creates a dilemma: a share of the family income must be allocated to the purchase of this good and it could not be used to buy basic goods and services. Expenditure reductions on food, housing, or even health and education can occur, to the detriment of the human capital accumulation that allows the breaking of poverty circles (John et al., 2011).

All these implications justify direct censorship against tobacco consumption. Thus, this study is based on a critique of consumer sovereignty and the externalities caused. It is based on a social value over the individual one. Likewise, voluntary control mechanisms tend to be ineffective, and this makes it possible to justify interventions (López & Viudes, 2009). Therefore, this paper aims to determine the effect of tobacco consumption on the expenditure patterns within Costa Rican households, through the reallocation of the expenditure to other goods and services, based on the 2018 National Household Income and Expenditure Survey (NHIES). Studies that have investigated these effects in countries such as Ghana, China, Turkey, Bangladesh, Zambia, and Chile, find that tobacco use is associated with a crowding-out effect in education, health, and housing expenditure. Moreover, it is common to find a crowding-in effect in alcoholic beverage expenditure.

As detailed below, the studies related to tobacco consumption in Costa Rica focus on the effects on health or the smoker profile, but there are no known studies of this nature. This is the first contribution proposed in the country that examines the effect of tobacco consumption on the expenditure share of other goods and services. This research uses information on household income and expenditure contained in the HNIES 2018 to estimate a Quadratic Almost Ideal Demand System (QUAIDS) and thereby approximate Engel curves by implementing the Seemingly Unrelated Regression (SUR) model. Given the presence of heteroskedasticity and endogeneity, it is necessary to adjust the errors and to use an instrumental variable, so that the estimate corresponds to the Generalized Method of Moment Three-Stage-Least-Squares (GMM-3SLS).

The results indicate that in Costa Rica tobacco-consuming households show a crowding-out effect on health expenditure. Likewise, there is evidence of a positive relationship between tobacco consumption and expenditure on alcoholic beverages, so tobacco-consuming households have a higher expenditure share on these substances. This crowding-in effect holds when analyzing every income quintile.

2. Literature review

Economic literature has addressed tobacco consumption and its effects from different perspectives. On the one hand, there are theoretical developments that model smoking as a rational decision by incorporating externalities. For instance, Becker & Murphy (1988) proposed a model of rational addiction, and Gruber & Köszegi (2001) proposed a model in which agents act with temporal inconsistency; that is, while making the decision, they value the immediate benefit of consuming tobacco more than the future health benefit. Nevertheless, there are also criticisms related to the market failures, such as that of López and Viudes (2009): assuming that smokers are aware of the costs and that they consider them is incompatible with the efforts made by consumers themselves to quit smoking. On the other hand, there are empirical studies that analyze the crowding-out generated by tobacco consumption. The latter studies base their methodology on empirical analyses that approximate demand curves in cases in which there is no information on prices and use appropriate curves for the data according to the several types of goods, as well as empirical studies that analyze the household expenditure structure.

In Costa Rica, there are studies on the possible impacts of restricting tobacco consumption and the factors that influence it. Monge (2012) described the Costa Rican tobacco market and the implications of strict tobacco control on employment and consumption. Fonseca et al. (2017) found that the susceptibility and prevalence of smoking have decreased, but also found an increase in the proportion of young people with tobacco addiction. Additionally, being a woman, and living with children under 15 years old or with older adults are factors that reduce the probability of consuming tobacco (Espinoza et al., 2020). Even so, there is no research on the crowding-out effect of tobacco consumption over the family budget allocation.

In contrast, there are empirical studies in other countries that analyze this relationship. Nevertheless, they use previous literature associated with Engel curves and QUAIDS. More specifically, Engel (1857) studied the behavior of families and the allocation of income in various categories of goods, using the expenditure share in these groups as a proxy for welfare; however, there was no established specification for the equation. Later, Working (1943) and Leser (1963) proposed a linear form that relates the expenditure share of one good in the family budget to the natural logarithm of total expenditure. Subsequently, Banks et al. (1997) observed that for some commodities the data did not fit precisely and then developed the QUAIDS, which consists of including a quadratic term for the logarithm of total expenditure. Thus, it is possible to obtain Engel curves consistent with consumer behavior. This considers goods and services that are a luxury for certain income levels and a necessity for others, which adapts the curves for budget shares whose relationship is strongly non-linear.

Several studies use the QUAIDS to estimate conditional Engel curves. The theoretical framework that justifies this implementation is summarized in the difficulty to estimate conditional demand functions since it is necessary a specification of a utility function and information on prices (John, 2008; Nguyen & Nguyen, 2020; Wang et al., 2006). In the literature, it is common to use the SUR model, but the estimation method may vary. The usual one is to implement Three-Stages-Least-Squares (3SLS), although some authors use methods such as Fractional Logit (Pu et al., 2008; Wang et al. 2006) or the combination of 3SLS with a Probit model (Chelwa & Walbeek, 2014).

John (2008) and San & Chaloupka (2016) included tobacco expenditure and total non-tobacco expenditure as covariates but found an endogeneity issue; this was solved through the instrumental variable (IV) method, using the ratio of adult men to adult women and total expenditure or income as instruments. Then, a 3SLS was implemented. Other studies found heteroskedasticity issues, so to correct the errors they used Feasible Generalized Least Squares (Husain et al., 2018; John et al., 2011). In cases where both endogeneity and heteroscedasticity problems arose, authors use GMM-3SLS (Masa et al., 2020; Nguyen & Nguyen 2020).

It is also common to consider the effects according to different income levels. Although John (2008) found that tobacco expenditure had a crowding-out effect on per capita food intake, education, and entertainment expenditure, the estimated effect was similar for low-income and high-income households. On the other hand, Pu et al. (2008) found that low-income households are most vulnerable since they are more affected in terms of food and health. John et al. (2011) found a crowding-out effect in education and clothing expenditure at the country level, and in food in low and middle-income households. Nguyen & Nguyen (2020) found a modest decline in the education expenditure share, but mainly in low-income households. Finally, Paraje & Araya (2018) obtained similar results: a lower expenditure share allocated to education and health for poorer households.

For some studies, the most affected categories were food, health, and education (Wang et al., 2006; San & Chaloupka, 2015; Masa et al., 2020), but in some cases, the crowding-out also occurred in categories such as recreation, transportation, communication, housing, clothing, and financial security (Wang et al., 2006; Husain et al., 2018; Masa et al., 2020). However, a crowding-in effect is also possible, as Husain et al. (2018) found; this crowding-in was in food and health, while others found it in alcohol (Masa et al., 2020; Nguyen & Nguyen, 2020; Paraje & Araya, 2017; Wang et al., 2006). Chelwa & Walbeek (2014), on the other hand, did not find any category in which tobacco implied a crowding-in effect.

3. Methodology

3.1. Approach

This study follows a quantitative and empirical approach. The primary focus is to determine the effect of tobacco consumption on the expenditure patterns within Costa Rican households. This investigation does not delve into the broader health or social implications of tobacco consumption but instead focuses on understanding how households prioritize spending when tobacco is consumed, through the estimation of Engel curves. In other words, it identifies which categories are influenced by tobacco consumption. Besides, since cross-sectional data is used, the aim is not to explain how individuals` behavior changes over time, nor is it intended to explain the mechanism through which expenditure in these categories changes.

3.2. Data

This research uses the 2018 Encuesta Nacional de Ingresos y Gastos de los Hogares (National Household Income and Expenditure Survey or NHIES), carried out by the Instituto Nacional de Estadística y Censos (INEC or National Institute of Statistics and Censuses), whose effective sample corresponds to 7,046 households selected through a probabilistic design of areas, stratified, two-staged, and replicated, with immediately captured interviews that included sociodemographic and socioeconomic characteristics, purchasing and consumption habits, and personal expenditures (INEC, 2020). This data is divided into three distinct datasets: expenditures, individuals, and households. However, only the household dataset is used, which consists of 341 variables and 7,046 observations and gathers information on expenditures and socioeconomic characteristics for each household.[1] Of the total, there are 184 households with reports of having consumed tobacco through cigarettes, tobacco cigars, chewing tobacco or snuff, pipe, and other tobacco products.

3.3. Selected variables

It is necessary to define the total household expenditure and its composition to divide it into various expenditure categories. In this case, the selected variable is total monthly consumption expenditure. This is made up of total expenditure on food and non-alcoholic beverages; alcoholic beverages; tobacco; clothing; housing; water and electricity, grouped as utilities; furniture and home maintenance, renamed as equipment; communication; recreation and culture; education; restaurants and hotels; and, finally, miscellaneous expenditures. All of these, except for total consumption and tobacco expenditure, are divided by net expenditure on tobacco (total consumption expenditure minus tobacco expenditure) so that the expenditure share can be obtained in every category.

As covariates, and following Banks et al. (1997), it is necessary to include the logarithm and the squared logarithm of total consumption expenditure, but this expenditure is eventually instrumented by current gross income. Moreover, control variables are essential. The age and years of education of the household head do not receive any transformation; the number of household members is log-transformed; marital status is converted into a set of dummy variables for each status except married; the number of older adults members is also transformed into a dummy, taking value 1 to indicate the presence of older adults and 0 if not; the number of minors is converted into a dummy to indicate their presence; and, finally, the gender variable is recoded so it takes the value 1 for women and 0 for men.

3.4. Estimation

To estimate the effect of tobacco consumption on expenditure shares, it must be possible to establish a statistical relationship between every expenditure category and a set of explanatory variables. In addition, the expenditure of every group does not have the same behavior, since there are goods that represent luxuries at certain income levels and necessities at other levels. For this reason, this research follows the empirical demand curves used by Engel (1857), which are presented in the form of a quadratic equation system by including a linear and a quadratic term for the logarithm of the expenditure, as developed by Banks et al. (1997). This makes it possible to simultaneously model household decisions and control for sociodemographic characteristics, and thus determine whether there is a different budget allocation among the distinct groups of goods and services in those tobacco-consuming households.

To estimate the Engel curves for the various categories, the functional form is specified as follows:

where represents the expenditure share excluding tobacco of category for household . is a dummy variable that indicates whether the household consumed tobacco and equals 1 if the household was observed with positive tobacco expenditure. is total household expenditure excluding tobacco. is a vector of characteristics that includes gender; age; years of education; marital status of the head of household (in 5 dummies terms: free union, divorced, separated, widowed, or single); a dummy that indicates the presence of minors and another for older adults. Finally, corresponds to the error term.

There is evidence of endogeneity across , since the expenditure structure changes as total expenditure increases (Pu et al., 2008). In this case, the Durbin-Wu-Hausmann test rejects the null hypothesis of exogeneity so total current income is used as the instrumental variable, following Banks et al. (1997) and Masa et al. (2020). Moreover, it is important to test the validity of this instrument since it can lead to inconsistent estimates; in particular, the first stage and the F statistic support that this is a valid instrument.

Due to the nature of these data, it is necessary to test for heteroskedasticity. The Breusch-Pagan test rejects the null hypothesis of homoscedasticity, so it is required to correct the standard errors. Besides, the allocation of household expenditure in one category is correlated with the other categories since households make these decisions simultaneously. This implies that it is likely that the error term of the various categories in (1) is also correlated. The common approach to resolve this is through an equation system using the SUR model. Nevertheless, to consider both heteroskedasticity and endogeneity, the estimation is via GMM-3SLS.

The idea is to determine whether there are categories affected by tobacco use through budget reallocation, considering the income levels. Specific categories are food; clothing; utilities; equipment; communication; recreation and culture; education; restaurants and hotels; and miscellaneous expenditures. This estimation is performed both for the total number of households and for every income quintile. However, all regression equations are estimated simultaneously, so the miscellaneous category is dropped to ensure the adding up restriction and, thus, to avoid collinearity issues.

In equation (1) the dependent variable corresponds to the expenditure share in the previous categories. The coefficient measures the effect of tobacco consumption on the specific share. Thus, is understood as the difference between the mean of the expenditure share in category of tobacco-consuming households and the non-tobacco-consuming households. Considering that is expressed as a percentage, the coefficient says how the expenditure share for category . changes in percentage points when the household consumes tobacco.

As a hypothesis, a crowding-out effect is expected in food, health, and education expenditures. Yet, this effect is expected to be higher in the first income quintiles. Similarly, a particular interest presumes to demonstrate whether there is indeed a crowding-in effect in alcoholic beverages; this relationship is expected to be positive regardless of income level.

4. Results

Table 1 and Table 2 provide the mean values and the corresponding standard deviations (s.d.) for the variables used in the estimation. These values are presented for three distinct groups: the entire country, households that do not consume tobacco, and households that do consume tobacco.

Table 1
Descriptive measures of household and household head characteristics

the author, based on NHIES 2018. since the last four variables are dummies, the mean represents the percentage of households that meet that characteristic.

Variable		Country			Non-consuming			Consuming
		Mean	s.d.		Mean	s.d.		Mean	s.d.
Head of household age		50.47	15.71		50.45	15.73		51.24	14,81
Years of education		7.89	4.42		7.88	4.42		8.08	4,50
Number of household members		3.22	1.63		3.22	1.63		3.21	1,86
Female-headed households		0.38	-		0.39	-		0.37	0,48
Tobacco-consuming households		0.03	-		0.00	-		1.00	-
Older adults’ presence		0.25	-		0.25	-		0.27	-
Minors’ presence		0.50	-		0.50	-		0.40	-

Table 2
Descriptive measures of income and the various expenditures in colones

the author, based on NHIES 2018 all amounts are rounded and expressed in colones

Variable		Country			Non-consuming			Consuming
		Mean	s.d.		Mean	s.d.		Mean	s.d.
Current income		841 966	1 106 432		840 010	1 107 859		914 375	1 052 539
Consumption		524 650	480 724		522 817	479 277		592 517	528 588
Tobacco		578	5 377		-	-		21 989	25 134
Net tobacco		524 072	480 558		522 817	479 274		570 528	525 203
Food		130 661	107 259		130 590	107 228		133 279	108 644
Alcohol		2 033	12 186		1 859	11 872		8 493	19 689
Clothing		24 234	44 437		24 215	44 518		24 948	41 418
Utilities		63 358	81 851		63 213	82 220		68 699	66 738
Equipment		30 808	55 581		30 786	55 724		31 613	50 127
Health		26 757	72 296		26 588	70 461		33 008	122 319
Transportation		82 401	176 759		82 289	176 483		86 559	187 140
Communication		32 286	31 560		32 172	31 503		36 494	33 407
Recreation		32 111	59 022		32 026	58 906		35 272	63 282
Education		22 137	80 282		22 114	80 319		22 979	79 127
Rest. /Hotels		38 108	68 417		37 799	68 094		49 569	78 842
Miscellaneous		39 177	59 371		39 165	59 618		39 614	49 546

Table 1 shows that around 3 % of households reported tobacco consumption during the survey period. Additionally, as seen in Table 2, both the average current income and the average of the several types of consumption expenditure are higher for tobacco-consuming households. The same occurs with the average age of household heads, but not with years of education. Concerning the number of household members, the average is the same, but tobacco-consuming households have a lower average number of minors. In contrast, the presence of older adults is similar, although it is higher in tobacco-consuming households. Finally, most households are headed by men, but the percentage headed by women is lower for tobacco-consuming households.

Table 3 shows the expenditure share in every category for the country and for every income quintile. In categories such as food and utilities, a reduction is observed as the income quintile increases. However, others such as alcohol, equipment, health, transportation, communication, recreation, education, restaurants, and miscellaneous expenditures rise as the quintile increases. The expenditure share on tobacco and clothing remains similar.

Table 3
Expenditure by category as a percentage of total net tobacco expenditure by income quintile

the author, based on NHIES 2018. Q1, Q2, Q3, Q4, and Q5 represent every income quintile, ordered from the lowest to the highest income, while the total represents the entire sample.

	Tobacco	Food	Alcohol	Clothing	Utilities	Equip.	Health	Trans.	Comm.	Rec.	Educ.	Rest. / Hotels	Misc.
Total	0.15	30.00	0.28	4.58	14.96	5.33	4.30	11.99	6.78	5.81	2.45	6.67	6.85
Q1	0.13	39.40	0.10	4.42	17.06	4.51	2.81	8.67	5.84	5.77	0.83	4.27	6.36
Q2	0.16	33.66	0.15	4.66	16.05	5.02	3.78	10.20	6.73	5.45	1.92	5.70	6.68
Q3	0.15	29.83	0.24	4.73	14.97	5.43	4.53	11.84	7.08	5.41	2.17	6.91	6.87
Q4	0.21	24.07	0.41	4.42	14.28	5.47	4.92	14.27	7.52	5.54	3.74	8.55	6.79
Q5	0.13	18.58	0.58	4.71	11.28	6.71	6.13	16.72	7.05	7.11	4.33	9.00	7.77

As mentioned before, the coefficient of interest in equation (1) is , since it allows estimating the effect of consuming tobacco on the expenditure shares. Thus, Table 4 presents the estimated effect in percentage points (p.p.) for every category, both for the country and for every quintile income. This shows that in most groups (food, clothing, services, furniture, transportation, communications, recreation, and education) there is no significant effect. However, at the country level, expenditure share on alcohol, health, and restaurants and hotels show statistically significant relationships with tobacco consumption in Costa Rican households.

Table 4
Effect of tobacco consumption on every expenditure category by income quintile in percentage points

author, based on the estimation using NHIES 2018 Q1, Q2, Q3, Q4, and Q5 represent every income quintile, ordered from lowest to highest income, while the total represents the entire sample. *p <0.10; **p <0.05; ***p <0.01

	Food	Alcohol	Clothing	Utilities	Equip.	Health	Trans.	Comm.	Rec.	Educ.	Rest. / Hotels
Total	0.10	1.34***	-0.49	0.38	-0.30	-0.97**	-0.42	0.01	-0.11	-0.45	1.47*
	(1.04)	(0.28)	(0.34)	(0.88)	(0.44)	(0.49)	(0.88)	(0.41)	(0.44)	(0.48)	(0.86)
Q1	0.90	0.50*	-1.31**	0.80	-0.17	-0.53	1.53	-0.01	-0.51	-0.28	1.03
	(2.53)	(0.30)	(0.63)	(1.87)	(0.67)	(0.70)	(2.26)	(0.89)	(1.05)	(0.40)	(1.31)
Q2	0.47	1.06***	-0.08	0.48	-0.37	-1.53*	2.22	-0.86	-0.92	-0.33	-0.56
	(2.34)	(0.40)	(0.91)	(2.07)	(0.91)	(0.91)	(2.31)	(1.13)	(0.73)	(1.31)	(1.72)
Q3	1.54	0.70**	-0.43	0.68	-0.77	-0.86	-2.55*	0.00	-0.64	-1.87***	3.75
	(2.55)	(0.34)	(0.81)	(2.19)	(0.72)	(0.75)	(1.50)	(0.98)	(0.71)	(0.50)	(2.29)
Q4	0.84	2.46***	-0.49	-0.37	0.08	-1.16	-3.75**	-1.12	-0.85	0.32	5.02**
	(2.01)	(0.91)	(0.66)	(1.86)	(1.22)	(0.92)	(1.87)	(0.72)	(0.73)	(1.42)	(2.48)
Q5	-3.85**	1.67***	0.04	-0.97	-0.57	-0.15	1.90	1.75	2.64*	-0.62	-1.50
	(1.52)	(0.63)	(0.94)	(1.86)	(1.17)	(1.95)	(2.32)	(0.97)	(1.45)	(1.26)	(1.32)

A striking result is a crowding-in effect in the expenditure share allocated to alcoholic beverages, which is 1.34 p.p. This means that, if the household consumes tobacco, the alcohol expenditure increases 478 %.[2] Moreover, this crowding-in is not only shown at the country level but in every income quintile. In the first one, the crowding-in effect is 0.5 p.p., a 500 % increase. In the second one, it is 1.06 p.p., a 707 % increase. In the third one, it is 0.7 p.p., a 292 % increase. In the fourth one, there is an increase of 2.46 p.p., a 600 % increase. Finally, in the fifth one, the increase is 1.67 p.p., a 288 % increase. Thus, a strong relationship between tobacco and alcohol consumption is evident, and it remains consistent across every income level. This can be reasonable explained by considering factors such as social habits, psychological interactions, and addictive tendencies.

Tobacco consumption shows a -0.97 p.p. effect in the health expenditure share, a 22.6 % decrease when the household consumes tobacco. However, this crowding-out is not found when analyzing every quintile except for the -1.53 p.p. effect in the second quintile, which is a 40.5 % decrease.

In restaurants and hotels, the effect is positive and reaches 1.47 p.p.; tobacco consumption leads to an increase of approximately 21.9 % in the expenditure share dedicated to restaurants and hotels. Similarly, this crowding-in is not found in every quintile but only in the fourth one; the effect is 5.02 p.p. or a 58.7 % increase.

In the first quintile, the tobacco consumption effect in the clothing expenditure share is -1.31 p.p., a 29.7 % decrease in the presence of tobacco consumption. In the second quintile, as mentioned, the only crowding-out effect is in the health expenditure share. In the third quintile, there is a result that stands out: the crowding-out in the transportation expenditure share, that is a -2.55 p.p. effect or a 21.5 % decrease, and the crowding-out in the education expenditure share, that is a -1.87 p.p. effect or an 86 % decrease. In the fourth quintile, there is a crowding-out in the transportation expenditure share, a -3.75 p.p. effect or a 26.3 % decrease. In the fifth quintile, the tobacco consumption effect in the food expenditure share is -3.85 p.p. or a 20,7 % decrease. This stands out because for these households there is a crowding-in effect in alcoholic beverages and recreation expenditure shares (2.64 p.p. or a 37.1 % increase), while other categories are not affected.

When comparing the hypotheses and the effects of tobacco consumption shown in table 4, there are mixed results. First, the literature indicates that tobacco involves a crowding-out effect in the expenditure share on food, education, health, and clothing, with special emphasis on the first three categories. This effect was expected to be observed in the expenditure share on education and health in lower-income households. However, this crowding-out effect was found mainly at the country level and in the health expenditure share but not in the food or education expenditure share. Second, the literature shows a crowding-in effect in the alcoholic beverages’ expenditure, which is, in this case, confirmed, and it even holds for the different income levels. Finally, there is no effect in categories such as utilities and equipment, neither at the country level nor in the income quintiles.

6. References

Banks, J., Blundell, R., & Lewbel, A. (1997). Quadratic Engel Curves and Consumer Demand. The Review of Economics and Statistics, 79(4), 527-539. http://www.jstor.org/stable/2951405

Becker, G., & Murphy, K. (1988). A Theory of Rational Addiction. Journal of Political Economy, 96(4), 675–700. http://www.jstor.org/stable/1830469

Engel, E. (1857). Die Productions- und Consumtionsverhaltnisse des Konigreichs Sachsen. Zeitschrift des Statistischen Bureaus de Koniglich-achsischen, Ministerium des Innern, 8(9), 1-5.

Espinoza, A., Fantin, R., Barboza, C., & Salinas, A. (2020). Características sociodemográficas asociadas a la prevalencia del consumo de tabaco en Costa Rica. Revista panamericana de salud pública, 44(17), 1-8. https://doi.org/10.26633/R- PSP.2020.17

Fernández, J. (2005). Algunas reflexiones en torno al problema de la creación de necesidades en publicidad y marketing. Revista Internacional de Comunicación Audiovisual, Publicidad y Estudios Culturales, 1(3), 101-124. https://dialnet.unirioja.es/servlet/articulo?codigo=2142127

Fonseca, S., Méndez, J., Bejarano, J., Guerrero, C., & Reynales, L. (2017). Tabaquismo en Costa Rica: susceptibilidad, consumo y dependencia. Salud Pública de México, 59(1), 30-39. https://doi.org/10.21149/7765

Gruber, J. & Köszegi, B. (2001). Is Addiction “Rational”? Theory and Evidence. The Quarterly Journal of Economics, 116(4), 1261–1303. https://doi.org/10.1162/003355301753265570

Husain, M., Datta, B., Virk-Baker, M., Parascandola, M., & Khondker, B. (2018). The crowding-out effect of tobacco expenditure on household spending patterns in Bangladesh. PloS one, 13(10), 1-15. https://doi.org/10.1371/journal.po- ne.0205120

Instituto Nacional de Estadística y Censos (2020). Encuesta Nacional de Ingresos y Gastos de los Hogares. Metodología. INEC. https://admin.inec.cr/sites/default/files/media/meenigh2018_0_2.pdf

John, R. (2008). Crowding out effect of tobacco expenditure and its implications on household resource allocation in India. Soc Sci Med, 66(6), 56-67. https://doi.org/10.1016/j.socscimed.2007.11.020.

John, R., Ross, H., & Blecher, E. (2012). Tobacco expenditures and its implications for household resource allocation in Cambodia. Tobacco Control, 21(3), 341-346. http://www.jstor.org/stable/41516453

John, R., M., Sung, H.-Y., Max, W. B., & Ross, H. (2011). Counting 15 million more poor in India, thanks to tobacco. Tobacco Control, 20(5), 349 352. https://doi.org/10.1136/tc.2010.040089

Leser, C. (1963). Forms of Engel functions. Econometrica: Journal of the Econometric Society, 694-703.

López N. & Viudes, A. (2009). El control del tabaquismo desde la perspectiva de la economía. Revista Española de Salud Pública, 83(1), 9-23. https://scielo.isciii.es/scielo.php?script=sci_arttext&pid=S1135-57272009000100002

Masa, A., Chelwa, G., & van Walbeek, C. (2020). Does tobacco expenditure influence household spending patterns in Ghana? Evidence from the Ghana 2012/2013 Living Standards Survey. Tobacco induced diseases, 18(48), 1-8. https://doi.org/10.18332/tid/120936

Monge, M. (2012). Economía del tabaco. Acta Médica Costarricense, 54(1), 8-14. http://www.scielo.sa.cr/scie-lo.php?script=sci_arttextypid=S0001-60022012000100003ylng=enytlng=es

Nguyen, N., & Nguyen, A. (2020). Crowding-out effect of tobacco expenditure in Vietnam. Tobacco Control, 29(5), 326-330. https://doi.org/10.1136/tobaccocontrol-2019-055307

Organización Mundial de la Salud (2008). Informe OMS sobre la epidemia mundial del tabaquismo, 2008. http://apps.who.int/iris/bitstream/handle/10665/43897/9789243596280_spa.pdf?sequence=1

Paraje, G., & Araya, D. (2018). Relationship between smoking and health and education spending in Chile. Tobacco Control, 27(5), 560 - 567. https://doi.org/10.1136/tobaccocontrol-2017-053857

Pu, C., Lan, V., Chou, Y. & Lan, C. (2008). The crowding-out effects of tobacco and alcohol where expenditure shares are low: analyzing expenditure data for Taiwan. Soc Sci Med, 66(9),1979-1989. https://doi.org/10.1016/j.socs- cimed.2008.01.007.

San, S., & Chaloupka, F. (2016). The impact of tobacco expenditures on spending within Turkish households. Tobacco control, 25(5), 558–563. https://doi.org/10.1136/tobaccocontrol-2014-052000

U.S. National Cancer Institute & World Health Organization (2016). The Economics of Tobacco and Tobacco Control. Bethesda. Department of Health and Human Services, National Institutes of Health, National Cancer Institute; and World Health Organization.

Wang, H., Sindelar, J. & Busch S. (2005). The impact of tobacco expenditure on household consumption patterns in rural China. Soc Sci Med, 62(6),14-26. https://doi.org/10.1016/j.socscimed.2005.07.032

Working, H. (1943). Statistical laws of family expenditure. Journal of the American Statistical Association, 38(221), 43-56.

Notas