You are on page 1of 19

Impatience and excess consumption of soft drinks, tobacco, alcohol and tortillas in Mexico

Carlos Romero Hernández1


Draft July 14 2019

Abstract

This paper analyzes the hypothesis that Mexican consumers are impatient in the consumption of
goods subject to excise taxes such as tobacco, alcohol and soft drinks. For comparing purposes, the
analysis includes the tortilla, the primary food of the Mexican meal. Estimates of hyperbolic discount
rate for household consumption, in 2016, 2014 and 2012, show different degrees of impatience in
consumption of tobacco, soft drinks, and alcohol, but not in tortillas. These estimates show different
behavior by four socioeconomic strata of households.

Key words: Present bias, self-control problems, non-rational, hyperbolic discount, soft drinks

JEL D91

Introduction

Consumer goods such as tobacco, alcoholic beverages and soft drinks are subject to excise taxes in
order to control their consumption through an increase in prices. Excess consumption is caused by
an addiction or by the instant gratification that these products provide to consumers.

Excessive consumption is a problem of self-control that is part of a set of irrational behaviors in


economic terms, that affect people's decisions (Reisch and Zhao, 2017). The problem of self-control
is analyzed as the behavior of impatient consumers who make purchases without considering the
future consequences of their decisions (Courtemanche, Heutel, and McAlvanah, 2014).

This paper analyzes the hypothesis that consumers are impatient in the consumption of goods
subject to excise taxes such as tobacco, alcohol and soft drinks. The behavior of impatient
consumers considers the future value of consumption according to a discount rate. Studies on the
subject point out that the higher the discount rate, or the lower the discount factor, the more
consumption is advanced to the present, and the consumers respond less to short-term changes in
market prices (Richards and Hamilton, 2012). An opposite perspective considers that the excess
consumption is a rational addiction created by the past, accumulated consumption of the good
(Becker, Grossman, and Murphy, 1991).

The studies that analyze the discount rate are based on experiments; a group of people, usually
students, who are presented with a decision problem and the answers are analyzed to estimate the

1
Independent Consultant in Economic competition and regulation issues. mail: cromero@eeconsultoria.com

Electronic copy available at: https://ssrn.com/abstract=3423028


discount rate (Rubinstein, 2003). The impatience of consumers is analyzed with hyperbolic discount
models, that is, with discount rates that vary over time (Heal, 2007).

This paper presents estimates of a hyperbolic discount model for the Mexican household
consumption of soft drinks, tobacco and alcohol. The consumption of tortilla2 is included to compare
the results. The analysis uses consumption data as reported in the National Household Income
Expenditure Survey, ENIGH for its acronym in Spanish, carried out every two years by the Mexican
statistical authority, INEGI. It is particularly interesting to note in the survey of 2016 the impact of
the specific tax on soft drinks established in 2014. The time variable is approximated by the average
age of household members. The model is estimated with quantile regression and the result of the
estimation of the median is presented.

The discount model is estimated for the total sample of households that report the consumption of
each product and four strata of households. The results obtained are:

• Estimates of the discount rate on soft drinks consumption for 2016, 2014 and 2012. The discount
rate is higher for the households of lower average age and the consumption is advanced to a greater
degree, in the following order, in the stratum of low, medium-low and medium-high income homes.
For stratum four, of high-income level, the consumption of soft drinks is not advanced.

• The discount rate on the consumption of tobacco, alcohol and tortillas for 2016 shows different
behaviors. The consumption of tobacco is advanced but not the consumption of tortillas and it can
even be interpreted that it is postponed for the four strata. The consumption of alcohol is advanced,
relatively, less than the consumption of soft drinks.

The discount rate of consumption, as noted, reduces the effect of taxes to effectively control excess
consumption of goods. However, the estimated model does not indicate in what magnitude or to
what extent the effect of taxes is neutralized. For this, it is necessary to estimate the elasticities of
demand.

Irrational behavior

The behavior of consumers is a variable that is rarely considered in the design of public policy. The
latter makes an implicit assumption about how people make decisions and process the information
available in the market, such as advertising and product standards. However, several studies show
the gap between what public policy assumes on the expected behavior of consumers and their
actual behavior.

A case of public policy is the frontal labeling of consumer goods used in several countries and in
Mexico (SE, 2014). This type of labeling assumes that any consumer can understand, interpret and
analyze data related to the contents of fats, sugars, sodium, or other items, presented in the labels

2
Tortilla is soft flat bread prepared mainly of corn dough. This meal is the main component of Mexican food
basket.

Electronic copy available at: https://ssrn.com/abstract=3423028


of products. However, some studies indicate that consumers do not use labeling to make and
regulate their decisions (INSP-SSA, 2016, pp. 11-12) (Storcksdieck and Wills, 2012, p.136).

Impatience in consumption is another behavior that is not considered in public policy. This problem
introduces the time factor when the consumption of goods a product is advanced or not. If the
future consumption of a good has a value of 100 units and at present that consumption has a value
of, for example, 90 or 80 units, then it implies discount rates of 10% or 25%. That is, the higher the
discount rate the lower the current value of a future consumption. This result is the basis of studies
on health problems, such as obesity, discussed by Courtemanche, Heutel, and McAlvanah (2014)
and Richards and Hamilton (2012).

Hyperbolic discount model of consumption

Heal (2007, p.69) discusses the discount models used in the analysis of intertemporal consumer
decisions. Particularly, he presents a model used in the natural sciences which can be adapted to
represent the economic discount rate. The model postulates that the response, R, to a stimulus, S,
is proportional, by a factor K, and not linear, represented by the logarithm of the stimulus, Ln (S),
that is: R = K Ln (S).

In this response-stimulus model, the economic discount factor is obtained by considering the time
T as a stimulus and, as it progresses, the response decreases. In a usual way, the discount factor is
expressed in its exponential form with a negative sign: 𝐹𝑇 = 𝑒 −𝐾𝐿𝑛(𝑇) = 𝑇 −𝐾 . In this way, the
discount rate is obtained from the rate of change of the discount factor, ie: F´/F = -K/T. This is a
hyperbolic discount because the discount rate varies over time.

The relevant economic value of the hyperbolic discount is in the range 0 <F <1: i) if F = 0 it implies
that future consumption has no value, that is, there is a complete loss of value. A small value of F or
approaching zero indicates a higher degree of impatience of consumers; and ii) if F = 1 or its value
tends to increase it implies that future consumption has no loss, unlike the previous case, meaning
that the consumer is not impatient or does not advance their consumption. A value of F> 1 is
possible and it can be interpreted as if present consumption is postponed.

Estimation of the consumer discount model

The amount of consumption of the products is obtained from the National Household Income
Expenditure Survey, ENIGH, conducted in Mexico every 2 years. The amount consumed of a product
provides a gratification to the consumer that explains the excess of consumption (Richards and
Hamilton, 2012). If we consider a group of variables X that represent characteristics of households,
the product consumption function is represented as C(X). It should be noted that the demand
function is not estimated which requires more market information.

The households represent families of different sizes whose members are young children, adults and
older adults. A relatively young or adult family would be indicating different age compositions with
greater or lesser presence of children and older adults. This composition would affect the amount
3

Electronic copy available at: https://ssrn.com/abstract=3423028


purchased of goods or the average household consumption. Young households can have higher
discount rates, just like young people, but that discount can be nuanced by adults. For example, if
adults suffer from diseases such as smoking, obesity or diabetes, excess consumption may be
tempered not only by people with illnesses but also by extension by young family members. This
would be expressed in different consumption discounts according to the average age of the
households.

To analyze household consumption and consumption discount, it is proposed to estimate a Cobb-


Douglas function that includes the hyperbolic discount factor based on the average age of
households; the latter would represent the implicit discount that reveals the consumption of the
goods at the time analyzed. The function of the consumption quantities is represented as:
𝐶(𝑋, 𝑒 −𝐾𝐿𝑛(𝐸) ) where X is the explanatory variables and the model is expressed as:

𝛼𝑗
𝐶 = 𝑒 −𝐾𝐿𝑛(𝐸) ∏ 𝑋𝑗

Beverage consumption behavior in 2012, 2014 and 2016

The consumption of soft drinks per person is the relevant indicator to analyze the consumption
pattern of this product. This work uses the average consumption per person per household. This
average consumption adds two effects. On the one hand, although consumption is personal, the
decision on how much to consume depends, on average, on the number of household members.
On the other hand, it is possible that the presence in the home of people with health conditions
generated by the consumption of a certain product has the effect of reducing the consumption of
that product. This may be because the family assimilated the consequence of the consumption or
because a part of the income is assigned to treat illnesses.

The consumption of soft drinks per person per household is obtained from the household surveys
of 2012, 2014 and 2016. Household sample uses quarterly consumption data. In 2016, the sample
of households is 45,693, in 2014 the sample is 12,235 households and in 2012 the sample is 5,792.
Annex 1 describes the variables used.

The consumption of soft drinks by socioeconomic level is classified in four income strata of
households: Low, Low-middle, High-middle and High, which identifies socioeconomic characteristics
of people and households. In 2016, 21.1% of households (9,623) are in the low stratum and 53.9%
(24,607) in the low-middle stratum.

Table 1 shows the consumption of soft drinks per person per household. On average between 2014
and 2016 consumption increased 1.5% and between 2012 and 2014 it decreased 4.3%. In 2014
consumption decreases in all strata and the opposite effect is shown in 2016, except in stratum 4.

Electronic copy available at: https://ssrn.com/abstract=3423028


Table 1. Consumption of soft drinks per household per person by socioeconomic strata of
households, 2016, 2014 and 2012

Stratum of 2016 2014 2012 Change in


households average
consumption
Avg. CV* House- Avg. CV* House- Avg. CV* House- 2016/ 2014/
Hold Hold Hold 14 12
Stratum 1 1.523 1.09 9,623 1.369 1.08 2,197 1.586 1.07 1,518 11.3% -13.7%
Low
Stratum 2 1.604 1.06 24,607 1.586 1.08 6,478 1.614 0.99 2,948 1.1% -1.7%
Low-middle
Stratum 3 1.688 1.03 8,591 1.689 0.95 2,599 1.767 0.99 1,016 0.0% -4.4%
High-
middle
Stratum 4 1.569 1.03 2,872 1.680 1.17 961 1.863 1.07 310 -6.6% -9.8%
High
Total 1.600 1.06 45,693 1.576 1.06 12,235 1.647 1.02 5,792 1.5% -4.3%
* CV: Standard deviation / Average
Source: ENIGH

Regarding prices, these are obtained from the proportion (expenditure / quantity of consumption).
Table 2 shows the increase in average prices between 2016 and 2014 by 7.6% and between 2014
and 2012 the increase was of 11.4%. In 2014, a special tax was applied to soft drinks. In this period,
the price increase is similar in all strata of households, except in stratum 4 between 2016 and 2014.

Table 2. Price per liter of soft drink by socioeconomic strata of households, 2012, 2014 and 2016

Stratum of 2016 2014 2012 Change in


households average prices
Avg. CV* House- Avg. CV* House- Avg. CV* House- 2016/ 2014/
hold Hold hold 14 12
Stratum 1 12.3 41.7 9,623 11.3 40.8 2,197 10.1 40.1 1,518 8.8% 11.5%
Low
Stratum 2 12.3 43.4 24,607 11.4 39.7 6,478 10.3 104.8 2,948 8.1% 10.5%
Low middle
Stratum 3 12.6 43.1 8,591 11.6 35.3 2,599 10.4 33.3 1,016 8.5% 11.4%
High middle
Stratum 4 12.9 46.3 2,872 12.7 44.9 961 11.5 51.6 310 1.9% 10.3%
High
All 12.4 43.2 45,693 11.5 39.7 12,235 10.3 79.5 5,792 7.6% 11.4%
* CV = Standard deviation / Average
Source: ENIGH

In 2016, as example, the households that consume soft drinks have as characteristics their income
and the expenditure made on other goods other than soft drinks. The presence or risk of diabetes
that is observed with blood tests carried out by household members is also considered. Other

Electronic copy available at: https://ssrn.com/abstract=3423028


variables are the members of the household who speak an indigenous language and the average
schooling of the members of the households.

The variables used in the model and their characteristics by socioeconomic stratum are shown in
Table 3. The variables are:

i. Age: Average age of household members. The deciles of age are considered to estimate the
discount rate on consumption.

ii. Test: Percentage of household members who had a blood test.

iii. Age Indicator: The variable measures the average age of the household members who makes
blood tests with respect to the average age in the home.

• Indicator> 1 if the average age of those who perform blood tests is greater than the
average age of the household.

• Indicator = 1 if blood tests are not performed at home.

• Indicator <1 if the average age of those who undergo blood tests is less than the average
age of the household

iv. Persons: Number of household members.

v. Total expenditure per person per household: value of consumer goods per person, that is, the
expense among household members of different consumer goods without including the cost of soft
drinks and medicines for diabetes.

vi. Income: Household income.

vii. Indigenous: Percentage of household members over 3 years who speak an indigenous language.

viii. Schooling: School level achieved by household members; no level, primary, secondary, bachelor,
master’s degree, PhD.

Table 3. Statistics of variables, ENIGH 2016

Variables Stratum 1 Stratum 2 Stratum 3 Stratum 4 Total


Households 9,623 24,607 8,591 2,872 45,693
Consumption per person per household
Average 1.523 1.604 1.688 1.569 1.600
s.d. 1.656 1.696 1.734 1.615 1.691
CV 108.7% 105.7% 102.7% 103.0% 105.6%
Prices
Average 12.3 12.3 12.6 12.9 12.4
s.d. 5.1 5.3 5.4 6.0 5.3
CV 41.7% 43.4% 43.1% 46.3% 43.2%

Electronic copy available at: https://ssrn.com/abstract=3423028


Variables Stratum 1 Stratum 2 Stratum 3 Stratum 4 Total
Expenditure: Total expenditure per
person per household, but different to
the product under analysis
Average 6,498 9,286 13,276 20,160 10,133
s.d. 5,936 10,004 13,123 20,921 11,513
CV 91.4% 107.7% 98.8% 103.8% 113.6%
Quarterly Income
Average 23,679 35,184 47,580 71,690 37,387
s.d. 30,110 46,774 64,915 69,954 51,035
CV 127.2% 132.9% 136.4% 97.6% 136.5%
Size: Households members
Average 3.96 3.81 3.57 3.47 3.78
Sum 38,141 93,802 30,629 9,977 172,549
s.d. 1.98 1.85 1.70 1.45 1.84
CV 50.0% 48.6% 47.6% 41.8% 48.6%
Age: average age of household members
Average 33.3 33.5 35.8 34.0 33.9
s.d. 17.4 16.1 15.7 14.5 16.2
CV 52.3% 47.9% 43.9% 42.7% 47.8%
Indicator: average age of members that
make a blood test/ average age of
household members
Average 1.254 1.239 1.218 1.218 1.237
s.d. 0.389 0.373 0.351 0.337 0.370
CV 31.0% 30.1% 28.8% 27.6% 30.0%
Test: people older than 12 years that
makes blood test
Average 1.134 1.134 1.198 1.174 1.149
Sum 10,912 27,909 10,295 3,373 52,489
% members household 28.6% 29.8% 33.6% 33.8% 30.4%
d.e. 1.128 1.091 1.065 1.055 1.092
CV 99.5% 96.2% 88.9% 89.8% 95.1%
Indigenous: people over 3 years who
speak an indigenous language
Average 0.756 0.104 0.026 0.019 0.221
Sum 7,279 2,555 221 56 10,111
s.d. 1.598 0.526 0.306 0.184 0.885
CV 211.2% 506.4% 1190.2% 942.2% 400.0%
Schooling *
Average 2.33 2.95 3.85 4.64 3.09
s.d. 0.97 1.21 1.50 1.61 1.40
CV 41.6% 41.1% 39.0% 34.7% 45.4%
* Schooling: 0 none, 1 preschool, 2 primary, 3 secondary, 4 high school, 5 teaching, 6 technical career, 7
bachelor, 8 master, 9 PhD.
Source: ENIGH

The consumption of soft drinks by deciles of average age of households are shown in Table 4.
Between 2016 and 2014, the increase in consumption is observed except for deciles 1, 3, 9 and 10,
which correspond to the deciles of the lowest and highest average ages.
7

Electronic copy available at: https://ssrn.com/abstract=3423028


Table 4. Consumption of soft drinks per household per person by deciles of age of households,
2016, 2014 and 2012

Decile of age* 2016 2014 2012 Change 2016/14 Change2014/12


< 17.25 1.204 1.229 1.149 -2.1% 7.0%
[17.25, 20.5) 1.251 1.211 1.279 3.4% -5.3%
[20.5, 23.5) 1.306 1.310 1.335 -0.3% -1.9%
[23.5, 26.75) 1.354 1.304 1.403 3.8% -7.0%
[26.75, 30.4) 1.472 1.447 1.583 1.8% -8.6%
[30.4, 35) 1.538 1.538 1.504 0.0% 2.2%
[35, 41.3) 1.727 1.633 1.790 5.8% -8.8%
[41.3, 50.6) 1.981 1.909 2.137 3.7% -10.7%
[50.6, 63) 2.233 2.237 2.437 -0.2% -8.2%
More than 63 years 2.200 2.202 2.181 -0.1% 1.0%
Total 1.600 1.576 1.647 1.5% -4.3%
* Range of deciles corresponds to household survey of 2016
Fuente: ENIGH

Estimation of the model

The model of consumption per person per household is Cobb-Douglas, multiplicative, and includes
the annotated variables and the hyperbolic discount factor based on household deciles of age. The
model is:
𝑑 𝛼 𝛼 𝛼 𝛼
𝐶𝑜𝑛𝑠𝑢𝑚𝑝𝑡𝑖𝑜𝑛ℎ = Π𝑒 −𝑘ℎ 𝐿𝑛(𝐴𝑔𝑒ℎ) 𝐼𝑛𝑑𝑖𝑐𝑎𝑡𝑜𝑟ℎ 1 𝑆𝑖𝑧𝑒ℎ 2 𝐸𝑥𝑝𝑒𝑛𝑑𝑖𝑡𝑢𝑟𝑒ℎ 3 𝐼𝑛𝑐𝑜𝑚𝑒ℎ 4

𝑒 𝛼5 𝑇𝑒𝑠𝑡ℎ 𝑒 𝛼6 𝐼𝑛𝑑𝑖𝑔𝑒𝑛𝑜𝑢𝑠ℎ 𝑒 𝛼7 𝑆𝑐ℎ𝑜𝑙𝑙𝑖𝑛𝑔ℎ 𝑈ℎ for h-households

The random variable Uh has a constant mean and variance, independence of errors is not required,
but independence of regressors and errors is required (Koenker, 2005).

In its logarithmic form, the model is expressed as:

ln(𝐶𝑜𝑛𝑠𝑢𝑚𝑝𝑡𝑖𝑜𝑛ℎ ) = Σ 𝑘ℎ𝑑 ln(𝐴𝑔𝑒ℎ𝑑 ) + 𝛼1 ln(𝐼𝑛𝑑𝑖𝑐𝑎𝑡𝑜𝑟ℎ ) + 𝛼2 ln(𝑆𝑖𝑧𝑒ℎ ) +


𝛼3 ln(𝐸𝑥𝑝𝑒𝑑𝑛𝑑𝑡𝑢𝑟𝑒ℎ ) + 𝛼4 ln(𝐼𝑛𝑐𝑜𝑚𝑒ℎ ) + 𝛼5 𝑇𝑒𝑠𝑡ℎ + 𝛼6 𝐼𝑛𝑑𝑖𝑔𝑒𝑛𝑜𝑢𝑠ℎ + 𝛼7 𝑆𝑐ℎ𝑜𝑜𝑙𝑖𝑛𝑔ℎ +
𝐿𝑛𝑈ℎ

The explanatory variables are exogenous, and the model can be estimated with ordinary least
squares method, but the estimation of the logarithmic model cannot be transformed, directly, to
the original model to obtain the factor and the discount rate. For this reason, quantile estimation is
used in order to make this transformation (Hao and Naiman, 2007, page 38).

Electronic copy available at: https://ssrn.com/abstract=3423028


The estimates of the models are presented in Annex 2. The estimation of the median is presented
to analyze the behavior of the discount of consumption, both in the sample of households and in 4
socioeconomic strata.

• The consumption model of soft drinks for 2016, 2014 and 2012. The last two years are estimated
to use as a reference the consumption behavior of soft drinks in other periods.
• The tobacco, alcohol and tortilla consumption model in 2016. The first two are products subject
to special taxes, as well as soft drinks. Tortillas are included because they are a basic
consumption product, which is not subject to this tax.

Consumption model of soft drinks for 2016, 2014 and 2012

The results of the estimation of discount of the consumption is presented in the following two
tables:

• Discount rate. This rate is higher in younger households and decreases with increasing age,
except in stratum 4 of households of high socioeconomic status whose estimated parameters
(K) are not significant or are statistically zero. For the first three strata and the total sample, the
result would indicate that young households can advance their consumption in greater
proportion than older households.
• Discount factor. This factor has similar values by decile of consumption and age of the
households taking as criterion the value of the coefficient of variation, CV. This coefficient shows
values lower than 7%, that is, there is not a wide variation of the discount factor for the deciles
of age of the households. The differences are present among the socioeconomic strata.

Table 5. Rate and discount factor for soft drink consumption by socioeconomic stratum, 2016

Discount rate Discount factor


Deciles of Total Stratum 1 Stratum 2 Stratum 3 Stratum 4 Total Stratum Stratum Stratum Stratum
age 1 2 3 4
1 3.3% 4.9% 4.0% 0.9% 0.0% 0.269 0.146 0.209 0.705 1.000
2 2.3% 3.5% 2.7% 0.7% 0.0% 0.274 0.144 0.219 0.689 1.000
3 1.9% 2.8% 2.3% 0.5% 0.0% 0.279 0.155 0.212 0.710 1.000
4 1.5% 2.4% 1.9% 0.3% 0.0% 0.286 0.147 0.218 0.789 1.000
5 1.3% 1.9% 1.5% 0.3% 0.0% 0.297 0.162 0.232 0.743 1.000
6 1.1% 1.6% 1.3% 0.2% 0.0% 0.301 0.158 0.229 0.788 1.000
7 0.9% 1.3% 1.1% 0.2% 0.0% 0.296 0.159 0.224 0.784 1.000
8 0.7% 1.1% 0.8% 0.2% 0.0% 0.296 0.157 0.231 0.749 1.000
9 0.6% 0.8% 0.7% 0.1% 0.0% 0.280 0.151 0.212 0.760 1.000
10 0.5% 0.7% 0.6% 0.2% 0.0% 0.232 0.129 0.176 0.618 1.000
Average 1.4% 2.1% 1.7% 0.4% 0.0% 0.281 0.151 0.216 0.734 --
s.d. 0.9% 1.3% 1.1% 0.3% 0.020 0.010 0.016 0.054 --
CV 64.1 63.2 63.1 73.5 7.3 6.4 7.6 7.4 --
Source: Own elaboration

Electronic copy available at: https://ssrn.com/abstract=3423028


The values shown by the rate and the discount factor indicate that consumers advance consumption
proportionally to the age of their members, with the effect that the consumption factor is similar
for different levels of consumption and different age levels of households, but differences are
observed between socioeconomic strata. Stratum 1, of households of low socioeconomic level
shows values that indicate greater impatience in consumption, following in that order the second
stratum of low-middle level and the third stratum of households with high-middle level. In the fourth
stratum, the estimated discount rate is zero, and the discount factor is 1, that is, these households
do not advance consumption.

Graph 1. Rate and discount factor for soft drink consumption by socioeconomic stratum, 2016

Discount rate Discount factor


6.0% 1.200

5.0% 1.000

4.0% 0.800

3.0% 0.600

2.0% 0.400

1.0% 0.200

0.0% -
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

Total Stratum 1 Total Stratum 1


Stratum 2 Stratum 3 Stratum 2 Stratum 3
Stratum 4 Stratum 4

Source: Own Elaboration

The estimated discount factor for 2012, 2014 and 2016 show the following behaviors.

• In stratum 4 of households with high socioeconomic status, the discount factor is unitary in the
three years analyzed, or its discount rates are zero. This would indicate that these consumers are
not impatient.

• The estimation of the 2014 model differs from the one obtained in 2012 and 2016. The estimated
results correspond to the decil 6 since in the median the parameters of the deciles of age are not
significant.

• Strata 1 and 2 whose socioeconomic conditions are low and medium-low, and includes 75% of
households in 2016, the discount factor is in the three years in a value range of 0.10 to 0.40. It means
that consumers are impatient and advance their consumption.

10

Electronic copy available at: https://ssrn.com/abstract=3423028


Graph 2. Discount factor for soft drink consumption by deciles of age and socioeconomic
stratum, 2012 and 2014
2012 2014*
1.200 1.200

1.000 1.000

0.800 0.800

0.600 0.600

0.400 0.400

0.200 0.200

- -
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

Total Stratum 1 Total Stratum 1


Stratum 2 Stratum 3 Stratum 2 Stratum 3
Stratum 4 Stratum 4

* Estimation of decil 6
Source: Own elaboration

Table 6 shows effects of variables that increases or decrease consumption (See Annex 2).

Table 6. Estimated parameters of the variables,


statistically significant values
Consumption increases by: Consumption decreases by:
• Income People older than 12 years that makes blood
• (total sample and stratum 1,2,3,4) test (total sample and stratum 2)
• Total expenditure per person per • Indicator of age
household, but different to the product (total sample and stratum 1,2)
analyzed.
• (total sample and stratum 1,2,3)
• Indigenous • Schooling
(total sample and stratum 1,2) (total sample and stratum 1,2,3,4)
• Size of household
(total sample and stratum 1,2,3,4)
Source: Own elaboration, See Annex 2

11

Electronic copy available at: https://ssrn.com/abstract=3423028


Consumption of tobacco, alcohol and tortillas in 2016

The discount of soft drink consumption is complemented by a consumption model for tobacco,
alcohol and tortillas. The same data set of the 2016 ENIGH is used and the median (decil 5) is
estimated.

The consumption of tobacco, alcohol and tortillas per household per person according to deciles of
age is presented in Table 7. It is not the purpose of this work to carry out a study of the consumption
patterns of these products, instead the data presented are meant to serve as a reference to analyze
the behavior of the discount rate of consumption of these products. The relevant feature is the CV
since the higher its value, the greater the heterogeneity between households grouped in each decile
of age. In this way, the consumption of tortillas is relatively more homogenous among the
households of each decile, unlike tobacco and alcohol. The highest CV values are highlighted in the
table, showing greater heterogeneity of households.

Table 7. Consumption of Tobacco, Alcohol and Tortillas per household per person by deciles of
age in households, 2016

Deciles of age Tobacco Alcohol Tortillas


Avg Households CV Avg Households CV Avg Households CV
< 17.25 0.011 323 162% 1.074 213 156% 0.983 5,625 67.2%
[17.25, 20.5) 0.016 301 202% 1.012 256 137% 1.061 5,871 69.8%
[20.5, 23.5) 0.017 347 199% 1.638 285 218% 1.100 5,813 68.5%
[23.5, 26.75) 0.016 351 175% 1.576 355 199% 1.112 6,120 67.1%
[26.75, 30.4) 0.019 431 197% 1.935 346 174% 1.178 5,955 68.5%
[30.4, 35) 0.018 396 146% 1.960 343 259% 1.227 5,701 71.1%
[35, 41.33) 0.028 487 161% 2.174 425 148% 1.334 6,119 77.8%
[41.33, 50.66) 0.041 507 200% 3.620 462 211% 1.555 5,811 79.6%
[50.66, 63) 0.048 501 161% 3.191 438 144% 1.815 5,645 78.4%
More than 63 0.053 339 160% 2.661 254 124% 2.017 5,596 80.9%
years
Total 0.028 3,983 200% 2.229 3,377 199% 1.334 58,256 81.3%
Average 0.027 2.084 1.338
s.d. 0.015 0.856 0.347
CV 56.9% 41.1% 25.9%
Source: ENIGH

The estimated consumption model is presented in Annex 2. The results of the estimation of the
median (decil 5) are reported for the complete sample of households in 2016. For tobacco
consumption, the discount factor takes values close to zero, which indicates that the consumption
is completely advanced, i.e it rather takes place at the present time instead of deferring or
postponing it. At the other extreme, there is the consumption of tortillas whose discount factor is
greater than 1. It can be interpreted as if consumption is postponed unlike tobacco. In the case of

12

Electronic copy available at: https://ssrn.com/abstract=3423028


soft drinks consumption, the discount factor indicates that its consumption is advanced in a greater
proportion than the consumption of alcohol in all the deciles of age.

Table 8. Rate and discount factor for consumption of tobacco, alcohol, soft drinks and
tortillas per household per person, 2016
Deciles of age Discount rate Discount factor
Tabaco Alcohol Soft drinks Tortilla Tabaco Alcohol Soft drinks Tortilla
1 11.9% 2.1% 3.3% -5.5% 0.0100 0.4245 0.2695 8.6
2 8.3% 1.4% 2.3% -4.1% 0.0097 0.4648 0.2736 9.4
3 6.7% 1.2% 1.9% -3.4% 0.0106 0.4584 0.2792 10.2
4 5.7% 0.9% 1.5% -2.9% 0.0102 0.4892 0.2865 10.5
5 4.7% 0.7% 1.3% -2.5% 0.0106 0.5060 0.2974 11.2
6 4.1% 0.7% 1.1% -2.2% 0.0099 0.4499 0.3005 11.5
7 3.3% 0.5% 0.9% -1.8% 0.0112 0.4910 0.2961 11.8
8 2.6% 0.4% 0.7% -1.5% 0.0110 0.5239 0.2962 12.5
9 1.8% 0.3% 0.6% -1.1% 0.0157 0.4998 0.2802 13.1
10 1.4% 0.3% 0.5% -0.8% 0.0142 0.3641 0.2319 12.1
Average 5.1% 0.9% 1.4% -2.6% 0.0113 0.4672 0.2811 11.1
s.d. 3.2% 0.6% 0.9% 1.4% 0.0020 0.0467 0.0204 1.4
CV 64.1 67.6 64.1 -55.7 17.9 10.0 7.3 12.8
Source: Own elaboration

Graph 3 shows the previous results for the discount of the consumption of the four products. For
tortillas only the unit value of the corresponding factor is shown.

Graph 3. Discount of consumption of Tobacco, Alcohol, Soft Drinks and Tortilla consumption by
deciles of age, 2016
Discount rate Discount factor
14.0% 1.2000
12.0%
10.0% 1.0000
8.0%
6.0% 0.8000
4.0%
0.6000
2.0%
0.0% 0.4000
-2.0%
-4.0% 0.2000
-6.0%
-8.0% -
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

Tobacco Alcohol Tobacco Alcohol


Soft drinks Tortilla Soft drinks Tortilla

Source: Own elaboration

13

Electronic copy available at: https://ssrn.com/abstract=3423028


The discount of the consumption of tobacco, alcohol and tortillas by socioeconomic stratum
presents the following results (see Graph 4).

• The discount factor for tobacco consumption is close to zero for the 4 strata of households.

• Consumption of alcohol is advanced to the present in households of low and low-middle strata,
strata 1 and 2, but in strata 3 and 4 with a factor equal to 1 they would not advance consumption.

• Regarding tortilla consumption, all household strata discount consumption with a factor greater
than or equal to 1, in the same order: strata 1, 2, 3 and 4. These households do not advance their
consumption and would rather postpone it.

Graph 4. Discount of consumption of tobacco, alcohol and tortilla consumption by socioeconomic


stratum, 2016
Tobacco Alcohol
1.000 1.000

0.750 0.800

0.600
0.500
0.400
0.250
0.200
- -
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

Total Stratum 1 Total Stratum 1


Stratum 2 Stratum 3 Stratum 2 Stratum 3
Stratum 4 Stratum 4

Tortilla
36.0
31.0
26.0
21.0
16.0
11.0
6.0
1.0
1 2 3 4 5 6 7 8 9 10

Total Stratum 1 Stratum 2


Stratum 3 Stratum 4

Source: Own elaboration

14

Electronic copy available at: https://ssrn.com/abstract=3423028


Conclusions

Excess consumption is analyzed in economics as a problem of impatient consumers that advance


future consumption of goods without considering consequences of their decision. This analysis is
carried out through estimates of a hyperbolic discount model of consumption for goods subject to
excise taxes such as tobacco, alcohol and soft drinks. Tortillas were included to compare the
behavior of consumption of this primary Mexican food. These models were estimated for the total
sample of households and for four socioeconomic strata of households defined in the ENIGH of
2016, 2014 and 2012.

The results indicate in general that the consumption is advanced in households of lower average
age and in the strata of households of low and low-middle socioeconomic level. The discount in the
consumption of tobacco and tortillas show opposite results, in the first case consumption is
advanced for all the strata of households, and in the second consumption is not advanced and it can
even be interpreted as if it is postponed to varying degrees according to the stratum of households.
On the other hand, the consumption of soft drinks shows a discount factor higher than alcohol
consumption, but the households of upper socioeconomic level do not advance consumption in any
of both goods, soft drinks and alcohol.

The estimation of the discount rate of consumption emphasizes the relevance of considering the
consumer behavior in the design of public policies oriented to controlling excess consumption.

References

Becker, G. S., Grossman, M., y Murphy, K. M. (1991). Rational Addiction and the Effect of
Price on Consumption. The American Economic Review, 81(2), 237-241.
Courtemanche, C., Heutel, G., y McAlvanah, P. (2014). Impatience, Incentives and Obesity.
The Economic Journal, 125(February), 1-31.
Hao, L., y Naiman, D. (2007). Quantile regression. California: Sage Publications.
Heal, G. (2007). Discounting: A Review of the Basic Economics. The University of Chicago
Law Review, 74, Winter(1), 59-77.
INSP-SSA. (2016). Encuesta Nacional de Salud y Nutrición de Medio Camino 2016. Informe
final de resultados. 31 Octubre. Consultado en México:
http://www.epidemiologia.salud.gob.mx/doctos/encuestas/resultados/ENSANUT.
pdf
Koenker, R. (2005). Quantile Regression (Vol. 38). New York: Cambridge University Press.
Reisch, L. A., y Zhao, M. I. N. (2017). Behavioural economics, consumer behaviour and
consumer policy: state of the art. Behavioural Public Policy, 1(02), 190-206.

15

Electronic copy available at: https://ssrn.com/abstract=3423028


Richards, T. J., y Hamilton, S. F. (2012). Obesity and Hyperbolic Discounting: An Experimental
Analysis. Journal of Agricultural and Resource Economics, 37(2), 181-198.
Rubinstein, A. (2003). "Economics and Psychology"?. The case of Hyperbolic Discounting
INTERNATIONAL ECONOMIC REVIEW, 44(4, November), 1207-1216.
SE. (2014). MODIFICACIÓN de la Norma Oficial Mexicana NOM-051-SCFI/SSA1-2010,
Especificaciones generales de etiquetado para alimentos y bebidas no alcohólicas
preenvasados-Información comercial y sanitaria, publicada el 5 de abril de 2010.:
Diario Oficial de la Federación, 14 ago 2014.
Storcksdieck, S., y Wills, J. M. (2012). Nutrition Labeling to Prevent Obesity: Reviewing the
Evidence from Europe. Curr Obes Rep, 1(3), 134-140.

Annex 1. Classification of expenditure and products of the household survey ENIGH

Variables ENIGH Variables of consumption model


Variable name Name code Variable name Name code
Expenditure monetary Gasto_tri Expenditure Gasto_tri + gasto_nm_tri
quarterly quarterly

Expenditure no Gas_nm_tri
monetary quarterly

Cost by total units Costo


Expenditure for total Gasto Expenditure for total Gasto + Costo
units units
Quantity of units Cantidad
Price (gasto + Costo) / Cantidad
Sources: Catalogue of products

Code Product
A004 Tortilla de maíz
A220 Refrescos de cola y sabores
A 223 Coñac y brandy
A224 Cerveza
A228 Aguamiel, pulque, Tlachique
A229 Aguardiente, alcohol de caña, charanda mezcal
A230 Ron añejo, blanco, con limón
A233 Tequila añejo, azul y blanco
A234 Vino de mesa blanco, rosado y tinto
A235 Vodka
A236 Whisky
A 239 Cigarros

16

Electronic copy available at: https://ssrn.com/abstract=3423028


Annex 2. Table 9. Estimation of the median of the soft drink consumption model per household per person, 2016
Total Stratum 1 Stratum 2 Stratum 3 Stratum 4
Variable Coeff. t-stat. Prob. Coeff. t-stat. Prob. Coeff. t-stat. Prob. Coeff. t-stat. Prob. Coef. t-stat Prob.
Test blood: % -0.049 -2.902 0.004 -0.049 -1.583 0.113 -0.051 -2.180 0.029 -0.026 -0.552 0.581 -0.074 -0.950 0.342
members household
Income 0.120 12.908 0.000 0.147 8.533 0.000 0.122 9.198 0.000 0.080 3.090 0.002 0.087 2.018 0.044
Total expenditure per 0.126 11.071 0.000 0.162 6.072 0.000 0.154 9.816 0.000 0.057 1.915 0.056 0.015 0.254 0.799
person per
household, no
includes product
analyzed
Age Indicator -0.084 -4.016 0.000 -0.074 -1.677 0.094 -0.096 -3.316 0.001 -0.020 -0.420 0.675 -0.034 -0.325 0.746
Indigenous: % 0.187 7.727 0.000 0.249 7.893 0.000 0.244 4.996 0.000 -0.086 -0.451 0.652 0.486 1.383 0.167
members speak an
indigenous language
Schooling -0.077 -18.526 0.000 -0.041 -3.294 0.001 -0.078 -11.823 0.000 -0.064 -6.261 0.000 -0.067 -3.929 0.000
Size: Household -0.589 -34.949 0.000 -0.633 -17.825 0.000 -0.584 -27.494 0.000 -0.512 -11.145 0.000 -0.607 -6.556 0.000
members
Age: decile 1 -0.489 -13.958 0.000 -0.716 -11.148 0.000 -0.583 -11.823 0.000 -0.130 -1.607 0.108 -0.018 -0.112 0.911
Age: decile 2 -0.441 -14.579 0.000 -0.660 -11.028 0.000 -0.517 -11.942 0.000 -0.127 -1.753 0.080 0.016 0.103 0.918
Age: decile 3 -0.413 -13.760 0.000 -0.603 -10.382 0.000 -0.503 -11.642 0.000 -0.111 -1.571 0.116 0.010 0.074 0.941
Age: decile 4 -0.388 -13.383 0.000 -0.594 -10.720 0.000 -0.473 -11.749 0.000 -0.073 -1.060 0.289 0.013 0.091 0.928
Age: decile 5 -0.362 -12.946 0.000 -0.544 -10.321 0.000 -0.436 -11.018 0.000 -0.089 -1.383 0.167 -0.011 -0.083 0.934
Age: decile 6 -0.345 -13.044 0.000 -0.530 -10.259 0.000 -0.424 -11.383 0.000 -0.068 -1.101 0.271 0.051 0.416 0.677
Age: decile 7 -0.335 -12.815 0.000 -0.506 -10.848 0.000 -0.412 -11.765 0.000 -0.067 -1.142 0.254 0.041 0.343 0.732
Age: decile 8 -0.319 -13.406 0.000 -0.485 -9.872 0.000 -0.384 -11.101 0.000 -0.076 -1.332 0.183 0.029 0.261 0.794
Age: decile 9 -0.316 -13.635 0.000 -0.469 -10.709 0.000 -0.385 -11.550 0.000 -0.068 -1.249 0.212 0.067 0.622 0.534
Age: decile 10 -0.343 -14.852 0.000 -0.480 -11.256 0.000 -0.407 -12.885 0.000 -0.113 -2.251 0.024 -0.041 -0.394 0.693
R-square 0.070 0.096 0.074 0.041 0.048
R-square Adjusted 0.070 0.094 0.073 0.039 0.043
Note: Estimates made with Eviews software

17

Electronic copy available at: https://ssrn.com/abstract=3423028


Annex2 Table 10. Estimation of the soft drink consumption model per household per person, 2014 and 2012 *
2014: decil 6 2012: Median
Coeff. t-stat Prob. Coeff. t-stat Prob.
Variable -0.083 -2.426 0.015 -0.033 -0.589 0.556
Test blood: % members household 0.145 8.140 0.000 0.117 4.146 0.000
Income 0.093 3.911 0.000 0.190 5.319 0.000
Total expenditure per person per household,
no includes product analyzed -0.035 -0.981 0.326 -0.171 -2.606 0.009
Age Indicator 0.165 3.126 0.002 0.429 2.797 0.005
Indigenous: % members speak an indigenous language -0.067 -8.199 0.000 -0.068 -3.986 0.000
Schooling -0.613 -17.504 0.000 -0.470 -8.079 0.000
Size: Household members -0.364 -5.626 0.000 -0.754 -8.157 0.000
Age: decile 1 -0.361 -5.965 0.000 -0.655 -7.322 0.000
Age: decile 2 -0.315 -5.741 0.000 -0.616 -7.294 0.000
Age: decile 3 -0.292 -5.587 0.000 -0.601 -7.892 0.000
Age: decile 4 -0.282 -5.291 0.000 -0.540 -6.996 0.000
Age: decile 5 -0.263 -5.207 0.000 -0.545 -7.117 0.000
Age: decile 6 -0.253 -5.137 0.000 -0.488 -7.081 0.000
Age: decile 7 -0.239 -5.254 0.000 -0.479 -7.052 0.000
Age: decile 8 -0.228 -5.251 0.000 -0.451 -6.861 0.000
Age: decile 9 -0.245 -5.810 0.000 -0.475 -7.358 0.000
Age: decile 10 0.080 0.082
R-square 0.079 0.080
R-square Adjusted 0.145 0.163
Note: Estimates made with Eviews software

18

Electronic copy available at: https://ssrn.com/abstract=3423028


Annex 2. Table 11. Estimation of the median for the model of consumption of tobacco, alcohol and tortillas per household per
person, 2016
Tobacco Alcohol Tortillas
Variable Coeff. t-stat Prob. Coeff. t-stat Prob. Coeff. t-stat Prob.
Test blood: % members household 0.002 0.026 0.980 -0.153 -1.899 0.058 0.042 3.314 0.001
Income 0.063 1.811 0.070 0.145 3.042 0.002 -0.012 -1.673 0.094
Total expenditure per household per person, 0.062 1.817 0.069 0.080 1.278 0.201 -0.166 -19.975 0.000
no includes good analyzed
Age Indicator -0.068 -0.983 0.326 -0.011 -0.122 0.903 -0.065 -5.057 0.000
Indigenous: % members speak an indigenous language -0.389 -1.846 0.065 -0.032 -0.280 0.780 -0.020 -0.887 0.375
Schooling -0.036 -2.453 0.014 -0.079 -4.130 0.000 -0.089 -27.930 0.000
Size: Household members -1.065 -17.237 0.000 -1.188 -13.905 0.000 -0.348 -27.376 0.000
Age: decile 1 -1.722 -14.443 0.000 -0.318 -2.092 0.037 0.799 35.243 0.000
Age: decile 2 -1.576 -15.182 0.000 -0.260 -1.890 0.059 0.765 35.988 0.000
Age: decile 3 -1.475 -15.132 0.000 -0.253 -1.830 0.067 0.752 36.824 0.000
Age: decile 4 -1.424 -15.000 0.000 -0.222 -1.817 0.069 0.730 38.232 0.000
Age: decile 5 -1.357 -14.956 0.000 -0.203 -1.658 0.097 0.723 38.629 0.000
Age: decile 6 -1.328 -14.400 0.000 -0.230 -1.970 0.049 0.703 38.400 0.000
Age: decile 7 -1.236 -14.255 0.000 -0.196 -1.717 0.086 0.680 38.273 0.000
Age: decile 8 -1.180 -14.088 0.000 -0.169 -1.604 0.109 0.662 38.039 0.000
Age: decile 9 -1.030 -12.487 0.000 -0.172 -1.788 0.074 0.638 40.514 0.000
Age: decile 10 -1.001 -12.424 0.000 -0.238 -2.472 0.014 0.584 39.264 0.000
R-square 0.196 0.181 0.077
R-square Adjusted 0.193 0.177 0.077
Note: Estimates made with Eviews software

19

Electronic copy available at: https://ssrn.com/abstract=3423028

You might also like