Regional infrastructure effects on the life satisfaction of rural and urban residents. A case study for Ecuador

In developing countries, many households lack basic household services and the provision of utility infrastructure is uneven across regions, leaving lagged regions behind. This lack of infrastructure in speci�c places can affect the welfare of its inhabitants. This paper aims to measure the in�uence of household basic services and sub-national infrastructure activities on the individual subjective wellbeing in Ecuador. To determine how important the geographical context is for individual welfare, a hierarchical ordered logistic multilevel model is conducted. The results show that the individual heterogeneity is explained in 7% by the variation across cantons. There are 52 cantons which are above the average life satisfaction and 43 cantons below it. Findings regarding infrastructure evidence that sub-national utility projects and road infrastructure have a positive signi�cant effect on the life satisfaction of rural residents whereas it is not signi�cant for urban residents, indicating diminishing marginal utility of urban people. Once a satiation point is achieved, marginal utility increases are lower. As for household services, access to internet has a higher positive welfare effect than access to sewerage and access to water via pipelines.


Introduction
The conceptualization of welfare has been a necessary and hard task for societies, with a particular urgency in developing countries towards a better understanding about the conditions under which poverty persists.In theoretical grounds, wellbeing can be conceived as an interplay of three main components: i. the resources that a person has, ii.what they can achieve with those resources, and iii. the future goals and aspirations that they have (Gough & McGregor, 2007).Given the complexity of this phenomenon, there is not a universally accepted measure of welfare (Ferrata & Nisticò, 2015).Several approaches have been undertaken, ranging from the widely criticized income measure to subjective measures such as happiness and life satisfaction.While happiness is a self-assessment of pleasant effects of a particular aspect, life satisfaction takes into account several domains of life (Diener et al. 1999).To study welfare, objective circumstances, subjective perceptions, which are shaped by a given context (social, economic, political, geographical, etc.), need to be considered (Gough & McGregor, 2007 Knight & Gunatilaka, 2010).With respect to objective aspects, the satisfaction of basic needs is a key element to analyze.Basic needs must be satis ed rst, and then, higher capabilities allow people to meet new challenges and enhance their wellbeing.However, in many countries, especially in developing ones, basic needs are hardly satis ed.These lacks compromise the achievement of wellbeing of their population.Particularly, in Latin America, 26% of households struggle to satisfy their basic housing needs such as sanitation, water and space (Interamerican Development Bank, 2012).So far, the literature considers that income is a key element to determine the satisfaction regarding basic needs for food, water and shelter (Florida et al, 2013).However, the infrastructure for household basic services is an external factor that do not depends on the individuals' income but on governments.This study aims to determine the wellbeing effect of the access to basic housing services and the marginal wellbeing increase of sub-national infrastructure such as road construction and utility projects in Ecuador.It also aims to distinguish the marginal effect of such infrastructure for urban and rural residents.
The effect of the basic housing services on wellbeing is expected to be higher in developing countries, especially in rural areas where a satiation point regarding these needs is not reached yet (Stevenson & Wolfers, 2013).At low levels of consumption of housing services, additional units make individuals happier.Once the satiation point is reached, generally in urban areas, there would be small increases in wellbeing (Diener & Seligman, 2004).To analyze the levels of satiation, the indicator of unsatis ed basic needs (UBN) is useful.According to this indicator, rural-urban differences are clearly evident in Ecuador with an index of poverty in terms of UBN of 61.6% in rural areas and 21.4% in urban ones in 2019.This shows that in rural areas, compared to urban areas, a higher proportion of the population have households that lack for adequate conditions and basic services.Moreover, the basic service coverage in Ecuador varies across types of cities.While metropolis have 92% and 78% of households with water served by pipelines and sewerage, respectively, small cities record lower coverage in those services with 78% and 54%, respectively (MIDUVI, 2015).Such geographical disparities in terms of basic infrastructure dotation might in uence on the wellbeing of individuals in those areas.As long as people have unmet needs, they can hardly have good levels of well-being.Given this, infrastructure projects to provide basic household services in places with low provision are key elements to improve the life quality of inhabitants.In Ecuador, infrastructure activities are present in 175 cantons out of 221.In 2015, 138 cantons recorded less than 10 infrastructure projects in spite of the fact that their population lack basic household services.The resources are mainly devoted to big cities whereas small cities with low coverage of basic infrastructure have less funding which leave them even more lagged.
By focusing on the case of Ecuador, this study seeks to determine the individual's welfare effect of both the basic service infrastructure of the household and the infrastructure of the canton where he/she lives by distinguishing the rural and urban context of the individual.This approach allows to analyze the welfare effect considering the existent and direct access and consumption of infrastructure, on the one hand, and the additional and indirect infrastructure of the location where individuals live, on the other hand.To do so, a multilevel ordinal logistic model is conducted, using a module on subjective wellbeing of the ENEMDU survey for 2014 and 2015 of Ecuador.Results evidence that the heterogeneity of life satisfaction is explained by differences across cantons in 7%.The availability of sewerage in a household is important for the individual welfare but access to internet has a higher positive effect.Regarding the contextual infrastructure, the effect of the construction of utility projects is higher for rural residents than for urban ones, re ecting the diminishing marginal utility principle.The consumption of additional units of utility infrastructure makes individuals of urban places less happy than the consumption of previous ones.As the satiation point in this type of infrastructure is not reached in rural areas, their residents report increasing utilities for marginal increases.While an increase from $0 to $8192 of productivity in utility projects makes rural residents more likely to reporting a high level of life satisfaction by 2.2 p.p, the same productivity increase makes urban residents more likely to report a high level of life satisfaction only by 1.7 p.p.As for Other infrastructure activities such as industrial facilities, it positively affects urban residents and negatively affects rural ones.
The remainder of this paper is structured as follows.Section 2 presents a brief literature review on the role of infrastructure in the wellbeing of population.Section 3 describes data and econometric methodology.Section 5 discusses the results and Section 6 concludes.

Literature Review
The infrastructure, besides of being a growth determinant, is an important wellbeing driver since it facilitates human capital development, increases the standard of living and in turn, improves the distribution of wealth when is also focused on deprived areas (Popova, 2017;Schwartz, 2011).The infrastructure investment can improve the quality of life for the population without having a direct impact on their income (Haughwout, 2001).For instance, transport infrastructure such as roads are necessary to carry food between rural and urban places, workers to workplaces, children and young people to education institutions, etc.Therefore, infrastructure is a means to do daily activities that have an important welfare role.For instance, highway infrastructure allows travelling and trading at lower costs, which have a positive impact on welfare.Such impact varies across highway segments (Allen & Arkolakis, 2019).Additionally, access to publicly provided goods and services is an important source of welfare (Boltvinik, 2003).The wellbeing of people, especially in rural areas, would improve when they have access to household sewerage connections, to drinking and irrigation water, to roads, among other utility facilities.In this sense, infrastructure is a powerful tool to reduce regional disparities making some lagged territories more attractive for people and industrial settlement (Popova, 2013).From the cost-bene t analysis of water supply and sewerage projects, their bene ts are multidimensional in elds such as economics, health, amenity and ecology (Vucijak & Ceric, 2010).In line with the health edge, Soares (2009) obtained that mortality has reduced in Latin America due to provision of treated water and sewerage services.More literature about utility projects focuses on the e ciency differences between public and private provision of these services (Galiani et al, 2005;Smith et al, 2018).Regarding rail infrastructure, Knaap and Oosterhaven (2011) show that it has an effect of redistribution of employment towards the invested area.
Especially, in developing countries, the existing poverty in lagged areas may reduce with the provision of infrastructure since it allows satisfying some basic needs.Besides, infrastructure in rural areas can have an impact on the reduction of peri-urban poverty of cities, generated by urban-rural migrations.Thus, infrastructure can kill two birds with one stone.
It is expected that the rural residents in less advantageous conditions in terms of construction of roads and utility projects have a higher marginal utility as compared to urban residents.For those people with a low provision of infrastructure, the effect of having one more unit increases their utility to a larger extent than for those people with high provision of infrastructure.Sapkota (2018) focusing on rural Nepal observed that perceived impacts of access to infrastructure were higher in more remote areas.Once physical infrastructure is su cient, generally achieved in urban areas, additional bene ts are marginal (Stone & Strutt, 2009).On the contrary, rural population have double bene ts of infrastructure on the current standard of living and future additional gains (Popova, 2013).To the best of our knowledge, the differences of the effect of infrastructure on life satisfaction across urban and rural residents have not been studied from a holistic perspective, only from an isolated rural one with Sapkota (2018).Helliwell, 2003) that analyze the effects of aggregate variables such as national income, inequality and education.The infrastructure, however, has been neglected despite that it is a key element that provides basic conditions for the functioning of a society.By using novel data of rms dedicated to infrastructure, the speci c effect of this item is assessed on urban and rural residents.

Estimation strategy
To identify the determinants of self-reported life satisfaction, ordered logit models and multilevel models have been used.The former does not deal with the hierarchical structure of the data where individuals are nested within cantons (Deleeuw and Meijer, 2008).To deal with the resulting intra-canton correlation, a twolevel variance components ordinal logistic regression is employed: where LS i is the life satisfaction of individual i,X i is a vector of individual variables and X canton ( i ) is a vector of cantonal variables,u canton ( i ) represents the cantonal random effects and e i is the error term.
The individual variables are socio-economic and demographic characteristics, typically used in the literature (Blanch ower & Oswald, 2004).The sociodemographic variables are sex with female as the reference category, the age, the age squared, the ethnic auto-identi cation (mestizo, afro-descendent, indigenous, montubio, white and other) with mestizo as the category of reference, the civil status (married, single, free union, widowed, divorced, separated) with single as the ference category.The socio-economic variables are the personal income in logarithm, whether the person worked the last week, the level of education (no education, primary education, secondary education and tertiary education) with no education as a reference category, whether the person accounts for medical insurance, whether the individual do sport activities, whether the person is homeowner and whether the person has been victim of a crime.In addition, variables associated to the satisfaction with other aspects of life are included.They are the satisfaction level with social relations, with the government and with the family.The lowest levels of satisfaction are the reference categories.These satisfaction variables are recategorized in the same way as the dependent variable.Contextual variables are given by household variables and cantonal variables.Household variables are related to the type of access route, sewerage, pipelines and urban/rural area.The cantonal variables are the population density, the gross valued added and the labor productivity of activities [1] related to basic infrastructure in year t.The basic infrastructure refers to i. construction of roads and railways [2], ii.construction of utility projects [3], iii.construction of other civil engineering projects [4].
The use of the labor productivity of infrastructure in the current year allows us to determine the additional welfare bene t of infrastructure utilities that people obtain given an existent level of infrastructure.In big cities, the additional bene t of infrastructure would be marginal since these cities already account for a su cient or almost su cient level of infrastructure meanwhile in small cities the additional bene t of infrastructure would be high as those places account for low levels of infrastructure.
To determine the validity of the multilevel model over the single-level ordered logit, the Likelihood ratio test is employed.The results show that the statistic chibar2(01) = 1451.49with a p-value of 0.000, which indicates that the second-level cantonal random effects are signi cant, so a multilevel model is appropriate.

Dependent Variable
The measure of subjective wellbeing is based on the self-reported life satisfaction obtained from the perception module of the ENEMDU survey.Life satisfaction considers several domains such as work, family, leisure, health, etc. (Diener et al. 1999).According to this survey, this variable is reported into a scale from 0 to 10 where 0 represents the lowest level of life satisfaction and 10 the highest level.Due to parsimony reasons, the 0-10 scale is sub-divided into three categories so that each category has at least 10% of individuals.According to this, the rst 5 levels are re-categorized as Low-Life-Satisfaction-Level (LLS), the 6th and 7th levels are re-categorized as Medium-Life-Satisfaction-Level (MLS), and nally, the 8th to 10th levels are re-categorized as High-Life-Satisfaction-Level (HLS) as shown in Table 1.As shown in Table 1, people tend to report high levels of satisfaction with more than 87.7% reporting at least a level of 6 in the 0-10 scale and only 1.6% of people report lower levels of life satisfaction than 3.The average of life satisfaction is 2 on a three-point scale.

Variable of interest: Infrastructure
The labor productivity of activities related to basic infrastructure in year t is obtained from the Directory of rms.The basic infrastructure refers to i. construction of roads and railways [5], ii.construction of utility projects[6], iii.construction of other civil engineering projects [7].The number of establishments dedicated to infrastructure can also be used.However, the number hides how big the infrastructure projects are, limiting the analysis of the effective impact of infrastructure on wellbeing.
In Ecuador, infrastructure activities are present in 175 cantons out of 221.More speci cally, road infrastructure activities, utility projects and Other infrastructure are present in 143 cantons, 103 cantons and 43 cantons, respectively.In 2015, 138 cantons recorded less than 10 infrastructure projects in spite of the fact that their population lack for basic household services.For instance, the cantons General Antonio Elizalde and Colimes recorded only one infrastructure project despite that 51% and 80% of their populations, respectively, lack sewerage systems.Cantons with more infrastructure projects were Guayaquil (1100), Quito (712), Cuenca (155) and Portoviejo (147).The rst three cities, however, have high coverage of sewerage of 91% in Quito, 85% in Cuenca and 82% in Guayaquil.The resources are mainly devoted to big cities whereas small cities with low coverage of basic infrastructure have less funding which leave them even more lagged.
In 2014, the labor productivity of basic infrastructure projects was USD$59,709 and in 2015, it decreased to USD$42,333.In particular, the average labor productivity in Other infrastructure activities was lower than the average productivity in road infrastructure and utility infrastructure in both years.As for road infrastructure, in 2015, 46 cantons record higher productivity levels than the average (USD$ 48,917) and 95 cantons record lower productivities.Regarding the utility projects, only 37 cantons record higher productivities than the average (USD$ 43,209) and 64 cantons record lower productivities.With respect to Other infrastructure, only 14 cantons record higher labor productivities than the average (USD$ 34,872) and 27 cantons record lower productivities.
[1] The number of establishments dedicated to infrastructure can also be used.However, the number hides how big the infrastructure projects are, limiting the analysis of the effective impact of infrastructure on wellbeing.
[2] The activity of construction of roads and railways (F4210) includes construction of motorways, streets, roads and other vehicular and pedestrian ways, surface work on streets, roads and highways, bridges or tunnels, construction of bridges, including those for elevated highways, construction of tunnels, construction of railways and subways and construction of air eld runways.
[3] This activity of utility projects includes the construction of distribution lines and related buildings and structures that are integral part of these systems.These civil engineering constructions are for longdistance pipelines, communication and power lines, urban pipelines, urban communication and power lines; ancillary urban works, water main and line construction, irrigation systems (canals), reservoirs.This class also includes construction of sewer systems, including repair, sewage disposal plants, pumping stations and power plants.[4] This activity (F4290) includes the construction of industrial facilities such as re neries and chemical plants, construction of waterways, harbor and river works, pleasure ports, dams and dykes, dreading of waterways, outdoor sport facilities.This class also includes land subdivision with land improvement such as adding of roads and utility infrastructure.
[5] The activity of construction of roads and railways (F4210) includes construction of motorways, streets, roads and other vehicular and pedestrian ways, surface work on streets, roads and highways, bridges or tunnels, construction of bridges, including those for elevated highways, construction of tunnels, construction of railways and subways and construction of air eld runways.
[6] This activity of utility projects includes the construction of distribution lines and related buildings and structures that are integral part of these systems.These civil engineering constructions are for longdistance pipelines, communication and power lines, urban pipelines, urban communication and power lines; ancillary urban works, water main and line construction, irrigation systems (canals), reservoirs.This class also includes construction of sewer systems, including repair, sewage disposal plants, pumping stations and power plants.[7] This activity (F4290) includes the construction of industrial facilities such as re neries and chemical plants, construction of waterways, harbor and river works, pleasure ports, dams and dykes, dreading of waterways, outdoor sport facilities.This class also includes land subdivision with land improvement such as adding of roads and utility infrastructure.

Discussion
Table 2 shows the results of the simple ordered logistic multilevel model in column ( 1), the multilevel model including individual characteristics in column ( 2), the multilevel model including infrastructure variables at the cantonal level in column ( 3) and the multilevel model including interaction terms between urban and infrastructure variables in column (4).For dichotomous variables, the reference groups are in parentheses.
Only the signi cance of the coe cients are interpreted from these results.Tables 3 and 4 record the interpretable marginal effects using average adjusted predictions of the individual and household variables, and the infrastructure variables, respectively.Standar errors are in parentheses and signi cance is given by: * p < 0.05, ** p < 0.01, *** p < 0.001 The results of the empty multilevel ordinal logistic model show that the heterogeneity of the self-reported life satisfaction is explained by between-canton variation in 7% (0.252/(0.252+3.29)),as in Pontarollo et al. (2019).This means that individuals of the same canton tend to report a similar level of life satisfaction which could be explained by the availability of the same services and amenities that a given canton provides.While life satisfaction is a speci c individual measure, regional disparities matter to explain such heterogeneity.The canton-speci c impact on life satisfaction is larger in the empty model than in the model including individual (4.2%) and in the model including contextual variables (4.3%).It implies that these variables explain 3% of the heterogeneity.The remaining 4%, which is not explained, might be due to unreported infrastructure related to urban amenities such as hospitals, parks, entertainment centers, etc. or a possible sorting of people that are highly satis ed with their lives living in speci c cities, generally, big ones.By means of this methodology, a ranking of the predicted cantonal random effects ( ) with their con dence intervals, is obtained using the empty model.Cantons are, then, ranked according to their level of life satisfaction.There are 95 cantons whose con dence interval do not overlap the zero line, indicating that life satisfaction level is signi cantly different from the average probability.There are 52 cantons which are above the average and 43 cantons below it.Among the cantons with the highest rank (see Table 3 and Figure 1) and therefore the highest level of life satisfaction, we have San Cristobal, Isabela and Santa Cruz, which are all the cantons of the Galapagos region.Surprisingly, the metropolitan cities of Quito, Guayaquil and Cuenca are not ranked in the rst positions, instead they are in the 108 th , 119 th and 77 th positions, respectively.This indicates that big cities, in spite of providing many services and amenities, could cause discomfort on the population due to congestion.Among the cities with the lowest levels of life satisfaction, we have three cities of the rainforest region of the country (Gonzalo Pizarro, Cascales and Putumayo), four coastal cities (Paján, Puebloviejo, Vinces and Muisne) and three cities of the highland region (Nabón, Penipe and Bolívar).
Regarding the cut points of the parsimonious "empty" model ( 1), the log-odds of reporting a medium (MLS) and a high level of life satisfaction (HLS) is 1.121 which corresponds to a probability of 75.4% (exp(1.121)/(1+exp(1.121)).Differently put, the probability of reporting a low level of life satisfaction (LLS) is 24.6% (1-0.754).The second cut point is estimated at 1.132 which indicates that the probability of reporting a HLS is 24.4% (exp(-1.132)/(1+exp(-1.132)).The distribution of the life satisfaction records 25% of the population in LLS and 25% of the population in HLS.The results on the socio-economic and demographic variables are in line with the existent literature.As the education level increases the likelihood of reporting a HLS increases from 12% (secondary education) to 28% (tertiary education) with respect to people that do not have any level of formal education.People that invest in human capital can have better opportunities that allow them to reach higher levels of welfare (Helliwell, 2003;Jiang et al., 2012).As the literature for developing countries suggest, income is an important source of welfare (Veenhoven, 1991;Clark et al, 2008).In the same line, employed individuals, compared to the unemployed people, have 13% higher likelihood of reporting a HLS (Clark & Oswald, 1994;Graham, 2008c;Ravallion & Lokshin, 2001).Furthermore, people who play sports are more likely to report a HLS than people who do not play sports.As a recreational activity, sports play an important role for welfare (Ferrer-i-Carbonell & Gowdy, 2007).Contrary to Clark et al. (2008), our result shows that the effect of income is four percentage points lower than the effect of leisure activities.Another condition to have an adequate level of life satisfaction is the tenancy of insurance as the likelihood of reporting a HLS 7% higher than people who do not have any insurance.
As the wellbeing involves surrounding environment of individuals, variables regarding the household conditions are analyzed.When a household accounts for sewerage and the water that receives is via pipelines, the individual living there is more prone to report HLS with respect to an individual whose household lack these services.The signi cant effects of sewerage and pipelines reduce when including the dichotomous variable urban, which was expected, since households located in urban areas are more likely to have those services.By contrast, whether the main access route to the household is paved or not does not matter for the wellbeing of the individual.
Regarding the cantonal contextual variables, while the gross value added and the population density are not signi cant, the infrastructure variables have a signi cant effect on the individual life satisfaction.Speci cally, road infrastructure and utility projects increase the level of welfare of individuals.To determine whether this effect changes depending on the place where people live, urban or rural, an interaction term with the urban variable is included.The associated coe cient of the interaction term indicates the effect of the different types of infrastructure for urban residents and the coe cient of the simple infrastructure variables indicates their effect for rural residents.The results show that the simple coe cients of construction of roads and utility projects infrastructure is signi cant, which shows that these types of infrastructure increase the life satisfaction of rural residents.For urban residents, the construction of roads and utility projects do not affect their life satisfaction whereas for rural residents, these works increase.This result appeals for a satiation point for urban residents as they account for enough basic infrastructure, additional works do not increase their wellbeing (Stevenson & Wolfers, 2013).As for rural residents, such satiation point is not achieved with existent basic infrastructure so that additional works do increase their wellbeing.Only the effect of other works of infrastructure (industrial facilities and outdoor sport facilities) is signi cantly positive for urban residents which indicates that such satiation point regarding these facilities is not yet satiated.
To interpret the magnitudes of the coe cients of the multilevel ordered logistic model, the marginal effects using Average Adjusted Predictions are analyzed in Table 4 for continuous variables and Table 5 for dichotomous variables.It is worth noting that the marginal effects of LLS, MLS and HLS sum zero because as some options become more likely, others become less likely.
As the effect of infrastructure varies by the place where a person lives, the marginal effects of area (urban, rural) at different levels of infrastructure are calculated.The marginal effects of changes from 0 to the mean productivity in logarithm and incremental units are presented in Table 4.For instance, an increase from 0 to 14 ($0 to $16 384) of productivity in logarithm of road infrastructure activities leads to a higher probability of reporting HLS by 1.55 p.p for rural residents and by 1.54 p.p for urban residents.If the productivity in roads doubles from $16 384 to $32 768, the increment in the probability to report HLS is only 0.1 p.p for both rural and urban residents.The differences in welfare increase between rural and urban residents for an increment in this type of infrastructure are not signi cant.By contrast, remarkable differences between urban and rural residents exist when the productivity of utility projects increase.While an increase from $0 to $8192 of productivity in utility projects makes rural residents more likely to reporting HLS by 2.2 p.p, the same productivity increase makes urban residents more likely to report HLS only by 1.7 p.p. Likewise, the increase of the probability of reporting HLS for rural residents by doubling the productivity of utility infrastructure is higher for rural residents (0.176%) than for urban ones (0.13%).These results might be explained by a decreasing marginal utility as stated by Easterlin (1974), indicating that rural residents with low provision of infrastructure have a large increase in their life satisfaction for increases in utility infrastructure.In addition, utility projects in these areas not only generate a direct effect of improved water quality but also long-term and indirect effects in terms of improved quality of life, job creation, increased property value and avoided environmental damage (Vucijak & Ceric, 2010).In complement, life satisfaction increases from utility infrastructure is not signi cant for urban residents as their basic needs might be less unsatis ed than those of rural residents.In fact, poverty by UBN in the urban area was 21.4% in 2019 whereas in the rural area, it was 61.6%.By contrast, an increase of the productivity of other infrastructure activities from $0 to $2048 decreases the probability of reporting HLS for rural residents by 0.23 p.p. and increases that of urban residents by 0.9 p.p.In this regard, other infrastructure activities include industrial facilities such as re neries and chemical plants, construction of waterways, harbor and river works, pleasure ports, dams and dykes, dreading of waterways, outdoor sport facilities.This class also includes land subdivision with land improvement such as adding of roads and utility infrastructure.In general, those facilities are built in peripherical areas that affect rural residents due to pollution and damage to the landscape.Thus, urban residents, located far away from those facilities, are not negatively affected.They are even more satis ed with the construction of infrastructure for pleasure for their use outside the city.
These ndings shed more light on the debate about the differences in life satisfaction across rural and urban places stated by Easterlin et al. (2011).Taking into account speci c aspects such as infrastructure clari es why differences in life satisfaction across rural and urban places exist.5 presents the predicted probabilities to report low, medium and high life satisfaction for a change from 0 to 1 of dichotomous variables and 1 unit/standard deviation increase of continuous variables.
Regarding the household characteristics, the internet access constraints reduce the life satisfaction of people.Those individuals living in households without access to internet are 1.7 p.p more likely to report LLS with respect to those with access to internet.This service is even more important for welfare than the water via pipelines and sewerage services.This result is especially interesting since it shows that people value the access to internet more than basic services such as sewerage and tapping water.One reason could be the fact that internet has no substitutes meanwhile sewerage and tapping water have been replaced by other forms.For instance, according to the WSP (2011), on the basis of four big cities in Latin America (Guatemala city-Guatemala, Managua-Nicaragua, Santa Cruz-Bolivia and Tegucigalpa-Honduras), shows that people in peri-urban places of these cities that do not account for sewerage, use, instead, in-situ sanitation systems such as latrines, septic tanks and toilets without drainage.Regarding the tapping water, people without this service use water wells or rainwater collectors.This reality is also present in Ecuador.It should be noted that not all people value internet access.According to our statistics, 70% do not has access to it due to different reasons.The main reason is the lack of economic resources (61%).Other reasons are ignorance on how to use internet (13%), no need or interest (20%) and no coverage (6%).Thus, people without internet access may also lack of other services and urban amenities, which are out of the scope of this study.The access to water via pipelines increases the probability of reporting HLS in only 0.7 p.p. Altogether, these aspects re ect how urbanized are the places where households are.Therefore, the location of the individuals' households plays an important role in determining their welfare.
As for individual characteristics, compared to women, men are less likely to report LLS in 1.1 percentage points, 0.1 p.p more likely to report MLS and 1.2 p.p more likely to report HLS.Likewise, employed people, in comparison with unemployed people, have higher probability of reporting HLS in 1.9 percentage points.As people get more education, the probability to report HLS increases from 1.7 p.p (secondary education) to 4.2 p.p (higher education) with respect to people with lower levels of education.People that invest in human capital can improve their standard of living (Helliwell, 2003).When individuals are home-owners, their probability of reporting HLS is 1.5 p.p higher than those who pay a rent for housing.Recreation activities are important elements to have a high level of life satisfaction.Thus, people that practice sports are more likely to report HLS by 1.8 p.p than those who do not practice any sport.People that have medical insurance are less likely to report a LLS in 1 p.p.

Conclusions
The subjective wellbeing does not only depend on individual characteristics but also on the context.Thus, the location of an individual matters for individuals' welfare.In Ecuador, the heterogeneity of the selfreported life satisfaction is explained by between-canton variation in 7%.This means that individuals of the same canton tend to report a similar level of life satisfaction which could be explained by the availability of the same services and amenities that a given canton provides.By means of the multilevel methodology, cantons are ranked according to their speci c effect of welfare.There are 52 cantons with signi cantly higher levels of life satisfaction than the average and 43 cantons with signi cantly lower levels of life satisfaction.Interestingly, the main cities of Ecuador are not ranked in the rst positions.
Infrastructure proves to have a signi cant positive effect on life satisfaction.Such effect varies depending on the rural-urban perspective.On the one hand, compared to urban residents, rural residents obtain higher life satisfaction effects of construction of utility projects, showing a diminishing marginal utility.As the satiation point in this type of infrastructure is not reached in rural areas, their residents report increasing utilities for marginal increases.On the other hand, urban residents would report higher life satisfaction when the productivity of other infrastructure activities such as industrial facilities increases whereas rural residents are negatively affected by these facilities.Household variables are also important determinants of individual welfare.The access to internet is the prevailing household characteristic that increases the probability to report high levels of life satisfaction, more than the access to water via pipelines and sewerage.This result indicates that people value internet access more than basic services because no substitutes exist for this service whereas substitutes have been found for the lack of sewerage and tapping water.In addition, it is worth noting that the effect of internet access might not be the same across people since some of them have a higher preference for internet than others.This aspect, which is out of our scope, can be further explored in future research, considering the internet use and need of internet.
Our ndings have two policy implications.First, to reduce the unequal regional infrastructure provision so that all the population be satis ed.And second, to ensure that bene ts of big civil engineering works such as roads, pipelines, irrigation, sewer systems, among others, reach not only urban but also rural residents of cantons.
While important ndings are obtained, this investigation has some limitations.To analyze the role of infrastructure, the productivity of infrastructure activities of the last 5 years is used rather than the stock of infrastructure.Since the focus of this study was the civil engineering infrastructure, healthcare and associated infrastructure was not considered.However, it de nitively constitutes a future line of research.
In addition, further research could be conducted to determine the welfare effect of infrastructure for different types of population in terms of age and occupation.
Cantonal random effects Map of life satisfaction (cantonal random effects) by cantons ; Berry & Okulicz-Kozaryn, 2009; Easterlin et al. 2011; DataThe database combines two types of data: individual data from the National Survey of Employment, Unemployment and Underemployment of 2014 and 2015 of Ecuador (ENEMDU, Spanish acronym); and aggregate data from the Directory of rms and the Regional Accounts from the Central Bank of Ecuador.

Table 1
Distribution of the life satisfaction Source: National Survey of Employment, Unemployment and Underemployment, 2014-2015.

Table 3
Ranking of cantonal random effects

Table 4
Marginal effects of area at different levels of productivity by types of infrastructure.