Bad Jobs Versus Good Jobs: Does It Matter for Life and Job Satisfaction?

Abstract

Using panel data from the National Employment, Unemployment and Underemployment Survey (ENEMDU) for Ecuador, we analyze the outcomes of life and job satisfaction whether moving from bad job to a good job—and vice versa—on life and job satisfaction. In contrast with bad jobs, good jobs are characterized by being employed in the formal sector, with social security registration, and earning at least the minimum wage. Using a conditional logit estimate, we found that workers who move from a bad to a good job increase job satisfaction by 9.5%, whereas when the transition is from a good to a bad job, job satisfaction decreases by 8.5%; in terms of gender, the effect is greater for men than women. Finally, we did not find any significant effect of job transitions on life satisfaction.

Introduction

Past research for European countries has shown a negative effect of becoming unemployed on life satisfaction, based on the idea that unemployment is an adverse situation because people experience a loss of income and reduced subjective well-being. In terms of happiness, Layard (2004) argues that any job is better than unemployment; in this sense, Grün et al. (2010) found that even low-quality jobs are better off than unemployment.

However, this dichotomy between unemployment and life satisfaction is not that straightforward in the case of developing countries, like Ecuador (the focus of our paper). Specifically, Dewan and Peek (2007) argue that in developing countries, the employment vs. unemployment dichotomy is not a good proxy for the variations in the composition of good and bad jobs in the labor market. These authors mention two reasons: first, employment covers a wide range of wages, working hours, and types of contracts and does not differentiate the formal from the informal sector. Second, poverty influences unemployment decisions since poor people must take bad jobs. Likewise, ILO (2019) argues that being in employment is far from having a decent standard of life; it forces workers to accept any job, even in the informal sector with low remuneration and practically neither labor rights nor social protection.

In fact, in Ecuador, unemployment is a minor concern for three reasons. First, unemployment represents only 3.2% of the labor force in 2013–2014 (the years under evaluation). In this sense, Canelas (2019) argues that in Ecuador, those who have an informal job are relatively low educated with low wages, which supports the idea of Dewan and Peek (2007) that these workers have no better option than having a poor-quality job (informally). Second, there is no unemployment insurance1 in Ecuador; therefore, unemployment duration is relatively low (3.3 months2). Third, as shown in Table 1, in Ecuador, people in unemployment experience higher life satisfaction, are more educated, and have better health status than those employed in bad jobs.3 Hence, it could be argued that those in unemployment are the ones best able to avoid bad jobs.

Table 1.

Descriptive statistics

Panel A. Individual characteristics				Panel B. Job characteristics			Panel C. Job characteristics
	Good job	Bad job	Unemployment		Good job	Bad job		Good job	Bad job
Life satisfaction	7.94	7.47	7.61	Working hours	44.47	37.91	Job satisfaction	3.78	3.45
	(1.44)	(1.59)	(1.61)		(15.19)	(8.91)		(0.56)	(0.83)
Female	0.37	0.36	0.49	More than one job	0.03	0.06	Dissatisfaction with low income	0.97	0.85
	(0.48)	(0.48)	(0.5)		(0.18)	(0.24)		(0.35)	(0.18)
Age	40.43	45.08	31.26	Permanent Job Contract	0.69	0.13	Dissatisfaction with working hours	0.34	0.17
	(12.35)	(15.95)	(13.1)		(0.45)	(0.33)		(0.47)	(0.37)
Primary	0.17	0.57	0.19	Temporal job contract	0.32	0.28	Dissatisfaction with working schedule	0.23	0.09
	(0.38)	(0.49)	(1.25)		(0.46)	(0.45)		(0.42)	(0.28)
Secondary	0.36	0.33	0.53	Vacations	0.86	0.06	Dissatisfaction due working overload	0.28	0.09
	(0.47)	(0.47)	(0.49)		(0.34)	(0.25)		(0.44)	(0.28)
University	0.46	0.08	0.28	Health insurance	0.07	0.01	Dissatisfaction with working instability	0.31	0.61
	(0.50)	(0.28)	(0.44)		(0.26)	(0.06)		(0.46)	(0.48)
Marital status	0.66	0.64	0.34	Training	0.40	0.02	Dissatisfaction with working environment	0.12	0.07
	(0.47)	(0.48)	(0.47)		(0.49)	(0.14)		(0.32)	(0.26)
Years working	10.01	15.95		Working establishment size	0.6	0.02	Dissatisfaction with working at street	0.05	0.1
	(10.25)	(15.92)			(0.48)	(0.16)		(0.21)	(0.29)
Racial minority	0.08	0.17	0.12	RUC	1	0.17	Dissatisfaction with working accidents	0.04	0.10
	(0.27)	(0.38)	(0.32)	Employed in the formal sector	1	0.19		(0.20)	(0.31)
Health status	7.34	6.63	6.72			(0.31)	Dissatisfaction with working activities	0.16	0.15
	(1.84)	(2.03)	(2.01)	Social security	1	0.07		(0.35)	(0.37)
Income	795.25	278.49				(0.25)	Dissatisfaction with possibilities to progress	0.59	0.78
	(1226.19)	(353.84)						(0.49)	(0.41)
Income perception	0.49	0.26	0.27				Dissatisfaction with bad labor relations	0.11	0.03
	(0.56)	(0.31)	(0.27)					(0.32)	(0.14)
Coast region	0.54	0.51	0.53
	(0.49)	(0.49)	(0.49)
Highland region	0.42	0.45	0.44
	(0.49)	(0.49)	(0.49)
Amazon region	0.03	0.02	0.03
	(0.18)	(0.16)	(0.16)

Source: authors’ own calculations

this table shows the average of individual characteristics and job characteristics for people in good jobs, bad jobs, and unemployment using ENEMDU panel 2013–2014

In contrast, the main concern in Ecuador is the workers employed in bad jobs. That is why we focus this research paper on dichotomy good jobs vs. bad jobs.

Regarding the labor market regulation, Latin America exhibits high levels of informality and relatively rigid regulation (David et al., 2020). Thus, an employment contract is a key instrument for accessing a good job, with legal and social security benefits (Ramos et al., 2015; Roethlisberger & Weller, 2011). In 2008, Ecuador’s labor market regulation changed severely since the subcontracting of complementary services (outsourcing workers) and hiring workers by the hour were abolished.4 For some employed workers, this new regulation implied accessing higher pay, earning equal or higher than the minimum wage, and accessing social security benefits. From the social point of view, this was considered a gain since people might be changing precarious work to decent work (the winners). On the other hand, this regulation might have harmed workers employed in the informal sector with poor-quality job conditions, since for firms, it is more costly to offer a formal job (the losers).

In this framework, we suggest that people in bad jobs experienced pecuniary cost (loss of income) and non-pecuniary cost (reduced life satisfaction and job satisfaction) similar to unemployment costs studied in previous literature for European countries. To test this hypothesis for Ecuador, we estimate whether job transitions into a different quality of jobs (based on objective measures) correlates to job satisfaction (subjective measure). To the best of our knowledge, this is the first study in Latin America on this topic. We selected Ecuador because in this country, earning the minimum wage and having working benefits seem to be an attainment for people employed in bad jobs. Besides, the official labor market survey in Ecuador has complete information on person’s wellbeing in the panel data structure, which allows us to deal with the potential econometric problem of endogeneity.

In brief, this paper presents research results of transitions between good jobs and bad jobs in Ecuador using a two-period panel 2013–2014 using a conditional logit model. The results show that job changes, from bad to good jobs (and vice versa), impact on job satisfaction, but not on life satisfaction. To the best of our knowledge, this is the first study on the topic for Latin America that brings the consequence of job transitions of subjective well-being. In this line, the results of this paper are important from the policy perspective in terms of fostering quality of jobs.

The rest of the paper is structured as follows. First, we present a brief literature review, followed by a description of the data and methodology employed. The results are then presented, and conclusions are drawn.

Literature Review

Employment and Well-being

The literature on worker’s well-being has paid particular attention to changes in life satisfaction between unemployed and employed people. When a person becomes unemployed, his welfare falls for two reasons: (i) the loss of income and (ii) the loss of self-respect and sense of significance, called the psychic loss or non-pecuniary cost (loss of social relationship, identity in society, and individual self-esteem). Winkelmann and Winkelmann (1998) found that the non-pecuniary effect of unemployment is much larger than the economic loss of income. Unemployed people have much lower levels of mental well-being than those in work (Carroll, 2007; Clark & Oswald, 1994; Wilson & Walker, 1993). Besides, Clark (2003) found that unemployed well-being is correlated with the reference unemployment group, which determines the level of adherence to the social norm (regional, partner, or household level). These authors likewise, Marcenaro-Gutierrez et al. (2010), argue that unemployment always hurts, but it hurts less when there are more unemployed people around. In the same way, Stutzer and Lalive (2004) found that unemployed people experienced lower levels of life satisfaction because of the social norm to work; the larger their reduction in life satisfaction is, the stronger the norm is.

A question that goes beyond is what happens when unemployed people find a job? Gielen and Van Ours (2014) suggest a scarring effect, in the sense that unemployed people, who experienced a drop in life satisfaction, do not completely recover life satisfaction when finding a job. These authors argue that unhappiness does not affect the probability of finding a job; rather, it increases only search job effort. Furthermore, Chandola and Zhang (2018) found that re-employed people into poor quality jobs were associated with higher levels of stress biomarkers in comparison with those who remain unemployed. However, Grün et al. (2010) found that people who take up a new job, even in bad jobs, report higher life satisfaction compared to those who remain unemployed.

An important related question that arises is whether people choose to be in certain labor status. Clark and Oswald (1994) suggest that because of the negative effect of unemployment on happiness, unemployment is predominantly involuntary. In search models, workers could become dissatisfied with their jobs and then become voluntarily unemployed, so the association between unemployment and life satisfaction is questionable (Kassenboehmer & Haisken-DeNew, 2009). From the methodological point of view, this discussion is crucial because if unemployment is involuntary, it can be treated as an exogenous variable, whereas if it is voluntary, it should be treated as an endogenous variable. Kassenboehmer and Haisken-DeNew (2009) distinguish between voluntary and involuntary unemployment using questions on the reason for job termination.5 Similarly, Chadi and Hetschko (2018) found that switching workplace affects job satisfaction positively only in the short term. To address endogeneity, the authors use the plant closure as an exogenous trigger of job changing. This is a relevant issue we deal with in the “Results” section.

Quality of Employment

Definition of Quality of Jobs

The definition of quality of jobs started decades ago from the concepts of decent work (ILO 1999) and quality of employment (Presidency Conclusions Lisbon European Council 2000). There have been several attempts to construct a definition of quality of job, based on objective measures and subjective measures.

The concept of quality of employment is a complex definition; it might depend on the perspective taken, i.e., whether it assesses the quality of employment from the societal, the corporate, or the individual point of view (Vermeylen, 2005 as cited in United Nations Economic Commission for Europe, 2015). Burchell et al. (2014) explore the debate around the quality of employment and decent work through the years and criticize the definition of decent work because it is vague and all-encompassing. There are various attempts to measure decent work (ILO, 2008; Anker et al., 2003; Bescond et al., 2003; Bonnet et al., 2003) and quality of employment (European Commission, 2008). Both concepts suffer the same problems such as (i) difficulty to measure across countries, (ii) lacking information on relevant variables/indicators, and (iii) not a standardized methodology with insights from diverse academic disciplines (Burchell et al., 2014).

Subjective Measures of Quality of Jobs

One of the most common approaches to measure the quality of employment is job satisfaction that relates to hedonic values of jobs and expresses the worker’s experienced preference revealing over opportunities or mentally experienced alternatives (Lévy-Garboua & Montmarquette, 2004). Clark and Oswald (1996) found that job satisfaction is negatively related to comparison levels of earnings. Hence, people’s reported job satisfaction could be treated as a proxy of utility from work. Layard (2005) agrees that it might approach the quality of a job through job satisfaction surveys made to workers. Green (2010) found that job satisfaction predicts subsequent quitting better than mental health assessments. Lévy-Garboua and Montmarquette (2004) found that job satisfaction correlates with wage gaps in the past and present and recommends the use of job satisfaction in econometric studies of job mobility.

Nonetheless, the use of job satisfaction as a proxy for worker well-being has two important critics. First, people employed in bad-quality jobs report being satisfied with their job even when the job quality measured objectively is poor (Brown et al., 2007 and Brown et al., 2012). This occurs because the responses in satisfaction questions may be influenced by adaptation and norms; workers may adapt to poor-quality jobs, which makes them satisfied with bad jobs (Brown et al., 2012). So, there is skepticism about whether job satisfaction decreases when objective conditions worsen (Eichhorst et al., 2015).

Lora (2008) describes that the quality of a job in Latin America is disappointing in terms of wages, working hours, productivity growth, informality, and social security affiliation, but job satisfaction is high. The author argues that one plausible explanation is that workers may have a misleading perception of their reality, probably because of low expectations. Despite their situation, poor people give a positive answer (aspiration paradox). De Bustillo Llorente and Fernández Macias (2005) conclude—for Spain—that job satisfaction cannot measure the quality of job. First, in case of being unsatisfied with their job, people will try to change it until finding a good one that fits their expectations. Second, unsatisfied workers get used to their actual job, lower their expectations, and declare a positive level of job satisfaction. Chadi and Hetschko (2021) describe adaptation as the honeymoon-hangover phenomenon; they found that job switchers report high level of satisfaction, but they adapted quickly returning to their initial level of satisfaction. Besides adaptation, another concern is preference drift, which means that higher wages increase job satisfaction, but this effect fades away over time (Groot and Van den Brink 1999). This happens because the preferences and aspirations also change, and hence, the higher wage-job satisfaction effect disappears.

Objective Measures of Quality of Jobs

Alternatively, the quality of work can be assessed using objective characteristics that capture what is important for workers (job values) and why they get into the labor market. Brown et al. (2007) define job quality in terms of pay, creative content of work, the interest of work itself, relations with colleagues, position within organizational and class hierarchy, influence and discretion over work, skill, and effort levels. Similarly, the ISSP Research Group (2017) uses six dimensions: pay, hours of work, future prospects, hard work, job content, and interpersonal relationships. OECD (2014) defines a good job in terms of earnings quality, labor market security, and quality of working environment. For Latin America, Roethlisberger and Weller (2011) propose twenty-one indicators to measure employment quality, divided into six groups: income source, stability, social protection, organization, social integration, and personal development. However, the authors recognize that their proposal suffers from several data restrictions, limiting its implementation. Nevertheless, they suggest that low-quality jobs with low productivity are concentrated mainly on the informal sector.

Similar to subjective measures, objective measures are not exempt from critics. Budd and Spencer (2015) mention two important critics: first, there is no universal agreement over what variables define job quality and how these variables should weigh it to construct a multidimensional measure. Besides, the selected variables may depend on data availability on surveys and the researcher’s discretion. Second, job objective measures are focused mainly on job characteristics (job-centric) and ignore what work means in people’s lives (worker-centric). A recent criticism of the objective and the subjective approach can be found in Nikolova and Cnossen (2020), who argue that job quality does not consider work as a source of meaning, called eudaimonic dimension. As subjective well-being science pays attention to eudaimonia as a measure of a person’s well-being, something similar should be incorporated to evaluate job quality.

Therefore, we believe that there is still a debate in the literature on what a job quality represents for a worker. Moreover, it is unlikely that a definition of job quality could be extrapolated from one country to another. For instance, Grün et al. (2010)—for West Germany—define a bad job in terms of (i) low wages, (ii) type of contract (permanent or fixed-term), and (iii) job satisfaction. For these authors, if the wage is two-thirds below the median wage of a full-time employee, it is considered low paid. In contrast, this criterion would hardly work for Ecuador, considering the income distribution is more asymmetrical than in Germany; hence, the low-paid threshold would be underestimated. Besides, the type of contract seems less relevant in Ecuador. Instead, what should matter is whether the worker has a contract or not because it determines labor stability and access to social security benefits.

Data and Methodology

Data

Data comes from the National Employment, Unemployment and Underemployment Survey (ENEMDU), which is a quarterly labor demand survey that covers urban and rural sectors for Ecuador. The survey uses a “two-two-two” rotating-panel design, which means a panel is visited in two consecutive quarters; it makes no visits in the next two quarters and in the next two quarters raises the second visit. In consequence, we constructed a two-period panel starting from December 2013 to December 2014. During this period, no significant macroeconomic or political shocks had occurred in Ecuador. For some years, ENEMDU incorporated subjective well-being modules, including a life satisfaction question; this is the reason why the 2013–2014 period was selected as it is the only in which this module coincides. Likewise, panel 2009–2010 coincides, but the matching is only possible for the urban sector (we will use it somehow as robustness check).

We define a good job if the person meets all of the three following conditions: (i) earns equal or more than the minimum wage, (ii) is employed in the formal sector, and (iii) is affiliated to social security (formal employment). Alternatively, if either good job conditions are not met, it is considered a bad job.

To classify formal and informal jobs, we use the guideline of the Molina et al. (2015), which is based on the ILO (2013).6 It considers a person in the formal sector if the working establishment uses a taxpayer identification number (RUC) or if the working establishment has equal or more than 100 employees, whereas a worker is considered informal if the working establishment has less than 100 employees and does not use RUC. On the other hand, there are several criteria to classify informal employment across countries such as labor contract, provident funds, social security registration, health, and other social services by virtue of their job (ILO 2013). In Ecuador, there is not a formal definition of formal employment; then, we use the criteria of social security registration since social security registration allows workers multiple benefits such as retirement pension, health insurance, risk insurance related to the job, and access to loans. It is crucial to notice that being employed in the formal sector (establishment perspective) is different from having formal employment (worker perspective). These terms are not interchangeable, but they complement each other for our good job definition.

Table 1 shows comparative statistics between good and bad jobs. Panel A shows that people in good jobs are younger, are less likely to be a racial minority, are more educated, have a better health status, receive a higher wage, and are more satisfied with their income compared to people in bad jobs. Panel B shows the difference in terms of job characteristics; by construction, people in good jobs have considerably more access to formal features such as permanent job contract (labor stability), social insurance, 13th salary, 14th salary, training, and vacations. Panel C indicates that people in good jobs experience overall higher job satisfaction and are less unsatisfied because of low income, labor stability, possibilities to progress, and working on the street. Conversely, people in good jobs are more unsatisfied because of work schedules, working overload, and working environment, which is reasonable since they work more hours, must follow a schedule, and have to deal with workmates. Table 2 shows the differences, in terms of characteristics, for those who move between good jobs and bad jobs before transitions (in 2013). We can observe that people who move from a bad job to a good job (G = 1) are more educated, with higher income and work more hours, compared to those who remain in bad jobs (G = 0). Similarly, people who move from a good job to a bad job (B = 1) are different compared to those who remain in good jobs (B = 0). In other words, those in better conditions are best able to find a good job, whereas those in worse conditions are the most likely to lose a good job.

Table 2.

Mean test difference before transition (in year 2013)

	Bad to good			Good to bad
	G = 0	G = 1	p-value	B = 0	B = 1	p-value
Job satisfaction	3.42	3.46	0	3.42	3.73	0
Life satisfaction	7.46	7.63	0	8.06	7.85	0
Female	0.37	0.27	0	0.37	0.34	0.87
Age	48.41	36.22	0	40.50	40.88	0.71
Primary	0.70	0.32	0	0.15	0.27	0
Secondary	0.25	0.43	0	0.33	0.47	0
University	0.03	0.24	0	0.50	0.25	0
Work hours	36.27	40.09	0	44.81	44.22	0.66
Health status	6.44	6.92	0	7.47	7.15	0
Labor income	151.15	201.18	0	802.29	520.51	0
Income perception	0.23	0.33	0	0.57	0.37	0
Marital status	0.61	0.57	0.28	0.68	0.62	0
Racial minority	0.22	0.13	0	0.07	0.11	0
Coast region	0.42	0.52	0.03	0.40	0.47	0.14
Highland region	0.51	0.47	0.04	0.54	0.55	0.77
Amazon region	0.02	0.03	0.72	0.03	0.05	0.23

this table shows the average of a person’s characteristics when employed in good and bad jobs in 2013. The p-value corresponds to test for differences between means. The null hypothesis is that the difference between means equals zero

Econometric Framework

In this paper, we consider that having a bad job could be considered involuntary or exogenous for people in adverse conditions, who do not have an outside option because unemployment is costly, and they may be forced to take any job, even a poor-quality job or an informal job (Canelas, 2019; Dewan & Peek, 2007). Nonetheless, this is hardly the case for all workers; there are also better-educated workers and entrepreneurial workers in informal jobs (poor quality jobs). On the contrary, having a good job could be considered endogenous since on average, more educated people, not racial minorities, with better health status are employed in good jobs (see Table 1). In either case, considering that this is an observational study, we do not make any causal inference of job transition on the outcome variables.

Regarding life satisfaction, it is measured through the question “In a scale of 1 to 10, meaning 1 completely unhappy and 10 completely happy, how do you feel taking into account all aspects of your life?” Job satisfaction is measured through the question “How do you feel at work” in a scale of 1 to 4, meaning 1 completely dissatisfied, 2 somewhat dissatisfied, 3 somewhat satisfied, and 4 satisfied. There is a lengthy discussion about the use of ordinal data regarding life satisfaction and job satisfaction as dependent variable for econometric analysis. Ferrer-i-Carbonell and Fritjers (2004) argue that the assumption of cardinality or ordinality in the dependent variable is less important. Instead, what we should care about is taking into account time-invariant unobserved factors. In this sense, it is important to take individual fixed effects. Consequently, we use a conditional logit model allowing for fixed effects, a similar approach to Booth and Van Ours (2008), Kassenboehmer and Haisken-DeNew (2009), and Grün et al. (2010). Each dependent variable (y₁ = life satisfaction, y₂ = job satisfaction) is transformed into a binary variable that equals one if the individual score is above the average and zero otherwise.

Table 2 performs a mean difference test for covariates used in the econometric model’s pre-transition in the year 2013. The results indicate that people who move from a bad job in 2013 to a good job in 2014 were different from people that remain in bad jobs in terms of job satisfaction, education, health, income, income perception, and racial minority. A similar pattern occurs for people who move from a good job to a bad job. To solve for selection bias, we use the differences in difference estimator to allow for unobserved heterogeneity (the unobserved difference in mean outcomes between people who find/lose a good job). The estimation equation is

$${Prob(y}_{\mathit{it}}=1)=\alpha +\beta T_{i1}\ast t+\delta T_{i1}+\rho t+{\epsilon }_{\mathit{it}}$$ 1

where $T$ indicates a transition variable from good (bad) to bad (good) job, $t$ is the time dummy, and the interaction of $t$ and $T$ is called the differences in difference estimator. When allowing for individual fixed effects, ${\theta }_i$ the equation takes the following form:

$${Prob(y}_{\mathit{it}}=1)=\alpha +{\theta }_i+\beta T_{i1}\ast t+\rho t+{\epsilon }_{\mathit{it}}$$ 2

For estimates in transitions from a bad job to a good job, $\beta =\pi$, and in transitions from a good job to a bad job, $\beta =\kappa$.

Results

Table 3 reports estimates for a (i) conditional logit model and (ii) conditional logit model allowing for fixed effects for both dependent variables and including control variables.

Table 3.

The effect of job transitions on life satisfaction and job satisfaction

	Bad job to good job				Good job to bad job
	Life satisfaction		Job satisfaction		Life satisfaction		Job satisfaction
	Logit	Conditional logit	Logit	Conditional logit	Logit	Conditional logit	Logit	Conditional logit
	1	2	3	4	5	6	7	8
t	− 0.154***	− 0.154**	− 0.168***	− 0.128*	− 0.200***	− 0.225***	0.098*	0.110
	(− 4.40)	(− 3.18)	(− 4.47)	(− 2.51)	(− 5.22)	(− 4.34)	(2.11)	(1.80)
T	0.134**		0.348***		-0.023		− 0.197***
	(2.64)		(5.57)		(-0.50)		(− 3.70)
(T*t) π	0.117	0.101	0.314***	0.333***
	(− 1.83)	(− 1.35)	(4.02)	(3.80)
(T*t) κ					− 0.034	− 0.084	− 0.289***	− 0.299***
					(0.56)	(1.16)	(− 4.18)	(− 3.77)
Hours	− 0.001	− 0.002	0.013***	0.008**	− 0.001	− 0.002	0.013***	0.007**
	(− 1.25)	(− 1.11)	(10.08)	(3.26)	(− 1.03)	(− 1.10)	(9.81)	(3.14)
Female	− 0.001		0.336***		− 0.003		0.348***
	(− 0.03)		(8.44)		(− 0.11)		(8.72)
Age	0.008	− 0.112	− 0.06***	− 0.119	0.009	− 0.114	− 0.06***	− 0.118
	(1.57)	(− 1.66)	(− 8.87)	(− 1.66)	(1.64)	(− 1.69)	(− 8.86)	(− 1.65)
Age squared	− 0.000	0.000	0.001***	0.000	-0.000	0.001	0.00***	0.001
	(− 0.91)	(1.90)	(11.14)	(1.84)	(-1.00)	(1.93)	(11.12)	(1.83)
Secondary	0.200***	0.172	0.0973*	0.180	0.209***	0.175	0.116**	0.169
	(5.29)	(1.45)	(2.23)	(1.45)	(5.55)	(1.47)	(2.65)	(1.36)
University	0.507***	0.530**	0.311***	0.355	0.537***	0.532**	0.415***	0.363
	(10.15)	(2.61)	(5.03)	(1.57)	(11.11)	(2.63)	(6.81)	(1.60)
Health status	0.373***	0.389***	0.082***	0.066***	0.373***	0.389***	0.082***	0.066***
	(39.64)	(23.97)	(9.18)	(4.31)	(39.66)	(23.98)	(9.18)	(4.31)
Marital status	0.116***	0.084	0.068	0.127	0.118***	0.081	0.061	0.128
	(3.44)	(0.52)	(1.70)	(0.74)	(3.48)	(0.50)	(1.53)	(0.75)
Racial minority	− 0.102*	0.086	− 0.061	− 0.124	− 0.104*	0.084	− 0.057	− 0.119
	(− 2.33)	(0.83)	(− 1.22)	(− 1.10)	(− 2.38)	(0.81)	(− 1.15)	(− 1.05)
Income perception	0.243***	0.143*	0.842***	0.589***	0.245***	0.145*	0.835***	0.588***
	(7.40)	(2.43)	(20.61)	(8.82)	(7.46)	(2.46)	(20.44)	(8.81)
Income	0.000	0.000	0.000***	0.000***	0.000	0.000	0.000***	0.000***
	(1.48)	(1.46)	(10.88)	(4.13)	(1.75)	(1.48)	(10.74)	(3.69)
Region 2	0.062*		− 0.071		0.062*		− 0.074*
	(1.98)		(− 1.90)		(1.97)		(− 1.97)
Region 3	0.097		− 0.037		0.096		− 0.031
	(1.09)		(− 0.35)		(1.08)		(− 0.29)
Industry occupation dummies	Yes	No	Yes	No	Yes	No	Yes	No
Fixed effects	No	Yes	No	Yes	No	Yes	No	Yes
Intercept	− 1.627**		0.812		− 1.608**		1.015
	(− 2.80)		(1.23)		(− 2.76)		(1.53)
N	23,009	7934	23,007	5798	23,009	7934	23,007	5798

t statistics in parentheses; *p < 0.05; **p < 0.01; ***p < 0.001

excluded categories

^aPrimary or less education

^bRegion 1 (highland region)

Hausman test: $H_o:{\beta }^{\mathrm{Logit}}-{\beta }^{\mathrm{Conditional}\mathrm{Logit}}=0$

Bad job to good job (life satisfaction)

Prob > ${\lambda }^2=0.00$

Bad job to good job (job satisfaction)

${\lambda }^2=63.24$

Prob > ${\lambda }^2=0.00$

Good job to bad job (life satisfaction)

Prob > ${\lambda }^2=0.00$

Bad job to bad job (job satisfaction)

Prob > ${\lambda }^2=0.00$

When performing the Hausman test7 to determine whether the difference between fixed and random effect coefficients is systematic, we reject the null hypothesis, in favor of the alternative hypothesis that the fixed effects model is appropriate. Panel A shows that moving from a bad job to a good job increases job satisfaction, whereas the coefficient for life satisfaction is not statistically significant. On the other hand, Panel B indicates that moving from a good job to a bad job reduces job satisfaction as before the result is not statistically significant on life satisfaction. In this sense, it is interesting to note that subjective income perception (Caporale et al., 2009; Clark et al., 2008; Ferrer-i-Carbonell, 2005) and health status (Helliwell, 2003; Ljunge, 2016; Mizobuchi, 2017) explain life satisfaction rather than absolute income or job transition.

In conditional logit models, computing marginal effects with fixed effects panel data is problematic since it calculates the probability of a positive outcome assuming that fixed effects are zero. As an alternative following Kitazawa (2012), we compute average semi-(elasticities) in Table 4. When a person moves from a bad job to a good job, job satisfaction increases by 9.5% (column 1) whereas when moving from a good job to a bad job, job satisfaction decreases by 8.5% (column 2). To validate the results, we estimate the same model, but with 2009–2010 panel data that contain information only of the urban sector.8 The effect of finding a good job is 8.6% (column 3) and the effect of losing a good job is − 11.9% (column 4), which are pretty close to the results in the 2013–2014 panel.

Table 4.

Average (semi-) elasticities for labor transitions on job satisfaction

	2013–2014 panel		2009–2010 panel
	Bad job to good job	Good job to bad job	Bad job to good job	Good job to bad job
	(1)	(2)	(3)	(4)
t	− 0.0364*	0.0313	− 0.0528	0.00316
	(− 2.51)	(1.80)	(− 1.06)	(0.06)
(T*t) π	0.0950***		0.0857*	− 0.119***
	(3.79)		(2.49)	(− 3.71)
(T*t) κ		− 0.0851***
		(− 3.76)
Hours	0.00229**	0.00220**	0.00206*	0.00205*
	(3.26)	(3.14)	(2.42)	(2.40)
Age	− 0.0339	− .0337	− 0.0307	− 0.0285
	(− 1.66)	(− 1.65)	(− 1.19)	(− 1.10)
Age squared	0.000358	0.000356	0.000320	0.000301
	(1.84)	(1.83)	(1.18)	(1.11)
Secondary	0.0514	0.0482	0.0396	0.0394
	(1.45)	(1.36)	(0.81)	(0.80)
University	0.101	0.103	0.0768	0.0812
	(1.57)	(1.60)	(0.92)	(0.97)
Health status	0.0187***	0.0187***	0.0340***	0.0334***
	(4.31)	(4.31)	(6.17)	(6.04)
Marital status	0.0361	0.0365	0.0231	0.0167
	(0.74)	(0.75)	(0.36)	(0.26)
Racial minority	− 0.0352	− 0.0338	− 0.116*	− 0.125*
	(− 1.10)	(− 1.05)	(− 2.30)	(− 2.46)
Income perception	0.168***	0.167***
	(8.78)	(8.77)
Income	0.000147***	0.000130***	0.000379***	0.000346***
	(4.13)	(3.69)	(5.48)	(5.03)
Industry occupation dummies	No	No	No	No
Fixed effects	Yes	Yes	Yes	Yes
N	7518	7518	3590	3590

t statistics in parentheses; *p < 0.05; **p < 0.01; ***p < 0.001

excluded categories

^aPrimary or less education

Region variables are omitted because of fixed effects

Some evidence suggests a job satisfaction differential in favor of women. Clark (1997) examines the paradox that women, despite having worse jobs than men, report higher job satisfaction. The author concludes that the only plausible explanation is that women have lower expectations since their jobs have been worse in the past, ruling out sample selection bias and differences between types of jobs by gender. In a similar line, Sloane and Williams (2000) also find higher job satisfaction in women than men. However, the differential does not seem to be innate; instead, it reflects self-selection into jobs. Other studies (Bender et al., 2005; Gazioglu & Tansel, 2006) also confirm that women report higher satisfaction than men do. Sousa-Poza and Sousa-Poza (2000) find that the gender job satisfaction paradox only holds in eight of twenty countries, although they use cross-sectional data, limiting the notion of causality since one cannot observe job satisfaction during employment transitions.

Table 5 shows the average semi-(elasticities) differentiating by gender for job satisfaction. The effect of job satisfaction for men is greater than for women when finding a good job (10.9% vs. 6.1%; column 2 vs. column 1) or losing a good job (− 9.5% vs. − 6.8%; column 4 vs. column 3). As fewer women than men work (64% vs. 36%), one plausible explanation is selection bias since dissatisfied women might be out of the labor force. A solution for this problem would be to use a Heckman model to solve for selection bias as Clark (1997). However, data about the number of children, husband wage, or care economy within the household is not available. Alternatively, another explanation for future research is despite finding a better job, gender discrimination in favor of men in terms of salary, positions, benefits, etc. would make women value less the gain of finding a better job. Our research is not comparable with the above-mentioned studies (Clark, 1997; Sloane & Williams, 2000, etc.) since we focus on the gain/loss in job satisfaction by gender when moving between bad and good jobs.

Table 5.

Average (semi-) elasticities for labor transitions on job satisfaction by gender

	Bad job to good job		Good job to bad job
	Female	Male	Female	Male
	(1)	(2)	(3)	(4)
t	− 0.0170	− 0.0442*	0.0399	0.0287
	(− 0.64)	(− 2.51)	(1.16)	(1.40)
(T*t) π	0.0612	0.109***
	(1.38)	(3.53)
(T*t) κ			− 0.0683	− 0.0950***
			(− 1.66)	(− 3.43)
Hours	0.000551	0.00291***	0.000522	0.00277**
	(0.43)	(3.40)	(0.41)	(3.23)
Age	0.00374	− 0.0427	0.00381	− 0.0431
	(0.10)	(− 1.73)	(0.10)	(− 1.75)
Age squared	− 0.000115	0.000482*	− 0.000117	0.000487*
	(− 0.30)	(2.06)	(− 0.31)	(2.08)
Secondary	0.0692	0.0487	0.0687	0.0450
	(0.93)	(1.18)	(0.92)	(1.08)
University	0.116	0.0951	0.117	0.0965
	(1.04)	(1.18)	(1.05)	(1.19)
Health status	0.0215**	0.0177***	0.0217**	0.0178***
	(2.79)	(3.32)	(2.81)	(3.33)
Marital status	0.162	-0.0253	0.166*	− 0.0257
	(1.93)	(-0.41)	(1.97)	(− 0.42)
Racial minority	− 0.0897	− 0.0222	− 0.0921	− 0.0199
	(− 1.31)	(− 0.60)	(− 1.34)	(− 0.54)
Income perception	0.168***	0.171***	0.168***	0.170***
	(5.11)	(7.20)	(5.11)	(7.16)
Income	0.000223**	0.000126**	0.000205**	0.000110**
	(2.93)	(3.14)	(2.67)	(2.76)
Industry occupation dummies	No	No	No	No
Fixed effects	Yes	Yes	Yes	Yes
N	1590	4208	1590	4208

t statistics in parentheses; *p < 0.05; **p < 0.01; ***p < 0.001

excluded categories

^aPrimary or less education

Region variables are omitted because of fixed effects

Besides our good and bad job definition, we construct a scale of quality of job (QJ) from 1 to 4, being 4 the best possible job and 1 the worst possible job. Let $\mathit{Formal}$ be a variable whether a person is employed in the formal sector and has formal employment ($Formal=Yes)$; if either condition does not hold, then $Formal=No$. Our quality of job measure is constructed as in Table 6.

Table 6.

Our quality of job measure

Quality of job (QJ)	Earns equal or more than the minimum wage	Formal
1	No	No
2	No	Yes
3	Yes	No
4	Yes	Yes

In Table 7, we show the average semi-(elasticities) transitions from and into each QJ category. The only significant effect is when (i) moving from QJ1 to QJ3 and (ii) moving from QJ1 to QJ4. In both cases, the job transitions that imply a gain in accessing the minimum wage are the only significant. Furthermore, a good job implies a better salary allowing people to satisfy their basic needs and have a better living standard. According to the Molina et al. (2015), on average between 2013 and 2014, the Basic Food Basket (BFB)9 calculated for a household of 4 members was $623 per month and considering that 1.6 members of the household work and earn exactly the minimum wage, the household income is $614. This is a negative $9 gap to access the BFB. In our sample, on average, people in a good job earn $795 and people in a bad job $278. This implies on average, with less than one member of the household with a good job, the household can access BFB, whereas 2.2 members of the household with bad jobs should work to buy the same basket. When moving from a bad job to a good job, the gain in salary is on average $265, approximately 43% of the cost of BFB.

Table 7.

Average (semi-) elasticities for labor transitions into different quality of jobs

	Life satisfaction (1)	Job satisfaction (2)
Job quality 1 to job quality 2	0.0566	0.0935
	(0.73)	(1.51)
	N = 2658	N = 2286
Job quality 1 to job quality 3	0.0448	0.0562
	(0.68)	(0.98)
	N = 3118	N = 2576
Job quality 1 to job quality 4	0.0486	0.103***
	(1.27)	(3.59)
	N = 5008	N = 3522
Job quality 2 to job quality 3	0.123	0.119
	(1.21)	(1.44)
	N = 704	N = 518
Job quality 2 to job quality 4	0.0755	0.0788*
	(1.29)	(2.19)
	N = 3404	N = 2004
Job quality 3 to job quality 4	0.0154	0.0178
	(0.27)	(0.52)
	N = 3318	N = 1912

t statistics in parentheses; *p < 0.05; **p < 0.01; ***p < 0.001

Conclusions, Theoretical Implications, and Limitations

This paper explores the idea that in developing countries, like Ecuador, unemployment is a minor concern since it is relatively low. Unemployment is not an option for economically disadvantaged people; they are forced to take bad jobs. Then, the real concern is those people in the labor force who are employed in bad jobs, i.e., earning less than minimum wage and employed in the informal sector with no social security protection.

In this paper, using data from the ENEMDU 2013–2014 panel for Ecuador, we explore whether labor transitions from bad to good jobs and vice versa impacts on life satisfaction and job satisfaction. We use a conditional logit model that allows for fixed effects to take into account time-invariant unobserved factors. Considering that before the transition in 2013, people who find/lose a good job are different in observable characteristics than people who remain in the same quality of jobs, consequently, we use the differences in differences estimator to control for endogeneity.

From the results, we can infer that labor transitions between good and bad jobs do not affect life satisfaction. We suggest that work is only one aspect of people’s life; other factors such as health status and subjective perception of income are determinants of life satisfaction in concordance with the subjective well-being literature.

We also found that labor transitions between quality of jobs influence on job satisfaction. The effect of moving from a bad job to a good job (9.5%) is slightly greater than moving from a good job to a bad job (− 8.5%). Using data from 2009 to 2010, which contains information only for the urban sector, we found similar results. Although objective income is not significant for life satisfaction, having a good job on average guarantees a decent standard of living since on average, a household can access a BFB and reduces the perception of poverty within the household.

Previous studies (e.g., Clark, 1997) had found that women experience greater life satisfaction than men; we found that the effect of labor quality transitions is greater for men than for women. An interesting idea for future research is exploring to what extent gender discrimination in the labor market might justify why those women value less than men the gain of finding a good job.

From the theoretical point of view, compared to previous studies, we could not observe exactly the adaptation and preference drift to past conditions considering the short-scale panel data. We suppose in terms of life satisfaction, there is some adaptation to bad jobs; as seen in Table 1, on average, people in bad jobs experience slightly less life satisfaction than people in good jobs. Therefore, people in bad jobs might be adapted to their conditions and having misleading perceptions, so they do not report poor life satisfaction, which could explain why the econometric estimates of labor transitions on life satisfaction are not statistically significant. However, in the absence of large panel data, we cannot rule out adaptation. Another alternative is that there might be other domains in life satisfaction that compensate for the fact of having a bad job. In terms of job satisfaction considering that the difference between good jobs and bad jobs is greater and the econometric estimates are statistically significant, adaptation seems less plausible.

A critical discussion is the relationship between both outcomes (life satisfaction and job satisfaction). Life satisfaction and job satisfaction might be positively correlated because people in poor-quality jobs would not be satisfied with their lives (spillover theory). Another alternative is that these variables might be negatively correlated since people find other pleasant activities in their lives to compensate for a poor-quality job (compensation theory). Bowling et al. (2010), based on a broad literature review, argue that the correlation between job satisfaction and life satisfaction ranges from 0.16 to 0.68, and the authors using meta-analysis support the spillover theory. In the case of Ecuador, the correlation is 0.11 and, thus, the evidence on the spillover theory is not conclusive. This low correlation might also indicate that there might be other more important determinants of life satisfaction than job satisfaction, for instance, income and health, according to subjective well-being studies. Life satisfaction encompasses many domains in life in a general score, whereas job satisfaction alludes only to job-related aspects.

Managerial Implications

Job quality is a combination of various dimensions that impact on individual self-assessment of work. Wage is conceived as a key determinant of job satisfaction; however, Clark (2015) based on empirical evidence mentions that workers care more about job security and interest in work compared to wages. From the firm perspective, promoting job quality could be considered costly; however, there are also potential benefits from it in terms of an increase in efficiency and productivity. Then, the challenge of firms is to concentrate their efforts on what is effective to foster working conditions. For instance, Ruubel (2021) found that workers who have flexibility in their working conditions experience a gain in their working environment. Flexibility is a wide concept, but the most notorious case is working from home. Mukherjee and Narang (2022) found that more than half of those who work from home have a gain in productivity and efficiency. In the same line, a key lesson from the COVID-19 pandemic was that Information and Communications Technology (ICT) was crucial to mitigate the negative effects of the lockdown on labor market outcomes and people’s well-being. Mofakhami (2021) found that the effects of ICT on job quality are positive in terms of optimizing the use of time and physical constraints, but it might boost the job pressure.

Limitations

This paper is not free of limitations. Specifically, we did assume implicitly that only one labor transition had occurred between December 2013 and December 2014. However, more than one transition could have occurred during that year; i.e., people move from a good job to a bad job and then to a good job. We believe that it is more likely the possibility of one transition because, after 1 year, the probability of remaining in a good job is 78%, and to remain in a bad job is 89%. This relatively low rate of transitions could be explained because Ecuador has a rigid labor market. According to Gwartney et al. (2016), Ecuador was ranked as the fifth most rigid labor market out of 159 countries analyzed in 2014. Considering that labor market rigidity makes it difficult to hire and fire and we use a short-scale panel (1 year), it is straightforward that labor mobility is low; it is expected that in a rigid labor market, job transitions into a different quality of jobs occur in longer periods.

Unlike Kassenboehmer and Haisken-DeNew (2009) and Chadi and Hetschko (2018), in our study, we could not distinguish between voluntary and involuntary job transitions. Therefore, despite taking into account time-invariant unobserved factors, the results should not be interpreted as a causal effect since job transition is not exogenous. Besides, from a macroperspective, macroeconomic trends and institutional performance affect the probability of a good job. From the labor economics perspective, search effort and asymmetric information between workers and firms also determine the job-finding process.

Future Research

We suggest that future research should incorporate the role of innovation, communication, and technology as criteria for job quality. Previous evidence indicated that KWs report higher job satisfaction than non-KWs (Torrent-Sellens et al., 2018; Viñas-Bardolet et al., 2020). Our good job definition is based on financial and social benefits received by workers and somehow is related to the basis of KWs since they are more educated, with higher wages and job satisfaction. In this line, it would be interesting to test the extent to which our good definition proxies with skills, education, autonomy, and use of ICT and to test whether the magnitude of effect of job transition on job satisfaction holds.

Appendix. Variable description

Job satisfaction (overall): 1 completely dissatisfied, 2 somewhat dissatisfied, 3 somewhat satisfied, 4 satisfied

Life satisfaction: in a scale of 0–10, where 0 = completely happy, 10 = completely unhappy

Female: 1 = female, 0 = male

Age: age in years

Primary: 1 = if primary is the highest degree completed, 0 = otherwise

Secondary: 1 = if secondary is the highest degree completed, 0 = otherwise

University: 1 = if university is the highest degree completed, 0 = otherwise

Hours: number of work hours last week

Self-reported health: in scale of 0–10, where 10 = very good health, 0 = very poor health

Income: labor income perceived last month in dollars

Income perception: 1 = if person considers actual income very good or good, 0 = if person considers income fair, poor, or very poor

Marital status: 1 = if married or in free union, 0 = if single or divorced

Racial Minority: 1 = if afro-Ecuadorian or indigenous, 0 = if white or mestizo

Region 1: 1 = lives in the coast region

Region 2: 1 = lives in the highland region

Region 3: 1 = lives in the Amazon region

Years working: number of years working

Working hours: number of working hours

More than one job: 1 = if worker has more than one job, 0 = otherwise

Permanent job contract: 1 = if worker has a permanent job contract, 0 = otherwise

Temporal job contract: 1 = if worker has a temporal job contract, 0 = otherwise

Vacations: 1 = if worker has vacation at work, 0 = otherwise

Health insurance: 1 = if worker has social insurance, 0 = otherwise

Training: 1 = if worker receives training at work, 0 = otherwise

Working establishment size: 1 = if working establishment has more than 100 employees, 0 = otherwise

RUC: 1 = if working establishment has more than 100 employees, 0 = otherwise

Formal sector: 1 = if worker is employed in the formal sector, 0 = otherwise

Social security: 1 = if worker is affiliated to social security, 0 = otherwise

Declarations

Research Data

Acosta-González, H. Nicolás (2020), “Bad jobs versus good jobs: Does it matters for life and job satisfaction? Evidence from Ecuador”, Mendeley Data, V1, https://doi.org/10.17632/6pfpvt78t9.1.

Conflict of Interest

The authors declare no conflict of interest.