University of Groningen – Joeri Everaers
Is labor flexibility
bliss?
1
Master Thesis - Economics
The effects of employment protection are often subject of debate between liberals and labor unionists. While the former favor efficiency, the latter favor security. This trade-off is hard to assess using the
standard revealed preference theory as the welfare effects of job security are difficult to measure. This thesis uses a radically new way to overcome this problem. By using subjective well-being (SWB) as a
trustworthy proxy for welfare, both the efficiency and security effects can be compared on an equal scale. The analysis confirms the hypotheses that employment protection positively influences welfare
and that this effect significantly differs between insiders and outsiders of the labor force.
Joeri Everaers
Eerste Helmersstraat 78-2
1054 DL Amsterdam
0031 – (0)654361757
j.everaers@student.rug.nl
S1559575
First supervisor:
Dr. A.A.J. van Hoorn
2
Outline
Outline ... 2
Chapter 1 – Introduction ... 4
Chapter 2 – Background of SWB ... 6
2.1 Introduction ... 6
2.2 History and definition of SWB ... 6
2.3 Different uses of SWB ... 7
2.3.1 Subjective evaluations in health care ... 7
2.3.2 SWB in a cost-benefit analyses for public policy ... 7
2.4 SWB as a complement to the revealed preference theory ... 9
2.4.1 De gustibus (non) est disputandum – revealed preference theory ... 9
Chapter 3 - Theory & Hypotheses ...11
3.1 Introduction ... 11
3.2 Labor flexibility and employment ... 11
3.2.1 Labor institutions and the labor market ... 12
3.2.2 Employment protection ... 13
3.3 The effect of unemployment on SWB... 14
3.4 Hypotheses ... 16
Chapter 4 – Data and method ...18
4.1 Introduction ... 18
4.1.1 Dependent variable ... 18
4.1.2 Independent variables ... 19
4.1.4 Descriptive data ... 20
4.2 Model ... 22
4.2.1 Model 1 – The overall effect of employment protection on the SWB of society ... 22
3
Chapter 5 – Results ...26
5.1.1 The overall effect of employment protection on the SWB of society (Model 1) ... 26
5.1.2 Insiders as individuals who are full-time employed (Model 2a) ... 27
5.1.3 Insiders as individuals with a high probability of employment (Model 2b) ... 28
5.2 Robustness checks ... 30
5.2.1 Broadening the definition of employment ... 30
5.2.2 Taking a different source for SWB ... 30
5.2.3 Including additional control variables ... 31
5.2.4 Taking a different source for employment protection ... 32
5.2.5 Using a different sample ... 32
Chapter 6 – Conclusions ...35
Chapter 7 - Limitations ...36
References ...37
Appendix I - Debate SWB as overarching policy goal ...48
I.I Conclusion on SWB as overarching government goal ... 51
Appendix II – The rationality assumption ...55
II.I Rational Choice Theory ... 55
II.I.I Rational Choice Theory itself ... 55
II.I.II Utility ... 56
II.II Difficulties of rationality... 56
II.II.I Explanations for inconsistencies ... 58
4
Chapter 1 – Introduction
Typically, economists are eager to elaborate on the advantages of labor market flexibility. By lowering employment protection, labor becomes cheaper. This causes the demand for labor to rise, and with it employment and welfare. But not all agree. Other economists stress that upholding a certain level of consumption is needed to spur economic growth. By lowering employment protection, employees are laid off during economic crises, which causes their expenditure to drop. This drop in expenditure in turn aggravates the economic downfall. Besides economists, politicians also disagree. While the liberals tend to follow the laissez faire economist, left wing politicians argue that labor should be protected. Not just to stimulate consumption, but also to contribute to the feeling of safety and thereby to the welfare of the population.
So there exist contrasting approaches regarding employment protection and welfare. How can be assessed which point of view is correct? Employment protection has several implications for society and affects insiders differently than outsiders of the labor market. On the one side, it lowers job flows and thereby dampens economic efficiency. On the other side, it increases security for people who have a job and thereby protects the consumption levels from income volatility for credit-constrained and risk-averse agents. While the first effect is relatively easy to quantify, the second effect is often ignored by traditional economic research. Partly because of this inability to incorporate all the effects of employment protection into one model, the debate concerning employment protection has been stuck. This thesis uses a novel method of DiTella et al. (2001) to internalize both the efficiency as well as the security effects by measuring the direct effect of employment protection on welfare by using data on SWB. In addition to using SWB as a proxy for welfare, an instrumental variable is created which calculates the probability of employment for each agent. This probability of employment is used to distinguish between insiders and outsiders of the labor market and to compare both groups. By combining these two unorthodox methods, new insights are discovered. Insights that supplement the knowledge of labor institutions and that enable the debate concerning employment protection to reach a new level.
DiTella et al (2001) analyze the assumed effect of inflation and employment on the welfare of a society and succeed in providing evidence that both factors indeed influence the societal welfare function. They succeed because they directly measured the effect of both inflation and employment on welfare, measured as SWB. Instead of focussing on indirect factors such as employment levels, job flows, or economic growth, this thesis follows their example by examining the direct effect of employment protection on SWB.
5
estimates how likely someone is to be employed), and comparing the effect of employment protection on the SWB of agents with a high probability of employment with agents who have a low probability of employment.6
Chapter 2 – Background of SWB
2.1 Introduction
This thesis builds heavily on the literature regarding SWB. Because SWB is relatively new in economics, the beginning of this section outlines a short history of SWB measurement. This is followed by three different ways in which SWB is used: in cost-benefit analyses, health issues, and as a complement to revealed preference theory and analysis.
2.2 History and definition of SWB
Happiness has been a topic of concern and contemplation for over two millennia. In the ancient writings of Aristotle in the fourth century B.C.E., happiness is described as an ‘activity of the soul that
expresses virtue’. Aristotle claimed that true happiness is achieved by fulfilling life’s purpose, identified
by Aristotle as reason. Since Aristotle, many philosophers have examined the concept of happiness. While some merely focus on happiness in the afterlife, others follow Locke in stating that happiness equals good just as pain equals evil (McMahon, 2004).
In his book ‘The principles of morals and legislation’, Jeremy Bentham (1832) introduced the greatest happiness principle. This moral principle entails that: ‘The greatest happiness of the greatest
number is the measure of right and wrong’. Bentham thereby claims that ethics should be about the
maximization of happiness. The actual measurement of happiness took flight approximately four decades ago. What started with a paper by Wilson (1967), turned into an extensive branch of research. Through twin studies and behavioural economics, much has been learned concerning the determinants of SWB.
When the literature concerning SWB is examined, the definition of Diener et al. (1999) stands out. They claim that: ‘SWB is a broad category of phenomena that includes peoples’ emotional
responses, domain satisfactions, and global judgment of life satisfaction’. According to this definition,
SWB is not a single specific construct, but rather a broad area of interest. Within this concept of SWB there are two different components. One of these components focusses on the affective part. Feelings of happiness and other positive affects stimulate this part, while negative feelings, such as pain, suppress this component. The affective component thereby relies heavily on the moods and emotions of individuals (Diener et al. 1999). Some argue that this affective part should be divided into positive and negative affects as well (Diener and Emmons, 1984; Diener, Smith, and Fujita, 1995).
7
Although there are clear differences between the two components of SWB, this thesis follows most research done on SWB by using the different components interchangeably as SWB. Both components are highly correlated (Chamberlain, 1988 – Review SWLS Diener).2.3 Different uses of SWB
SWB can be used for multiple purposes and in this section three different applications are discussed. These applications are health care, cost-benefit analyses and as a complement for the widely used revealed preference theory.
2.3.1 Subjective evaluations in health care
The best-known use of subjective evaluations is in the health care sector. Physicians and psychologists use SWB evaluations to make prognosis concerning the health of individuals. Through relying on the answers given by patients, doctors are able to locate and identify certain illnesses or disorders. The subjective assessment of well-being is thereby considered to be of great importance and is heavily relied upon.
In recent years the health care sector has also adopted the concept of Quality Adjusted Life Years (QALY) to make a cost-utility analysis possible. In this analysis, the costs of a treatment are compared with the benefits measured in the number of life years added by the treatment and multiplied by correction for the quality of these life years. SWB plays an important role in determining the quality of the life years.
2.3.2 SWB in a cost-benefit analyses for public policy
Government policy is currently valued through a large number of indicators. While per-capita income is one of the most popular measures, unemployment, the Human Development Index, and the level of freedom are other examples which are used to measure the outcomes public policies. This section starts with a brief outline of the current political developments regarding SWB as measuring rod for the outcomes of public policies. This is followed by the criticism on using Gross National Product (GNP) as most important policy indicator and concludes with the advantages of using SWB as an important indicator for public policy.
8
of National Statistics to ask citizens what indicators they think should be used to measure happiness. In other nations (Germany, Italy, USA, etc.) reforms are under way in devising new indicators that could be used to measure the total effects of public policies. With this, they are clearly moving away from the opinion that GNP is the sole and most important indicator for government policy.Using mainly GNP as policy indicator has several major disadvantages. In his paper ‘Myths of economic growth’, Haque (2004) summarizes the different reasons why GNP is insufficient in determining the development of a country. These reasons vary from the difficulties measuring the underground economy to the impossibility to measure non-economic dimensions.
Because of the downsides of mainly using GNP as policy indicator, this thesis argues that SWB can be very useful in complementing or even overtaking GNP as main indicator. The advantages are summarized by Frey and Stutzer (2010) in ‘Happiness and Public Choice’. Frey and Stutzer give four reasons why they deem SWB is a more important cost-benefit tool to measure public policy than income, Gross National Product per capita, or extensions of GNP such as the Human Development Index. These reasons are:
(i) Unlike income, SWB also measures important non-material components of well-being. Components such as human relations, autonomy, and self-determination are very important to human well-being and should therefore be taken into account.
(ii) SWB measures the outcome of government activity rather than the input (number of hospital beds or number of teachers). This enables a better analysis of the different effects. By measuring the actual effect in a direct matter, one is able to get a better grasp of the effects than if one studies objective indicators (Diener and Suh 1997; 205).
(iii) Happiness measures and evaluates the subjective outcomes that are in line with the methodological approach of economics, this in contrast with the capabilities approach and the Human Development index.
(iv) SWB measures both outcome and procedural utility.
9
2.4 SWB as a complement to the revealed preference theory
This section shortly explains how revealed preference theory (RPT) is used to analyse economic issues and discusses two impracticalities regarding this framework.
2.4.1 De gustibus (non) est disputandum – revealed preference theory
Samuelson (1948) is considered to be the founding father of the RPT. In his papers, A note on the
pure theory of consumer behaviour and Consumption theory in terms of revealed preference ,
Samuelson states that the preference pattern of an individual can be deduced by examining which batch of goods is preferred to another batch. The utility curve of an individual is thus derived from the choices the individual makes.
Sigler and Becker (1977) take it one step further in their paper De gustibus (non) est
disputandum, which is Latin for tastes are not disputable. In this paper the authors claim:
’...that tastes neither change capriciously nor differ importantly between people. […] one does not
argue over tastes for the same reason that one does not argue over the Rocky Mountains—both are there, will be there next year too and are the same to all men. […] On the traditional view, an explanation of economic phenomena that reaches a difference in tastes between people or times is the terminus of the argument: the problem is abandoned at this point to whoever studies and explains tastes..’
Sigler and Becker conclude by stating that ‘we should apply standard economic logic as extensively as
possible’. The authors therefore claim that economists should only focus on differences in prices and
income to explain any changes in behaviour. The RPT combines the demand function with the utility function through observing the behaviour of agents. By assuming agents are utility maximisers, a utility function can be deduced by analyzing the choices made by agents.
Individuals are assumed to make rational decisions which optimize their own utility within the given constraints. By making a theory of consumer choice, it is claimed that even decisions which stem from an addiction or advertisement, are utility maximizing. The goal for economics and government policy is then to maximize the choice set for the individuals. If the individuals have more to choose from, they can raise their utility in the most efficient way. The method most commonly used to expand the individuals choice set is wealth. By increasing wealth, the individual faces an increase in choices which is then logically followed by an increase in utility.
10
2.4.1.1 Irrationality as an argument against revealed preference theory
As outlined by Gruber and Mullainathan (2003), addicts are likely to make choices that do not optimize their long term utility. Although it is theoretically possible to argue that in fact addicts are behaving in an optimal way, most would agree that addicts are irrational. But while most agree that individuals are not always rational, the problem is whether this forms a serious threat to traditional economic theory. If small inconsistencies, which have little effect, are the only problem, there is not much to worry about and basic economic theories (although not perfect) are still viable. However, the absence of rationality could just as well severely hamper the effectiveness of RPT. If people do not rationally optimize their utility, merely increasing the choice set of individuals does not always lead to an increase of their overall utility.
Behavioural economics has shown that people are only partly rational. There are several reasons that explain this irrationality. The reasons mentioned are the endowment effect, status quo bias, loss aversion, inability to predict future utility, and lack of willpower. The factors combined outline why people are likely to fail to make the choices that optimize their overall utility. An extensive discussion of these factors is found in Appendix II.
2.4.1.2 Non observable preferences as a problem for revealed preference theory
Besides irrationality difficulties, RPT cannot be used to analyze the welfare consequences for public policy or economic phenomena when agents do not actively make observable choices. For instance, when the effects of inflation on welfare are examined, it is not feasible to focus on the choices made by individual agents. This is simply because they do not actively chose the level of inflation. While it is possible to look at the choices made by democratic governments, and assume that they act in the interest of the utility of the individuals who chose them, the actual welfare effects of for instance inflation are still very difficult to deduce.
The same goes for the effect of job security on the welfare of society. Although it is likely that agents are risk-averse (see Appendix II), RPT is unable to clearly quantify the positive effect of job insurance on welfare as no observable choices are made. Therefore, the effects of the trade-off between insurance and efficiency cannot be fully examined using RPT.
11
Chapter 3 - Theory & Hypotheses
3.1 Introduction
The goal of this thesis is to analyze the welfare effects of labor flexibility on society. Labor flexibilization has two different effects on the labor market. On the one hand, labor flexibility raises efficiency as labor is able to move around more easily. Because of this, labor is better able to maximize its productivity. On the other hand, a high level of labor flexibilization means a lower level of insurance. Employees are less certain they can keep their job in the future, and because they are assumed to be risk-averse, this uncertainty lowers their welfare. In sum, labor flexibilization is a trade-off between efficiency and insurance (Blanchard et al. 2006). This trade-trade-off is impossible to examine through RPT as the welfare received through job security cannot be measured within this framework. Therefore this thesis chooses to use SWB as a proxy for welfare. By using SWB as a proxy for welfare, this thesis is able to deduce directly the effect of labor flexibilization on welfare, thereby circumventing the infeasibility of using RPT to analyze the earlier mentioned trade-off.
This chapter focusses on the relationship between employment protection and welfare, measured as SWB. It is widely assumed that a more flexible labor market, and thus a lower level of employment protection, leads to a lower level of structural unemployment and to a higher level of temporary unemployment. The strong positive relationship between employment and SWB is undisputed as well. Unemployment leads to a relatively large drop in SWB, a drop far larger than the loss of financial wealth accounts for. The question remains whether less employment protection results in a change in SWB.
This section analyzes the literature regarding employment protection, unemployment and SWB, and accordingly states two hypotheses. The first part analyzes the relationship between labor flexibility and employment. This part is followed by a section which specifically examines the relationship between employment protection and employment. The third part focusses on the relation between employment and SWB. The final part differentiates the effects between insiders and outsiders in the labor force and uses all the information to formulate the hypotheses.
3.2 Labor flexibility and employment
12
3.2.1 Labor institutions and the labor market
After thirty years of extensive research regarding the labor market, no clear and widely accepted theory on employment has arisen. While quite a few theories have come and gone, it is still unclear how exactly the labor market functions. As Blanchard et al. (2006) state: ‘we have learned a
lot, but we still have a lot to learn’. This section aims to give a short and global overview of the existing
theories regarding the labor market institutions. This is done on the basis of the overview provided by Blanchard et al. (2006).
Sachs and Bruno (1985) were among the first to argue that unemployment could be explained through the interaction between shocks and two types of rigidity, real and nominal. After a negative shock, real wage rigidities hindered the wage to drop to the competitive level (in which the wages equal productivity). The longer it took for the wages to drop, the harsher the effect would be on unemployment. Slow adjustment of the nominal wage leads to a longer lasting difference in the real wage and the unanticipated increase in prices; a difference which also results in unemployment but can be partly overcome by monetary policy.
Calmfors and Driffil (1988) made an important contribution by showing that, after an adverse supply shock, the bargaining process has a significant effect on unemployment. Countries with either very centralized or very decentralized bargaining would be better off than countries with intermediate bargaining. If the countries had a very centralized bargaining process, both parties (employers and employees) integrated the need for low unemployment, thereby lowering their wage demand in order to save jobs. If a country had a very decentralized bargaining process, the bargaining power of the insiders (employees) was too small to demand high wages.
A theory of capital accumulation was the focus of Bean and Dreze (1991). They argued that a drop in productivity growth leads to unemployment due to sluggish wages. Because of sluggish wages, firms needed to reduce their employment rate. This drop in employment caused the profit rates to drop as well. This creates a vicious circle as lower profits result in less employment. As long as the profit rate is lower than the user cost, both capital stock and employment drops. Because of this vicious circle, a shock can result in a long period of unemployment.
The influential OECD Jobs Study of 1994 states that ill-adapted labor institutions are the main source for the high levels of unemployment. The labor rigidity caused by these institutions should be overcome and, as stressed by the OECD, labor markets should become more flexible. These reforms ought to entail a redesign of unemployment insurance and protection, the reduction of the tax wedge and the minimum wage, and more active labor market policy programmes.
This brings us to the empirics. In 1997, Nickell found that the difference in institutions was able to explain the large differences in the unemployment levels of twenty OECD countries during the 1980s and 1990s. The institutions which seem to matter most were the duration of the employment benefit and the degree of coordination of collective bargaining.
13
remained insignificant. Layard and Nickell confirmed these results in 1999. Although this explained a part of the labor market dynamics, a large part of the unemployment remained unexplained. Fialova and Schneider (2009) have found similar results. In their research high taxes were negatively correlated with employment and ALMPs seemed to stimulate employment. Fialova and Schneider did find that employment protection had a negative effect on both the participation rate as well as the employment rate.Blanchard and Wolfers (2000) claim that the interaction between macroeconomic shocks and labor institutions are the cause for cross-country differences in unemployment. While the shock initially boosts the unemployment, the institutions are responsible for the long-lasting effects. Because of the institutions, a transitory increase in unemployment is changed into a long lasting or permanent increase. The institutions are employment benefits, employment protection and a low level of bargaining coordination.
It seems safe to claim that some labor institutions, such as employment benefits and certain bargaining structures, have a negative effect on employment. For employment to remain high, a flexible labor market is needed which enables labor to move quickly to the most efficient position.
3.2.2 Employment protection
It has become clear that rigid market institutions are at least partly to blame for unemployment. Employment insurance and certain bargaining structures hinder the efficient allocation of labor and therefore contribute to higher levels of unemployment. But what about employment protection? Although it is clear that it contributes to the labor rigidity, the overall effects on unemployment remain unclear as it protects employment as well. This section provides a more extensive analysis of the literature regarding employment protection and shows that, although the effect on overall employment is ambiguous, it has a large effect on the character of the labor market.
3.2.2.1 Effect employment protection on job flows
In 2004, Gomez Salvador et al. concentrate on the difference in job flows between different European countries. Their data show that countries with high levels of employment protection have relatively low levels of job reallocation and creation. They found no effect on the job destruction rate. This result implies that higher employment protection leads to a more rigid labor market in which the allocation of labor is less efficient.
14
permanent jobs. That is, in Portugal less people actually lose their job, but if they lose their job they are bound to be longer unemployed than their American counterparts. This causes the overall level of employment to remain the same, but it has a dampening effect on the labor flows. Lower labor flows means that less productive jobs are harder to be replaced by more productive jobs. A dampening of the labor flows therefore reduces the efficiency of the labor market.Some puzzles regarding employment protection still remain. Blanchard and Jimeno (2005) examine the difference in unemployment between Spain and Portugal. While both have a similar level of employment protection, the unemployment levels of Spain and Portugal are very different. In the end they argue that the difference might be attributed to the difference in union power and unemployment insurance. But clear evidence remains absent.
3.2.2.2 Effect of employment protection on different groups
In addition to the change from structural to temporary unemployment, employment protection affects different groups differently (Alfonso and Gilles, 2006; Blanchard et al. 2006). The reforms on employment protection have been mainly aimed at getting outsiders (i.e. women, young and low-skilled workers) employed. It is therefore likely that the effects of employment protection have different effects on both types of labor participants.
In economic theory, employment protection protects the consumption levels from income volatility for credit-constrained, risk-averse agents. Employment protection increases the cost of labor for employers. These higher labor costs are mostly at the expense of groups with a high labor supply elasticity (youth and women) and people who face a high level of labor demand elasticity (youth and low skilled). That the disadvantages of employment protection are more significantly felt by certain groups (outsiders) is confirmed in research done by Alfonse and Gilles (2012). The reforms regarding employment protection in the EMU are followed by a positive and significant rise in their employment growth. This is likely to be the result of companies who find the employment protection to be the binding constraint, and who therefore build up a buffer stock of flexible workers if employment protection is lessened. This additional stock consists of workers who would otherwise have been unemployed.
3.3 The effect of unemployment on SWB
This section aims to outline the effects of unemployment on the SWB of both the individual and society. At first, the overall effect of unemployment on SWB is analyzed. The second part focusses on the question whether long term or short term unemployment has a different effect on the SWB of the individual. The final part outlines the different effect of unemployment on SWB for different groups in society.
15
beyond the mere loss of financial means. It is assumed that besides the financial loss, the loss of self-esteem and social relationships are responsible for this strong effect (Layard 2005). Besides the set-back of the individuals’ SWB, a high level of unemployment also lowers the SWB of individuals seemingly unaffected by the lay-off (Di Tella et al. 2003). For example, an average recession in the United States causes a loss of SWB which amounts up to $200 for the average individual (Frey and Stutzer, 2000). There exists a large amount of data that all confirm the strong relation between unemployment and low levels of SWB (Warr et al., 1988; Clark and Oswald, 1994; Winkelmann and Winkelmann, 1998 ; Frey and Stutzer, 2000; Di Tella et al., 2003; Blanchflower, 2001; and Clark, 2006).3.3.1 The effect of short- and long term unemployment
This thesis is interested in the effect of employment protection on SWB. As is stated before, a drop in employment protection is likely to lead to lower levels of structural unemployment and to higher levels of temporarily unemployment. This begs the question whether temporarily or permanent unemployment have different effects on the SWB of individuals.
While some argue that long periods of unemployment (>2 years) cause the individual to adapt to the situation (for instance through better budgeting and psychological adaptation), others state that long term unemployment leads to the diminishing of skills and the attendance of despair (Clark, 2006). The evidence for either diminishing or increasing effects of long-term unemployment remain ambiguous. Some, such as Warr and Jackson (2002) and Goldsmith et al. (1996), find an increase in SWB after a longer period of unemployment. By using the National Longitudinal Survey of Youth (NLSY), the latter finds that the duration of the unemployment has a positive effect on self-esteem. The longer the person is unemployed, the better he or she seems to feel (although the SWB does not reach the pre-unemployment levels). Thus, some kind of adaption seems to occur. Warr and Jackson (2002) find similar results and state that the mental health of the unemployed increased after a period of 25 weeks. Clark et al. (2001) broadened the definition of the period of unemployment by adding all the unemployment of the last three years (which could have been ‘interrupted’ periods of employment). With this broad definition of unemployment, the data also suggests that past unemployment lowers the negative effect of unemployment on SWB. The psychological effect of unemployment becomes less if the individual has more experience with being unemployed.
However, others, such as Winkelmann and Winkelmann (1998), fail to find a positive relationship. Their data shows no increase in SWB after a certain period of unemployment, something which would suggest that no adaption takes place. After conducting a large panel data research, which focuses on Europe in the 1990’s, Clark (2006) concludes that although there was mixed evidence in the past, ‘habituation to unemployment does not seem to be a widespread phenomenon
[...]’. While the negative effect of unemployment on SWB is evident, the question regarding the
16
3.3.2 Effect unemployment on different groups
In ‘Unemployment as a social norm’, Clark (2003) finds that the environment in which one lives, is of significant importance to the effect of unemployment on SWB. While unemployment is still negatively related to SWB, this effect becomes less strong when more people in the reference group are unemployed as well. While the unemployment of person A negatively affects the SWB of the employed person B, it has a positive effect on the SWB of the unemployed person C. In other words, if a person becomes unemployed the drop in SWB is influenced by the unemployment levels in his or her reference group. The SWB is less affected if a larger proportion of this group is unemployed as well, and vice versa.
This effect is endorsed by Stutzer and Lalive (2004). They assume that the percentage of citizens in a community who voted in favour of a reduction of employment benefits in a Swiss referendum is related to the social norm in that community. If more people were in favour of lower employment benefits, the social norm to work was assumed to be stronger. They find two results. The first is a significant negative relation between the strength of the social norm and the length of the average unemployment period. A stronger social norm to work leads to a shorter period of unemployment. They also find that the drop in life satisfaction following unemployment, is stronger when the social norm to work is higher.
Current literature thus suggests that the strength of the social norm to work is an important aspect when the effect of unemployment on SWB is considered. Because it is likely that the social norm to work is stronger among insiders than among outsiders, it is reasonable to assume that the SWB of outsiders is less affected by unemployment than the SWB of insiders.
3.4 Hypotheses
17
Table 3.4
The effect of employment protection on SWB Effect employment protection on employment Strength of effect unemployment on SWB Overall effect Whole society Insiders +++ Outsiders +
Combining this knowledge gives the following two hypotheses:
(1) Employment protection has a positive effect on the SWB of society.
(2) The effect of employment protection will be less positive for outsiders than for insiders.
The following chapters will test these hypotheses by using two models. In the first model, the effect of employment protection on the SWB of the whole society is analysed. Because the positive effect for insiders is stronger than the negative effect for outsiders, the overall effect of employment protection on the SWB of society is assumed to be positive. This effect is deduced by comparing the effect of employment protection on SWB between people who are active in the labor force and people who are not (and are thus likely to be unaffected by labor market institutions). If people who are active in the labor force benefit significantly more from employment protection than the people who are unaffected by it, the hypothesis is correct.
18
Chapter 4 – Data and method
4.1 Introduction
This section outlines both the data used by this thesis and the method to analyze the effect of employment protection on SWB. The first section focusses on the dependent variable, SWB, and the independent variables. The second section explains the models that are used to analyze the
relationship between employment protection and SWB.
4.1.1 Dependent variable
The dependent variable of this thesis is SWB. SWB is measured in a variety of ways, but most often through self-evaluating questionnaires. Researchers use both single item scales and multiple item scales. This thesis uses a single item scale which is outlined below.
Single item scales
The questionnaires can be composed out of several different questions but quite often consist out of just a single question. The two questions most frequently used are:
(1) Taking all things together, how satisfied are you with your life as a whole these days?
1 2 3 4 5 6 7 8 9 10
Dissatisfied Satisfied
(2) Taking all things together, would you say you are: 1- Very happy
2- Quite happy 3- Not very happy 4- Not happy at all
The World Values Survey, the survey used in this thesis, uses these two questions. By using this single item scale, one is able to easily compare different SWB levels between individuals or, when large samples have been taken, between countries. The first question, which focuses on satisfaction rather than happiness, is used as the main dependent variable of this thesis.
Reliability and validity of SWB
19
(Frey and Stutzer 2010). Multiple studies have found that the reported data on SWB is moderately stable, but sensitive to changes in the lives of individuals (Ehrhardt et al. 2000; Schimmack and Oishi 2005 – Frey and Stutzer 2010). Consistency tests indicate that people with high levels of SWB smile more (Fernandez-Dols and Ruiz-Belda 1995), are considered to be happy according to family, friends and spouses (Sandvik et al. 1993; Lepper 1998; Costa and McCrae 1988), express more positive emotions, are more optimistic, sociable, extrovert and sleep better (Frank 1997; Frey and Stutzer 2002). SWB is also heavily correlated with brain activity (Urry et al. 2004) and suicides (Daly and Wilson 2009).4.1.2 Independent variables
This thesis uses a variety of independent variables of which the most important are employment protection and the instrumental variable probability of employment. Other independent variables are used as control variables.
4.1.2.1 Employment protection
The employment protection of a country is measured with the indicators of employment protection of the Operation for Economic Co-Operation and Development (OECD) database. This indicator ‘measures the procedures and costs involved in dismissing individuals or groups of workers and the
procedures involved in hiring workers on fixed-term or temporary work agency contracts’. The indicator
is compiled from 21 items covering three different aspects of employment protection and is scaled between 0 (least stringent) and 6 (most stringent).
The first aspect is the individual dismissal of workers with regular contracts. This aspect covers items such as notice periods, severance pay, and the difficulty to dismiss workers. The second aspect covers the additional costs for collective dismissals. This measures the additional costs an employer has when a large number of workers need to be dismissed. The final aspect focusses on the regulation of temporary contracts. This quantifies the regulation of fixed-term and temporary work contracts and includes the regulation governing the establishment and operation of temporary work agencies.
4.1.2.2 Operationalisation of the distinction between insiders and outsiders
20
using the most important predictors of employment: year, gender, age categories, health, marital status, and education categories. These variables are found through examining current literature on employment and by testing their significance. Although these covariates only enable a rather rough estimate, it is sufficient to enable a distinction between insiders and outsiders of the labor force, i.e. between people with a high probability of employment and people with a low probability ofemployment.
4.1.2.3 Control variables
This thesis uses a large number of control variables to account for other factors that could influence SWB. These variables are: country, year, religion, gender, age categories, marital status, education categories, employment status, and income. These variables are found by analyzing current literature and testing for significance.
4.1.3 The data source – World Values Survey
Besides the data on employment protection, this thesis uses the World Values Survey (WVS) for data. The WVS is a research program that focuses on basic human values, ranging from happiness to opinions and beliefs. The survey started in 1981 and is conducted every nine years. The WVS uses standardized questionnaires in approximately 96 societies, containing roughly 90% of the world’s population. Because of the extensive amount of data, the WVS is ideal for a multilevel analysis focused on the levels of SWB in OECD countries.
4.1.4 Descriptive data
The 28 countries used in this research are Australia, Austria, Belgium, Canada, Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Japan, Korea, Mexico, Netherlands, New Zealand, Norway, Poland, Portugal, Slovak Republic, Spain, Sweden, Switzerland, Turkey, United Kingdom, and the United States. The years used are between 1989 and 2008. In table 4.1, the descriptives are given. These countries and years are used because of data availability.
Table 4.1.4
Descriptive Statistics
N Minimum Maximum Mean Std. Deviation
Year 113501 1989 2008 -- --
Satisfaction with your life
113501 1 10 7,24 2,105
State of health (subjective)
21
Religious person 106728 1 3 1,43 ,611 Sex 113384 1 2 1,52 ,499 Age/100 113238 ,15 ,98 ,44 ,171 Marital status 113120 1 8 -- -- Highest educational level attained 80323 1 8 4,52 2,236 Scale of incomes 96127 1 10 4,81 2,546Full time employed 113501 ,00 1,00 ,3953 ,48891
Part time employed 113501 ,00 1,00 ,0821 ,27458
Self employed 113501 ,00 1,00 ,0718 ,25821 Retired 113501 ,00 1,00 ,1815 ,38540 Housewife 113501 ,00 1,00 ,1347 ,34140 Student 113501 ,00 1,00 ,0580 ,23371 Unemployed 113501 ,00 1,00 ,0564 ,23069 Employed - other 113501 ,00 1,00 ,0202 ,14078 Inadequately completed elementary education 113501 ,00 1,00 ,0530 ,22404 Completed elementary education 113501 ,00 1,00 ,1312 ,33760 Incomplete secondary school: tech or voc type
113501 ,00 1,00 ,0810 ,27284
Complete secondary school: tech or voc type
113501 ,00 1,00 ,1056 ,30727 Incomplete secondary: university-preparatory 113501 ,00 1,00 ,0658 ,24788 Complete secondary: university-preparatory 113501 ,00 1,00 ,1110 ,31411
Some university without degree
113501 ,00 1,00 ,0616 ,24037
University with degree 113501 ,00 1,00 ,0986 ,29816
Eduction - other 113501 ,00 1,00 ,0516 ,22119 Married 113501 ,00 1,00 ,5843 ,49284 Living as married 113501 ,00 1,00 ,0509 ,21984 Divorced 113501 ,00 1,00 ,0490 ,21582 Seperated 113501 ,00 1,00 ,0165 ,12750 Widowed 113501 ,00 1,00 ,0711 ,25692 Single 113501 ,00 1,00 ,2243 ,41715 Predicted probability of employment 113329 ,04683 ,82222 ,3975167 ,14350643 Employment protection 87067 -,95 1,88 ,4552 ,76273
22
4.2 Model
This section outlines the two models used by this paper to examine whether employment protection has an effect on SWB and whether it affects insiders differently than outsiders. The data from the WVS is hierarchically structured, with individual observations (level 1) who have a certain level of SWB, who live in countries (level 2) that have a certain level of employment protection.
There are several methods to analyze hierarchically structured data. The first is to focus solely on the individual level. The data is disaggregated and all the characteristics are assigned to the individual. However, this method violates the independence assumption as groups of individuals share particular characteristics. It would be possible to add an independent categorical variable to account for the differences between countries and years. However, this would merely shift the mean level of SWB up or down because it still assumes that different factors (such as employment protection) have the same effect in all the different countries. In reality however, it is far more likely that different policies have different effects (both on the economy as well as on SWB) in different countries. A second method is to aggregate the country characteristics, thereby neglecting all the with-in group differences. This way, valuable information is wasted.
With a multilevel analysis, all the different (cross-level) effects are taken into account. A multilevel analysis thereby generates cluster robust standard errors, which could arise because of unobserved, state-level effects in the error term (Moulton 2001; Wooldridge 2003). Because this paper uses determinants from two levels, such a multilevel analysis is used in order to optimally use all the available data, including the cross-level interactions (Bryk and Stephen, 1992; Gelman and Hill, 2007). Because of these benefits, other studies regarding SWB also use multi-level analyses (Pittau et al, 2010; Hui et al., 2004).
The first model focusses on the effect of employment protection on the SWB of the whole society. This is done by comparing the effect of employment protection on SWB of people who are in the labor force with people who are outside the labor force (and who are thus unaffected by the labor policies). The second model determines whether insiders and outsiders are affected differently by employment protection. The second model uses two different methods to distinguish between insiders and outsiders. At first, insiders are individuals who are currently full-time employed. The second method uses an instrumental variable which calculates the probability of employment for each individual. Individuals with a high probability are considered insiders and vice versa.
4.2.1 Model 1 – The overall effect of employment protection on the SWB of society
The first model has the following characteristics. The is the SWB of individual who lives in
23
sample mean. This deviation is computed by subtracting the average level of EP from the level of EP indicated by the OECD.In the model, the effect of employment protection on people in the labor force is compared with people outside the labor force as it is likely that employment protection is only relevant for people who participate in the labor market. The distinction between people inside and outside the labor force is done by looking at their age. People who are either too young or too old to work (<18 and 65<) are assumed to be outside the labor force. People who are between the age of 18 and 65 are considered to be in the labor force. By solely looking at the difference in effect between the two groups, all other factors which could influence SWB are controlled.
This comparison is done by creating the dummy ACTIVELABOR, which includes all persons between the age of 18 and 65. is the coefficient of the interaction effect between employment protection and the active labor force. The first model is:
(1) = + µΩ + ACTIVELABOR + * ACTIVELABOR
To analyze whether there is a different effect of employment protection on the two groups (in- and outside the labor force), should be analyzed. It is important to notice that the coefficient rather than , is the most important element. The focus is on as it tells whether deviations in employment protection from their sample mean, cause a relative change in the SWB of people in the labor force relative to people who are neutral under changes in employment protection. If people in the labor force benefit from employment protection, this coefficient is positive (and vice versa).
4.2.2 Model 2 – The differentiated effect on insiders and outsiders
The second relationship that is analyzed, is the difference in effect between insiders and outsiders of the labor market. Again, is the SWB of individual who lives in country at time . is the
intercept, are the country fixed effects, are the year fixed effects, refers to deviation of
employment protection from the sample mean, and Ω are other factors that influence the SWB of the individuals. The variable X stands for the predictors of employment which are used to find the probability of employment in the second operationalisation of the model (in the first operationalisation they are left out of the model). These predictors of employment are taken into account as they may have an independent effect on SWB. In order to compare between insiders and outsiders of the labor market, the dummy variable INSIDERS is created.
24
4.2.2.1 Model 2.1 – Insiders as individuals who are full-time employed
A simple distinction between insiders and outsiders is whether they currently have a job. People who have a job can be considered insiders while people without a job can be considered outsiders. This paper uses full-time employment as the definition of employment.
However, this operationalisation has two limitations. Whether people are currently employed, is partly the result of employment protection regulations. In other words, current employment is a function of employment protection, so by using people’s current employment this operationalisation has a sample selection bias. The second difficulty focusses on the happiness of people who have a low chance of employment but are currently employed nonetheless and on people who have a high chance of a job, but are currently unemployed. When insiders are simply defined by people who currently have a job, these two groups are wrongly defined.
4.2.2.1 Model 2.2 – Insiders as individuals with a high probability of employment
A different way to determine whether an individual is an insider or outsider is by using their probability of employment. This probability is calculated for each individual by using the most important covariates of employment. The most important predictors of employment (X) used in this thesis are: country, year, gender, age categories, health, marital status, and education categories. These covariates are used to predict the probability of employment. In the table 4.2, the average probability of employment is shown for the different groups of the sample.
Table 4.2
Group Probability of employment (mean)
Full sample (N = 113329) .3975167
Full time employed (N = 44794) .4545508
Full time, part time and self-employed (N = 62250) .4402655
Unemployed (N = 6395) .4132958
Active in labor force (18<AGE<65) (N = 97381) .4185744
Not active in labor force (AGE<18 and AGE>65 (N = 15948) .2689348
26
Chapter 5 – Results
5.1.1 The overall effect of employment protection on the SWB of society (Model 1)
The first model compares the effect of employment protection between people within the labor force, who are affected by the legislation, and people outside the labor force, who are unaffected by the legislation. The model finds a positive and statistically significant result. If the average level of employment protection is raised by 50% (for instance from level 2 to level 3), the average level of SWB will increase with 4.5% (for instance from 7 to 7.3). Being active in the labor force increases the effect of employment protection on SWB with approximately 45% (.089/.195).
Being between the age of 18 and 65, has a negative effect on SWB in itself. Although this might seem counter-intuitive, it corresponds with the current knowledge concerning SWB and age as a U-shaped curve that is lowest in the middle (Economist 2010, Hawkes 2012, Frijters and Beatton 2012). Other significant factors on SWB are the four marital statuses Married (positive), Living together
as married (positive), Divorced (negative) and Separated (negative) and all health and education
dummies (besides EduMissing). These results are in line with earlier research, increasing the trust in the model. All and all, these results clearly confirm the first hypothesis that states that employment protection has a positive effect on the SWB of society.
Table 5.1.1
Model 1.1 Model 1.2 Model 1.3 Model 1.4
Intercept 7.009*** (.139) 5.963*** (.179) 6.513*** (.182) 6.669*** (.184)
Employment Protection .119*** (.021) .270*** (.027) .276*** (.027) .195*** (.031)
Active in labor force - - -.406*** (.025) -.580*** (.042)
Interaction
27
Very good health(subjective)
- .821*** (.024) .818*** (.234) .818*** (.024)
Good health (subjective) - .222*** (.022) .224*** (.022) .222*** (.022)
Fair health (subjective) - -.419*** (.023) -.416*** (.023) -.416*** (.023)
Bad health (subjective) - -1.277*** (.034) -1.278*** (.039) -1.273 *** (.034)
Very bad health (subjective)
- -1.910*** (.074) -1.921*** (.074) -1.916*** (.074)
Further controls No Yes Yes Yes
-2 Log likelihood 480862.873 431897.241 431624.455 431597.810
Notes: Standard errors in parentheses. * (**) (***) indicates significance at the 10% (5%) (1%). The sample consists of 113.501 individuals in 28 countries. Unreported control variables are eight education categories, sex, income, age, and age squared.
5.1.2 Insiders as individuals who are full-time employed (Model 2a)
The second model analyzes whether employment protection affects insiders differently than outsiders of the labor market. The operationalisation of the dummy variable INSIDERS is done in two different ways. At first, all individuals who are currently employed are considered to be insiders. By examining the interaction term, it can be deduced if employment protection has either a positive or a negative effect on insiders relative to outsiders. The model gives a positive and statistically significant result. If employment protection is increased, the SWB of insiders increases 3% more than the SWB outsiders.
Table 5.1.2
Model 2.1a Model 2.2a Model 2.3a Model 2.4a
28
Living together as married - .226** (.106) .222** (.106) .222** (.106) Divorced - -.340*** (.106) -.344*** (.106) -.345*** (.106) Separated - -.616*** (.113) -.620*** (.113) -.620*** (.113)Very good health (subjective)
- .821*** (.024) .820*** (.024) .819*** (.024)
Good health (subjective) - .222*** (.022) .221*** (.022) .220*** (.022)
Fair health (subjective) - -.419*** (.023) -.418***(.023) -.418*** (.023)
Bad health (subjective) - -1.277*** (.034) -1.273*** (.034) -1.272*** (.034)
Very bad health (subjective)
- -1.910*** (.074) -1.904*** (.074) -1.904*** (.074)
Further controls No Yes Yes Yes
-2 Log likelihood 480862.873 431897.241 431889.344 431883.229
Notes: Standard errors in parentheses. * (**) (***) indicates significance at the 10% (5%) (1%). The sample consists of 113.501 individuals in 28 countries. Unreported control variables are eight education categories, sex, income, age, and age squared.
This effect becomes stronger when intermediate groups, such as students or part-time employees are taken out of the sample. If people with full time employment are compared with people who are currently unemployed, the interaction term increases to .079*** (.024). Insiders benefit 8% more from employment protection than outsiders. The second hypothesis is thereby clearly confirmed.
5.1.3 Insiders as individuals with a high probability of employment (Model 2b)
The second method to distinguish between insiders and outsiders is to use the probability of employment. Instead of merely looking whether people have a job, an instrumental variable is created in which the probability of employment for each individual is calculated. This overcomes the sample selection bias that is created because employment is a function of employment protection, and defines outsiders and insiders more thoroughly.
29
conclusions done in the literature regarding SWB. This endorses the belief that the models are correct. These results again confirm the second hypothesis that employment protection affects insiders and outsiders differently and that insiders benefit more from employment protection than outsiders.Table 5.1.3
Model 2.1b Model 2.2b Model 2.3b Model 2.4b
Intercept 7.009*** (.139) 5.963*** (.179) 5.736*** (.220) 5.830*** (.221) Employment Protection .119*** (.021) .270*** (.027) .261*** (.028) .197*** (.035) Insiders - - .373* (.210) .143 (.225) Interaction Insiders & Employment protection - - - .144*** (.051) Control variables Married - .459*** (.103) .458*** (.103) .462*** (.103) Living together as married - .226** (.106) .226** (.106) .230** (.106) Divorced - -.340*** (.106) -.338*** (.106) -.336*** (.106) Separated - -.616*** (.113) -.615*** (.113) -.611*** (.113)
Very good health (subjective)
- .821*** (.024) .832*** (.025) .834*** (.025)
Good health (subjective) - .222*** (.022) .236*** (.023) .237*** (.023)
Fair health (subjective) - -.419*** (.023) -.397*** (.026) -.393*** (.026)
Bad health (subjective) - -1.277*** (.034) -1.223*** (.045) -1.209*** (.046)
Very bad health (subjective)
- -1.910*** (.074) -1.836*** (.085) -1.816*** (.085)
Further controls No Yes Yes Yes
-2 Log likelihood 480862.873 431897.241 431894.101 431886.048
30
5.2 Robustness checks
In order to check whether alternative explanations can refute the results obtained in section 5.1, several robustness checks are performed. To keep this section clear, only the interaction terms are published. Most of the variables used in the robustness checks are from the WVS, other sources are mentioned explicitly.
5.2.1 Broadening the definition of employment
Broadening the definition of employment is the first method to check the former results. While the first model defines employment as full time employment, it is also possible to define employment as being either full time, part time or self-employed. Changing the definition gives the following interaction terms for the second model.
Table 5.2.1
Model
Interaction term
Insiders * Employment protection
Only full time employment Full time, part time and self- employed
Model 2.4a .029** (.012) .008 (.011)
Model 2.4b .144*** (.051) .255*** (.047)
Model 2.4a gives an imprecisely estimated result. While the effect of employment protection on people who are employed is positive, the result is insignificant. Model 2.4b does give both a statistically and socially significant result. This result again confirms the second hypothesis that the effect of employment protection will be significantly lower for outsiders than for insiders.
5.2.2 Taking a different source for SWB
As is stated in section 2.2, Diener et al. (1999) claim that ‘SWB is a broad category of phenomena that
includes peoples’ emotional responses, domain satisfactions, and global judgment of life satisfaction’.
31
Although the effect in model 2.4a is slightly higher and the effect in model 2.4b is almost twice as low, the significant positive effects remain.Table 5.2.2
Model
Interaction term
Insiders * Employment protection
Life satisfaction Happiness
Model 2.4a .029** (.012) .032*** (.006)
Model 2.4b .144*** (.051) .081*** (.017)
5.2.3 Including additional control variables
Most economists agree that labor market institutions affect economic growth. This effect on economic growth could in theory affect the level of SWB in the long. In order to check this effect, economic growth is used as an additional control variable. Although the effect becomes slightly less strong, it is still socially and statistically very significant, thereby still supporting the two hypotheses.
32
Table 5.2.3
Model
Interaction term
Insiders * Employment protection
Without additional controls With GDP growth With GDP With GDP and GDP growth Model 2.4a .029** (.012) .089*** (.017) .090*** (.017) .090*** (.017) Model 2.4b .144*** (.051) .141*** (.051) .186*** (.051) .183*** (.051)
5.2.4 Taking a different source for employment protection
The OECD uses a very specific way to measure employment protection. By using labor freedom of the Economic Freedom Network, a broader definition of employment protection is used to check the results. As is show in table 5.2.4, both effects remain positive and significant. In fact, the effect even increases. This thereby again confirms the second hypothesis.
Table 5.2.4
Model
Interaction term
Insiders * Employment protection
OECD Economic Freedom
Network
Model 2.4a .029** (.012) .059*** (.015)
Model 2.4b .144*** (.051) .186*** (.046)
5.2.5 Using a different sample
33
employment protection regulations, is more likely to raise the labor costs above their productivity. This means that employment protection could become the binding constraint. This severely hampers their access to the labor market. Reproducing model 2b, this time by only using people who are currently unemployed, explicitly checks this reasoning.Table 5.2.5
Model 2.1b Model 2.2b Model 2.3b Model 2.4b
Intercept 5.966*** (.236) 5.704*** (.523) 6.564*** (.768) 7.212*** (.802) Employment Protection .142** (.086) .035 (.101) .046 (.103) -.293* (.160) Insiders - -1.340 (.864) -2.720*** (1.005) Interaction Insiders & Employment protection - - .732*** (.273) Control variables Married .314 (.429) .312 (.429) .324 (.428) Living together as married .204 .202 (.440) .212 (.440 Divorced -.502 (.438) -.508 (.438) -.511 (.438) Separated -.635 (.455) -.645 (.455) -.642 (.455)
Very good health (subjective)
1.264*** (.108) 1.223*** (.111) 1.237*** (.111)
Good health (subjective) .651*** (.096) .602*** (.100) .611*** (.100)
Fair health (subjective) -.056 (.104) -.134 (.116) -.123 (.116)
Bad health (subjective) -.937*** (.154) -1.135*** (.199) -1.109*** (.198)
Very bad health (subjective)
-1.083*** (.351) -1.349*** (.390) -1.272*** (.390)
Further controls No Yes Yes Yes
-2 Log likelihood 29052.188 25899.871 25897.506 25890.358
34
As the table shows, the effect of employment protection on SWB is extremely more positive for insiders than for outsiders. This effect is five times as strong as with the original sample. This confirms the above reasoning that unemployed outsiders are the ones who are hurt more by the negative effects of employment protection than people with ample opportunities on the job market.35
Chapter 6 – Conclusions
This thesis provides economists and politicians with a radically new way to assess the welfare implications of labor institutions. By using SWB as a trustworthy proxy for welfare, both the efficiency and security effects can be compared on an equal scale. This thesis thereby overcomes a problem that has severely hampered the research and the discussions regarding labor institutions. This novel method shows that employment protection has a positive effect on the welfare of a country and that insiders are affected more positively than outsiders. While the latter is in accordance with traditional economic thought, the former sheds a new light on the effect of employment protection on society.
By using global SWB data, all the factors that influence the utility of an agent are taken into account. When the effects of employment protection on SWB are compared between people who are in the labor force and people who are not, the first group benefits significantly more. Assuming that the second group is unaffected by employment policies, it is safe to state that employment protection has a positive effect on the welfare of a country.
The second goal of this thesis is to analyse whether traditional economic thought is correct in claiming that insiders benefit more from employment protection than outsiders. Two different methods are used to distinguish between the insiders and outsiders of the labor market. At first the effect of employment protection on SWB is compared between people who have a job and people who do not. The group individuals who have a job, benefits significantly more from employment protection than the group without a job. The second method creates an instrumental variable to estimate the probability of employment for each individual. This probability of employment is then used to differentiate between insiders and outsiders on the labor market. This model also shows a very significant result that confirms the difference in effect. Insiders benefit more from employment protection than outsiders.
36
Chapter 7 - Limitations
This research has several limitations. While the results of the different models correspond with the current literature, it is still possible that other uncontrolled factors influence SWB. Although the models control for quite a few of these factors, other (yet unknown) factors undoubtedly remain. These factors could influence the results and could change the effect of employment protection on SWB.
A second limitation is that the interaction effects measured in this research are relative. While it is established that the effect of employment protection on SWB is significantly different for different groups, the absolute effect of employment protection on SWB remains unknown. Although the first model shows that the average effect of employment protection on SWB is likely to be positive, the optimal level of employment protection for a country is undetermined. It is possible, or even likely, that some countries are already above their optimal level of employment protection. This implies that further increasing employment protection could have a negative effect on the welfare of that country.
A similar problem arises when the effect of employment protection on outsiders is considered. The model clearly states that insiders benefit more from employment protection than outsiders. However, that does not per se mean that this effect is positive. Because the first model indicates that the labor force benefits from labor protection, it is very likely that the effect of employment protection is positive for insiders. However, the same cannot be claimed for outsiders. Whether employment protection has a positive or negative absolute effect on their SWB remains unknown.
The unknown long term (indirect) effects are the third limitation. While there does not seem to exist a clear relation between yearly economic growth and SWB, a worsening in the competitiveness could lead to a weakening of public institutions, which can in turn affect SWB. Because this effect is indirect it is hard to estimate. However, that does not mean that the (potential) effect should be ignored.
