How different job search methods affect expected wages

How different job search methods affect expected wages.

Abstract

A lot of people think that knowing the right person will help getting a good job which pays a lot of money. Economic literature suggests that high-wage sectors use referrals to look for new employees, which supports what most people think. However, this paper is more interested in what people expect to earn at a potential new job. This paper develops a model which explains what people expect of their future earnings when making use of different job search methods.

JEL classification - J39

Author: Pim T. H. van Rooijen Student number: 10081984 Date: 9 July 2014

1 1. Introduction

It is generally accepted that personal contacts are fundamental for obtaining great jobs. The more high-placed friends someone has, the easier it is for this person to obtain a high position on the corporate ladder. Of course this is only true for the few lucky ones. It is hard for many job searchers to obtain a good, well-paying job. Job searchers can look for new jobs in numerous ways. However, all the methods job searchers have at their disposal can be assigned into one of two groups: the formal or informal information network. Job searchers who make use of the formal information network search by consulting: government services, private employment agencies, advertisements, union services and school or college placement bureaus. While the job searchers who make use of the informal

information network look for new jobs by asking around with friends and relatives, or the job searchers are referred by employees or employers (Rees, 1966).

This paper, is interested in what the job searcher expects when making use of these information networks. More specifically, this paper researches the question whether job searchers expect to earn different wages when making use of a formal or informal information network. For example, does someone who searches for a job and has high-placed friends expect a different wage than someone who searches for a job by reading advertisements. We expect to see a positive relationship between the informal search methods and expected wages. However, we have to construct a model to answer this research question. The model will consist of the various methods that job searchers have at their disposal including multiple control variables which can affect the expected wage. Furthermore, a regression will be performed on the completed model which allows us to look at whether there is a difference between the effect of a formal or informal network on expected wages.

The analysis in this paper links together social and economic literature and research done on the expected wage of job searchers. Social and economic literature stresses the importance of social networks and suggests that job searchers who make use of the informal network are hired in high-wage sectors. The model in this paper suggests that there is a difference between expected earnings between job searchers who make use of the formal or informal information network.

The paper is organized as follows. Section 2 present social and economic literature about the subject. Section 3 shows the descriptive statistics. Section 4 describes the model used in this paper and the regression results. Section 5 presents the discussion and section 6 concludes this paper.

2 2. Related literature

To answer the research question, some information has to be provided in order to obtain a firm understanding about the subject. This section will describe briefly how literature provides the necessary information in order to provide insight and awareness of different arguments. These arguments allow us to create an empirical model.

2.1 Information Networks in the Labor Market

In this paper we focus on the formal and informal information network in the labor market. The reason why the labor market is of interest is because in the labor market, people search for jobs, employers look for employees and wage rates are determined. Rees (1966) describes the formal or informal information network in more detail. Rees states that the formal information network in the labor market includes the state employment services, private fee-charging employment agencies, news-paper advertisements, union hiring halls, and school or college placement bureaus. The informal information network includes referrals from employees or employers, miscellaneous sources, and walk-ins or hiring at the gate.

2.2 Social Capital

A paper written by Alejandro Portes (1998) reviews the origins and definitions of social capital. In his paper, Portes reviews Bourdieu who gives the first systematic analysis of the term social capital. Social capital can be defined as Bourdieu asserts in his book The Forms of Capital (1985) as, "the aggregate of the actual or potential resources which are linked to possession of a durable network of more or less institutionalized relationships of mutual acquaintance or recognition" (as cited in Portes, 1998, p.3). The reason why this theoretical definition of social capital is important is explained by Portes. He argues that the concept of social capital, defined by Bourdieu, is instrumental and focuses on the benefits which result from participating in groups and from constructing sociability for the sole purpose of manufacturing this resource. Furthermore, Portes argues that the most common function of social capital is creating a network that goes beyond the immediate family. In addition, this network serves as an explanation of access to employment, position on the corporate ladder and the success of entrepreneurs. It is apparent that social capital is important for social interaction. The advantage of social capital and thus social interaction, is that social networks that people maintain already exist. These social networks don't need to be constructed which in turn implies that the cost is less than e.g., constructing a network by consulting private employment agencies (Granovetter, 2005).

3 2.3 Strength of Weak Ties

The connections between people can be described by interpersonal ties. Close friends are considered strong ties while acquaintances are considered weak ties. Interpersonal ties are important due to the fact that information flows to individuals through these ties. Granovetter (2005) states that more information is transmitted through weak ties than through strong ties. Granovetter argues that this is due to the fact that the information obtained from close friends overlaps with what we already know. On the other hand, acquaintances make contact with people we do not know and thus obtain more information. The result is that acquaintances may be better sources when trying to find a new job. Granovetter calls this the "strength of weak ties". (Granovetter, 1973, 2005).

Montgomery (1992) researches the implications of these weak-ties on labor market outcomes. He argues that the strength of weak ties hypothesis can be tested empirically by looking at the

reservation wage of job searchers. The reservation wage can be defined as the lowest wage rate at which a job searcher will accept a job. The reservation wage thus also corresponds to expected future earnings. Montgomery tests whether the reservation wage will rise as the proportion of weak ties increases while holding network size constant. He finds, by making use of Granovetter's argument that weak ties can be beneficial, that worker's reservation wage rises as the number of weak ties increases while holding network size constant. However, Montgomery states that this does not imply that making use of a weak tie results in higher expected wages. Making use of a weak tie results, according to Montgomery, in a relative lower expected wage. Montgomery points out that the expected wage increases as the number of job offers increases. He argues that a job searchers always receives a job offer through a weak tie, while a job searcher almost never receives a job offer through a strong tie. A job searcher who accepts the job offer through a weak tie, is more likely to have received only one job offer while a job searcher who accepts the job offer through a strong tie is likely to have received two job offers. The difference between the number of job offers results in the difference in expected wages.

2.4 The formal and informal information network

Social and economic literature discussed so far is positive about the effects of the informal information network. The general consensus, although premature, is that informal job search methods provide the job searcher more information than the formal counterpart. A paper written by Rees (1966) supports this view. Rees makes use of his own study of the Chicago labor market which was, at the time of this paper, still a work in progress. One observation made from this study is that in the white-collar occupations the informal information network is responsible for about half of hires. In the blue-collar occupations, informal hiring accounts for almost eighty percent of all hires. Most employers have, according to Rees, for different reasons a strong preference for the informal information network. Employee referrals are considered as the most important informal channel. One of the reasons why employers have a preference for referrals is because it provides screening. The increase in information

4 is obtained because employees are likely to refer people who look like themselves which provides screening. Furthermore employees have a reputation to maintain, their reputation is positively related to the quality of referrals. Another argument why employers favor informal channels is because it is simply costless. Of course it should be noted that some of the formal channels can be costless too.

Hiring costs and wages are inversely related. If an employer pays high wages, the quit rate by employees is reduced, but also high-quality workers will accept job offers. The inverse relationship is suggested by Stigler (1962). In addition Stigler points out that that wage rates and search costs are substitutes. Rees uses the relationship between wage rates and search costs to point out that when a low wage employer is looking for employees, he or she is forced to use high cost search methods, such as advertisements and private employment agencies.

2.5 Wage Premiums

Many workers think that personal contacts are important to obtain high salaries. Kugler (2003) states that referrals match high-paying jobs to workers who are well-connected, while formal job search methods match job searchers who are less-connected with jobs that pay less. In her paper, Kugler finds empirical evidence which suggests that the informal referrals channel lowers monitoring costs. The empirical model she uses shows that referred workers work harder, earn higher wages and have lower quit rates because referrals provide good jobs. In addition she finds empirical evidence which shows that at the industry-level, using referrals is correlated with industry wage premiums. It should be noted that the evidence Kruger finds suggests that referred workers earn higher wages because they are hired in high-wages sectors. It is not because they have some unobserved characteristic that can only be shown when a worker is referred.

Lin (1999) however, argues that informal channels are no better than other channels. He states that informal networks do not lead to better jobs and that informal networks are mostly used by the disadvantaged. The people who make use of the informal network rely on the statuses of the contacts. It implies that contact persons play a crucial role and that people who have great jobs do not rely on informal networks. Nevertheless, Lin finds mixed evidence when he researches the question whether the advantaged can apply directly to high-status positions.

5 3. Descriptive Analysis

This section provides basic descriptive information concerning the characteristics of workers who use different information networks. These information networks are either formal or informal. The tabulations are based on the DNB (Dutch Central Bank) household survey of 2012. The purpose of this survey is to study economic and psychological determinants of the saving behavior of households. A section of this survey contains a question about how respondents have been looking for jobs. Table I shows the questions that the respondents were asked. The respondents can choose from multiple information channels and each channel belongs either to a formal or informal information network. When a channel belongs, for example, to a formal network, the corresponding box in table 1 will be checked. Table 2 summarizes the characteristics of 129 respondents who were looking for a job. However, it must be noted that respondents are not restricted to one job search method, it is possible to make use of multiple formal and informal channels.

Table 1

Job search methods of respondents, DNB household survey 2012

Job search method Formal Channel Informal Channel

Answering advertisements 

Placing advertisements 

Reading advertisements 

Asking around with employers 

Asking friends and other relations 

Through a job center 

Through employment agency 

Table 2 shows that respondents use the formal information network to a greater extent than the informal network while searching for jobs. We see that there are some differences in the number of times that each channel is used. Answering and reading advertisements are by far the most used formal channels. On the other hand, placing advertisements is the least used formal channel, it has only been used two percent of the time. Asking friends and other relations is the most used informal channel. More respondents went looking for a job by asking their friends and other relations than by asking employers. If we analyze the difference between search methods used by men and women, we see that there are no big difference between genders. Men (women) made use of formal networks 74.3 (76.0) percent of the time while they made use of informal networks 25.7 (24.0) percent of the time.

6 Table 2

Descriptive statistics, DNB household survey 2012

Variable Formal Informal

Answering advertisements 70 (27.8)

Placing advertisements 5 (2.0)

Reading advertisements 69 (27.4)

Asking around with employers 24 (9.5)

Asking friends and other relations 39 (15.5)

Through a job center 19 (7.5)

Through employment agency 26 (10.3)

Total number of times channels used 189 (75.0) 63 (25.0)

Male 113 (74.3) 39 (25.7) Female 76 (76.0) 24 (24.0) Date of Birth 1945-1961 64 (76.2) 20 (23.8) 1962-1978 98 (74.8) 33 (25.2) 1979-1995 27 (73.0) 10 (27.0) Education High (≥ HAVO/VWO) 119 (78.3) 33 (21.7) Low (< HAVO/VWO) 70 (70.0) 30 (30.0)

Total Net Income

-500 - 25,000 61 (85.9) 10 (14.1)

25,001 - 50,000 30 (76.9) 9 (23.3)

50,001 - 130,000 11 (68.8) 5 (31.2)

N=129

The table presents the number of times that respondents, with various characteristics, used a formal or informal channel. Percentages are in parenthesis.

There are some minor differences when we compare the percentage gap in age between the formal and informal network. There are three different age groups, the percentage gap in age between the formal and informal network for the respondents who were born between 1945 and 1961 is 52.4 percent. The percentage gap in age for the respondents who were born between 1962 and 1978 is 49.6 percent, and when born between 1979 and 1995 results in a percentage gap of 46 percent.

The first noticeable difference arises when we look at the types of education respondents received. Education is split up into two groups, the high-group consists of respondents who finished at least higher general secondary education. The low-group consists of respondents who received

education which is considered lower than higher general secondary education. High (low) educated respondents used the formal network 78.3 (70.0) percent of the time while they made use of informal networks 21.7 (30.0) percent of the time. The second interesting difference arises when we look at the total net income received by respondents. The total net income is split up into three groups. We see that the percentage gap in net income between the formal and informal network for the respondents

7 who received between -500 and 25.000 euro is 71.8 percent. The percentage gap for respondents who received a net income between 25.001 and 50.000 euro is 53.6, and for respondents who received a net income between 50.001 and 130.000 euro this gap is 36.7 percent.

4. Regression Model and Results

In this section the job search methods of table 1 will be used to construct a function to answer the question: do job searchers expect to earn different wages when making use of a formal or informal information network? The analysis is based on the data from the DNB household survey 2012. Let Zi

represent a dummy variable that equals 1 if the respondent answers a job search question with "yes", and zero otherwise. The observation i's expected wage rate Wi is assumed to depend on Zi and some

observed control variables Xi plus an error εi. This will result in a linear regression that looks like

equation 1 where the parameters α and β will be estimated. Wi = βZi + αXi + εi (1)

Table 3 reports the results of performing OLS on equation 1.In the first column, all the independent dummy variables are included while all the control variables are excluded. In the second column, all variables are included. To prevent the dummy variable trap from occuring, one dummy variable has to be omitted from the regression. The most used job search method, based on table 2, will be omitted from the regression; therefore the dummy variable: answering advertisements will not be included.

Column1 shows the regression results only for the independent dummy variables. We can see that the only significant variable is the intercept, which measures the effect of omitted dummy variable on expected monthly wages. The independent dummy variables from the regression in column 1 basically explain nothing. However, if we include our control variables into the regression, some of the independent dummy variables become significant. Adding these control variables increases the fit of the model (increase from 0.018 to 0.77) while it decreases the mean square errors, consequently some variables which were insignificant now become significant . Column 2 shows the regression results including the control variables. The intercept measures the omitted dummy variable, but there is no statistical proof that the value of the intercept is different than zero. There are only two

significant dummy variables, these are: asking around with employers and through employment agency. Based on table column 2 in table 3, it implies that respondents who make use of the informal channel expect their wage to be 397.32 euro higher when compared to answering advertisements, and respondents who make use of the formal channel expect their wage to be 432.76 euro lower when compared to answering advertisements. The other job search dummy variables are not significant. However, the p-values of the dummy variables: placing advertisements, reading advertisements and through a job are very high, they are 0.97, 0.86 and 0.74 respectively. The high p-values imply that these variables are insignificant, but it may be the case that these three dummy variables are jointly significant. The hypothesis that these three variables are jointly significant and equal zero, results in a

8 Table 3

OLS regression estimates of the effect of formal/ informal search methods on expected monthly wage.

Variable (1) (2)

Intercept 1739.62*** -2293.37

(132.24) (12660.05)

Placing advertisements (1=Yes) 194.97 -10.62

(-208.39) (269.39)

Reading advertisements (1=Yes) 96.40 23.34

(152.49) (127.31)

Asking around with employers (1=Yes) 89.69 397.32**

(203.69) (178.65)

Asking friends and other relations (1=Yes) -14.39 215.30

(167.16) (142.18)

Through a job center (1=Yes) -208.61 74.89

(234.15) (226.99

Through employment agency (1=Yes) -95.41 -432.76**

(208.13) (206.76)

Date of birth - 1.03

(6.45)

Male (1=Yes) - 36.76

(143.16)

Main earner in the household (1=Yes) - -98.53

(162.81)

High education (1=Yes) - 186.51

(150.93)

Monthly net income - 0.29***

(0.045)

Expected monthly hours work at new job - 10.50***

(2.15)

Current fulltime job (1=Yes) - 163.54

(135.76)

R2 0.018 0.77

N=129

Standard errors are between parenthesis. *: p < 0.10, ** : p<0.05, ***:p<0.01

p-value of 0.99 (F=0.04). The null hypothesis that these three variables are jointly significant and equal to zero can't be rejected. Therefore we can assume that these three variables are not significant. Furthermore, the result that the searching for a job by asking friends and other relations is

insignificant, was not expected. The next section shall describe these results more extensively. There are some interesting results concerning the control variables. The only significant control variables in the regression of column 1 are: monthly net income and expected monthly hours work at a new job. The coefficient of the variable monthly net income is 0.29, and the coefficient of

9 expected monthly hours work at the new job is 10.50. It must be said that the effect of an increase in monthly net income is small, it seems unlikely that every euro increase in monthly net income is translated into an increase of expected monthly wage of 0.29 euro. Respondents expect to earn 10.50 euro more whenever they expect to work one additional hour per month. Furthermore, the fact that most control variables are insignificant is also surprising. For example, one would expect to earn more if he or she is the main earner, is older or has obtained higher education. These variables are far from significant as their coefficients are relatively low compared to their standard errors. A possible

explanation is that this model suffers from multicollinearity. We can test for multicollinearity by using the variance inflation factor (VIF). This factor measures the inflation of standard errors due to

multicollinearity. Low levels of VIF are desired as higher values indicate that the standard errors are larger than would otherwise be the case because of correlations between variables. O'Brien (2007) states that the value of 10 associated with VIF is a sign of serious multicollinearity. A VIF of 10 implies that the standard errors are larger by a factor of 10 than if there were no correlations between the independent variables. Variables with a VIF higher than 10 should be removed. Table 4 shows the results.

Table 4

Variance Inflation Factors. High to low.

Variable VIF

Main earner in the household 1.91

Monthly net income 1.86

Through employment agency 1.80

High education 1.77

Male 1.58

Expected monthly hours work at new job 1.49

Current fulltime job 1.44

Through a job center 1.40

Date of birth 1.39

Placing advertisements 1.36

Asking employers 1.34

Reading advertisements 1.28

Asking friends and other relations 1.10

Mean 1.52

As we can see from table 4, there are no values that are larger than 10. Therefore we assume that this model does not suffer from severe multicollinearity.

10 5. Discussion

This section will link the results obtained in table 3 with economic and social literature. Some variables which were expected to be significant, turned out to be insignificant. The purpose of this discussion is to look whether literature can provide additional information about the findings obtained from the regression.

First of all, it should be noted that it is possible that the respondents who make use of the formal information network are overrepresented. Rees (1966) states that for the white-collar workers, referrals account for almost 50 percent of all hires while for the blue-collar occupations, informal hiring accounts for almost eighty percent. In our sample, people are of course not yet hired, but the informal information channels were only used 25 percent of the time. Respondents used the formal information network 189 times, while the formal information network was used 63 times. Therefore it may be possible that respondents who make use of the formal information network are

overrepresented. One would expect to make more use of the informal information network because most of the hires occur through referrals. Our sample may therefore suffer from selection bias.

We now start the discussion by looking at the six job search dummy variables in column 2 from table 3. The job search variables: placing advertisements, reading advertisements and through a job center are not significant. First of all, the placing advertisements channel was only used five times in total. This channel may have only been used five times because it it simple very costly for the job searcher. For that reason, we may not have enough information about this formal channel to see any change in expected monthly wage when compared to answering advertisements. Furthermore, the expected monthly wage doesn't change when respondents search by reading advertisements and by consulting job centers. Although these formal channels were used more often, they remain

insignificant. Advertisements in general and job centers may be very costly for both employers and job searchers. According to Stigler (1962), we could expect a negative relationship between earnings and using advertisements or job centers. This negative relationship can't be confirmed by making use of these dummy variables. However, it is possible that these three job search variables are jointly significant. But as we have shown, the null hypothesis that these three variables are jointly significant and equal to zero can't be rejected at the p-value of 0.99. It implies that we can't find any statistical evidence which shows that the expected monthly wage is different when looking for a job by placing and reading advertisements and looking for a job through a job center when compared to reading advertisements.

A surprising results is that the variable which describes job search through friends and other relations is insignificant. Economic and social literature suggests that this information channel is one of the most important channels as people can get referred by friends and other relations. Furthermore Kugler (2003) finds empirical evidence that referred workers earn higher wages because they are hired in high-wage sectors. When workers receive higher wages, it seems rational to think that workers

11 would expect to earn higher wages at a new job. However, there may be a reason why the dummy variable is insignificant. Lin (1999) argues that informal networks are no better than other networks and that people who make use of informal networks rely on the statuses of contacts. The contact's role plays, according to Lin, a crucial role in obtaining a good job. It is possible that the friends and relatives of the respondents who were looking for jobs, just weren't that good of a contact and that consequently our sample suffers from this selection effect. The lack of knowing the right person would leave the respondents with potential not-such-great jobs, which makes this informal channel not better than reading advertisements.

Two of the six job search dummy variables are significant. These dummy variables which are significant are: asking around with employers and through employment agency. Montgomery (1992) researched the effect of acquaintances, rather than close friends and relatives, on labor market outcomes. The result from table 3 suggests that when respondents search for a job by asking around with employers, expected monthly wage increases by 397.32 euro when compared to reading

advertisements. The empirical model Montgomery uses looks at whether the expected future earnings increase as the proportion of acquaintances rises. He finds that the worker's expected future earnings rise as the number of weak ties in the worker's network increases. However, according to

Montgomery, making use of a weak tie lowers the expected wages. We find something different, our model suggests that searching for a job by asking around with employers increases the expected monthly wage. Nevertheless, our model looks only at employers and acquaintances can't be described by just employers. Furthermore, Montgomery comes to the conclusion that expected future earnings rise as the proportion of acquaintances rises while controlling for network size. Unfortunately controlling for network size in our model was not possible because this information could not be obtained from the DNB household survey. A possible result is that our model may suffer from omitted variable bias. It is possible that the dummy variable: asking around with employers is overestimated because of a positive correlation between network size and this predictor variable. By calculating the correlation between a dummy variable and a continuous variable this omitted variable bias can be tested.

The second significant job search dummy variable is: through employment agency, which has a negative coefficient of 432.76. As Stigler (1962) suggested in his paper, wage rates and search costs are substitutes. Rees (1966) used this relationship to point out that when a low wage employer is looking for employees, high cost search methods, such as private employment agencies must be used. The negative coefficient suggests that when respondents search for a job through employment

agencies, expected monthly wage decreases by 432.76 euro when compared to reading advertisements. However, given the possibility that respondents who make use of the formal information channels are overrepresented, we can't justify this conclusion. Furthermore, most of the control variables are insignificant which is an unexpected result. As this model doesn't suffer from severe multicollinearity it would be a good idea to perform this research again, this time with a larger sample size which is also

12 randomized, and with more control variables. Therefore, the model won't suffer from selection bias and the control variables have a better chance to become significant.

6. Conclusion

This paper investigates whether job searchers expected different wages when making use of a formal or informal network. One estimate suggests that job searchers who ask around with employers expect higher monthly wages when compared to searching by reading advertisements. However, further research should indicate whether this result is justified as this estimate may suffer from omitted variable bias. Another estimate suggests that job searchers who make use of temporary employment agencies expect lower wages than searching reading advertisements. Rees (1966) points out that when low wage employers are looking for employees, they are restricted to high cost search methods like employment agencies. Therefore, job searchers may expect to earn less when compared to reading advertisements. However, because this model may suffer from a non-randomized sample we can't be certain that the effect of these significant variables will be as described in table 3.

Furthermore four estimates suggest that there is no statistical evidence concerning the

difference in expected monthly wage between formal and informal channels.These four estimates are: placing advertisements, reading advertisements, using a job center and asking friends and other relations. The dummy variable: placing advertisements was only used 5 times, this was probably not enough to become a significant result in our model. Respondents used the dummy variables: reading advertisements and through a job center more. However, these variables too were not significant.

The insignificance of the informal information network: asking friends and other relations was rather interesting as economic theory suggests that this is an important informal channel. However, this paper makes use of a relatively small sample size. It is possible that if we run this regression again while this time we have our hands on a relatively large and randomized sample, these dummy

variables become significant.

Given all this information, we can conclude that we do not know whether job searchers expect to earn different wages when making use of a formal or informal information network. Two variables which are significant shown that there is some difference in wages. However, due to the problems of omitted variable bias and selection bias, we can't be certain that this is right. Therefore, the model designed in this paper should be run again, but this time with a larger sample size, which is also randomized.

13 Reference List

Granovetter, M. (1973). The Strength of Weak Ties. American Journal of Sociology, pp. 1360-1380. Granovetter, M. (2005). The Impact of Social Structure on Economic Outcomes. The Journal of Economic Perspectives, pp. 33-50.

Kugler, A. (2003). Employee Referrals and Efficiency Wages. Elsevier, pp. 531-556.

Lin, N. (1999). Social Networks and Status Attainment, Annual Review of Sociology, pp. 467-487. Montgomery, J. (1992). Job Search and Network Composition: Implications of the Strength-Of-Weak-Ties Hypothesis, American Sociological Review, pp. 586-596.

O'Brien, R. (2007). A Caution Regarding Rules of Thumb for Variance Inflation Factors, Quality & Quantity, pp. 673-690.

Portes, A. (1998). Social Capital: Its Origins and Applications in Modern Sociology, Annual Review of Sociology., pp. 1-22.

Rees, A. (1966). Information Networks in Labor Markets, The American Economic Review, pp. 559-566.