ARTICLE IN PRESS

0261-5177/$ - se

doi:10.1016/j.to

�Tel.: +47 61

E-mail addr

Tourism Management 29 (2008) 514–524

www.elsevier.com/locate/tourman

Earnings differentiation in the tourism industry: Gender, human capitaland socio-demographic effects

Christer Thrane�

Department of Social Sciences, Lillehammer University College, 2626 Lillehammer, Norway

Received 10 November 2006; received in revised form 16 May 2007; accepted 22 May 2007

Abstract

This paper draws attention to a somewhat neglected field within tourism research, namely the production side of tourism. Set within a

human capital framework, the purpose of the study is to examine if and how the variables education, work experience and the socio-

demographic variables parenthood and marriage affect Norwegian tourism employees’ annual wages in the 1994–2002 period.

Furthermore, the study specifically examines how the variables gender and time affect these wages. Among several findings, four are of

major importance. (1) Male tourism employees receive about 20% higher wages annually than their female counterparts, ceteris paribus.

(2) There is a concave (i.e. inverted U) relationship between work experience and annual wages for both female and male employees, but

this pattern is much less distinct for females. (3) Both parenthood and marriage affect wages, but their effects differ between male and

female employees. (4) All tourism employees experience a positive annual wage development in the period studied.

r 2007 Elsevier Ltd. All rights reserved.

Keywords: Wages; Earnings; Human capital; Tourism; Gender; Random effects

1. Introduction

Tourism research is for obvious reasons abundant withpapers describing, and sometimes explaining, the where-abouts and doings of tourists. For this reason, we know alot about tourism consumption and its determinants andconsequences. This is not the case for tourism consump-tion’s flip side: tourism production. That is, not nearly asmuch is known about the employees doing the tourism jobsand thereby providing the tourism services. The aim of thepresent paper is to make a contribution to the supply sideof tourism research by examining the differentiation intourism employees’ earnings. The study’s theoreticalrationale lies in the human capital framework as set forthin labor economics and sociology. The empirical analyses’point of departure is the examination of how the keyhuman capital variables education and years of workexperience, as well as certain socio-demographic variables,explain variation in tourism employees’ annual wages.Special attention is also given to the gender wage gap and

e front matter r 2007 Elsevier Ltd. All rights reserved.

urman.2007.05.017

28 82 47; fax: +47 61 28 81 70.

ess: Christer.Thrane@hil.no

how this is brought about. The empirical analyses arebased on longitudinal data from administrative registers inNorway and pertain to more than 100,000 tourismemployees covering the period from 1994 to 2002.The study is organized as follows. The next section

presents the general human capital model of earningsdifferentiation, some important insights from the litera-ture on gender differences in earnings, and the priortourism studies dealing specifically with remuneration.With Section 2 serving as a background, the present study’sspecific empirical goals are then spelled out in more detailin Section 3. Section 4 describes the data, variables andregression models, whereas Section 5 presents and discussesthe results. Section 6 is a short summary and a conclusion.

2. Earnings differentiation as a function of human capital,

gender and socio-demographic factors

2.1. Human capital and earnings

Human capital refers to skills and knowledge that areacquired and valued mainly for their productive potential

ARTICLE IN PRESSC. Thrane / Tourism Management 29 (2008) 514–524 515

in an economic sense (Becker, 1962). The two mostimportant human capital variables are (years of) educationand (amount of) work experience (Becker, 1975; Mincer,1958). In its core, the human capital model posits thatindividuals with more human capital are more productiveand, hence, more attractive from the viewpoint of potentialemployers. Therefore, since employers are willing to pay apremium for productive employees, more education ormore work experience is expected to be associated withhigher earnings. In this way the human capital explanationof earnings differentiation (or inequality) is that peoplereceive the payment they ‘‘deserve’’ according to theirhuman capital investments. Within this framework a greatnumber of studies in labor economics and sociology haveaimed to disentangle the various ways in which earnings orwage differentiation arises. The baseline empirical specifi-cation in this respect has been the so-called Mincerregression model named after its developer, Jacob Mincer(Mincer, 1974; Mincer & Polachek, 1974).1 In its mostbasic form, this model states that an individual worker’searnings can be expressed as

lnwi ¼ a0 þ b1Ei þ b2 Expi þ b3 Exp2i þ �i, (1)

where lnwi is the natural logarithm of earnings ofindividual i, Ei denotes years of completed schooling, Expi

denotes work experience in years, ei is a stochastic errorterm and a0, b1, b2 and b3 are the parameters to beestimated. The quadratic coefficient for work experienceindicates that earnings tend to rise steeply during theyounger years (where the human capital investments are attheir most intense), plateau out during midlife, andeventually deteriorate at older ages. In their latest world-wide review of the returns to investments in education,Psacharopoulos and Patrinos (2004) estimate the mean rateof return to another year of schooling to be around 10%.Yet they also find substantial variation in these returns andthat they tend to be highest in low-income and middle-income countries. No such point estimate can be identifiedfor the effect of work experience, but an overwhelmingnumber of studies have documented the inverted U-shapebetween experience and earnings (e.g. Murphy & Welch,1990; Robinson, 2003).

2.2. Human capital, gender and earnings

No matter which western country we focus on or howearnings are measured, whether it be hourly, monthly orannually, it is a well-established fact that men tend toreceive higher earnings than women on average (Almond &Rubery, 1998; Blau & Kahn, 1992, 1996, 2006). In otherwords, when the variable gender (coded 1 for males and 0for females) has been added to the Mincer regression modelin past research, its coefficient has tended to be positive and

1Indeed, Lazear (2000, p. 118) has claimed that Mincer’s regression

model ‘‘y has probably been estimated more times, by more researchers,

and on more different data sets than any other relation in economics.’’

significant. Accounting for this gender gap has been afrequent topic of research in both labor economics andsociology (Bayard Hellerstein, Neumark, & Troske, 2003;Datta Gupta & Rothstein, 2005; Meyersson Milgrom &Petersen, 2006; Petersen & Morgan, 1995). The a priorihuman capital explanation of this gender earnings gapwould be that women possess less human capital than men.In terms of education, this explanation does not appear tofit well with the available evidence; i.e. ‘‘y genderdifferences in education levels have never explained a largeportion of the overall gender pay gapy’’ (Blau & Kahn,2006, p. 42). Also, in many Western countries there areonly very small differences between the genders in terms ofeducational attainment. In contrast, due to maternityleave, marriage, and a traditional division of labor withinthe family (that also explains women’s higher propensityfor working part-time), women accumulate market-valuedwork experience more slowly than men (Blau & Kahn,2006; Mincer & Polachek, 1974; Polachek, 2006). Genderdifferences in work experience, thus, partly account for theoverall gender gap in earnings.In many of the Mincer regressions to date, separate

analyses for males and females have been carried out inaddition to the analysis pertaining to the pooled sample ofmen and women. This practice stems from the empiricalexperience that human capital variables and socio-demographic variables also have gender-specific effects onearnings. For example, according to Dougherty (2005), thereturns to education tend to be higher among women thanamong men. Furthermore, male employees tend to have amore distinct concave experience–earnings profile thanfemale employees, providing support for the so-called life-cycle human capital model (Ben-Porath, 1967). Theprediction from this theory, which has been tended to besupported empirically, is that men’s earnings shouldrise in a steeper fashion than women’s during the middleand late twenties and young thirties, due to women’sgreater involvement with childrearing and householdresponsibilities (e.g. Polachek, 2006).

2.3. Parenthood and marital status and earnings

Recent research has also examined how certain socio-demographic variables influence earnings. In particular, theeffects of parenthood and marital status on wages havebeen scrutinized (Budig & England, 2001; Høgsnes, Penner,& Petersen, 2006; Korenman & Neumark, 1992; Lundberg& Rose, 2000; Waldfogel, 1998). For both of these socio-demographic variables, the effects on earnings appear todiffer between men and women. Presence (or number) ofchildren in the family seems to be negatively associatedwith women’s earnings, all else equal, whereas no associa-tion or a positive association tends to be the case for men(Høgsnes et al., 2006; Lundberg & Rose, 2000; Waldfogel,1998). This effect has also been labeled the motherhoodpenalty (Anderson, Binder, & Krause, 2002). A number ofstudies have in a similar fashion documented the so-called

ARTICLE IN PRESSC. Thrane / Tourism Management 29 (2008) 514–524516

marriage premium; i.e. that married people typically earn ahigher wage rate than their unmarried counterparts, all elseequal. Whether a causal interpretation of this association islegitimate or not, however, is still a controversial issue(Loh, 1996). Also, since most of this research pertains tomen only, it is difficult to assess if this effect also holds forwomen. That said, Dougherty’s (2006) and Høgsnes et al.’s(2006) findings are suggestive of a somewhat smallermarriage premium among women.

2.4. Prior studies of remuneration in the tourism industry

Tourism scholars have repeatedly made the observationthat the tourism industry pays its employees poorly,relative to other industries (Church & Frost, 2004; Lee &Kang, 1998; Lucas, 2004; Riley, Ladkin, & Szivas, 2002;Riley & Szivas, 2003). Despite this, however, only a fewstudies have dealt empirically with earnings differentiationper se. Sturman (2001, p. 73) concluded for the US that‘‘y a one-point increase in human capital was associatedwith an average annual pay increase of $11,285 in otherindustries, but the same one-point increase garnered onlyan average increase of $9,155 in hospitality jobs.’’ Banulsand Rodrıguez’s (2005) recent study from the Spanishtourism industry is framed within the Becker/Mincerhuman capital tradition. Indeed, the first regression modelthey present exactly mirrors the Mincer specification in (1)above. Their results appear to be in line with most previousstudies, although the rate of return to another year ofschooling for tourism employees is only 3.3%. Further-more, once schooling measured in years is substituted withlevels of education (represented by a set of dummies), thenonlinearity of the education–earnings relation becomesevident. This is also in agreement with past research (e.g.Heckman, Lochner, & Todd, 2003). Contrary to Dough-erty’s (2005) general observation of a higher return toeducation for women, however, they find higher educa-tional returns for men. In keeping with previous research isalso the reported concave experience–earnings relationship.A possible limitation in Banuls and Rodrıguez’s (2005)study is that it does not incorporate gender among theindependent variables, although the results are presentedfor both the pooled sample and for males and femalesseparately. Hence, it is difficult to get an impression of thepotential gender differences in earnings based on theirresults.2

Marchante, Ortega, and Pagan (2005) also use theMincer model as starting point on their data fromAndalusia. Their aim is to relate monthly wages to thepossible mismatch between the required schooling for thehospitality employees’ present job and the actual schoolingthey possess. This research question differs from Banulsand Rodrıguez’s (2005) as well as from the literature cited

2It should be mentioned that the constants in the regression models for

females are generally smaller than the similar constants for men,

something that is indicative of lower female earnings.

above, but some of their reported results are nevertheless ofinterest in the present context. Their estimated return toanother year of schooling is quite small; i.e. less than 2%.In contrast, their reporting of a curvilinear effect of workexperience and a 10% male monthly wage advantage,ceteris paribus, are in line with most previous research.Santos and Varejao (2006) also focus on quantifying thegender differences in tourism employees’ earnings, usingmatched employer–employee data from Portugal. Amongtheir descriptive findings is that the hourly earnings formen are .40 Euros (or .16 log points) higher than they arefor women. Estimating Mincer-inspired wage models formen and women separately, they also find men’s experi-ence–earnings profile more distinctly concave thanwomen’s. Santos and Varejao (2006) also carry out thedecomposition technique suggested by Blinder (1973) andOaxaca (1973). This procedure identifies (i) the part of theoverall gender earnings gap that is due to the genderdifferences in the means of the independent variables(i.e. the endowment effect) and (ii) the part of the overallgender earnings gap that is due to gender differences inreturns to the regression coefficients (i.e. the discriminationeffect). The results indicate that gender difference in themeans of the independent variables (i.e. differences inendowments) account for 45% of the overall gender gap inearnings, whereas the remaining 55% is attributable todiscrimination. It is also suggested that the discriminationportion of the gender gap has more to do with men beingpaid above the non-discriminating wage structure thanwith women being paid below this structure.

3. Purpose of study

Studies examining tourism employees’ earnings differ-entiation are not abundant. This being the state of affairs,the general aim of the present study is to make acontribution to this literature. On a more specific note,this study extends previous work in three importantrespects. First, it explicitly examines and reports the effectsof both gender and human capital variables on tourismemployees’ earnings. Second, the study also introduces thesocio-demographic variables parenthood and maritalstatus as independent variables in the earnings models.Third, since the analyses are based on longitudinal data,the study also scrutinizes the time-dimension in earningsdifferentiation. The model serving as background for theempirical analyses, which draws heavily upon the precedingreview of the literature, can be expressed as

W ¼ f ðG;HC;SD;T ;CÞ, (2)

where W refers to annual wage, G represents the variablegender, HC represents human capital variables (i.e.education and work experience), SD represents socio-demographic variables (i.e. parenthood and marital status),T represents time (i.e. a set of observation year dummies)and C represents a vector of control variables.

ARTICLE IN PRESSC. Thrane / Tourism Management 29 (2008) 514–524 517

Based first and foremost on the results of preliminaryanalyses, two types of tourism employees are analyzedseparately in the study: tourism employees (i.e. personsemployed within tour operation, at travel agencies, attourist information offices, at camping sites, and at lodgingfacilities other than hotels) and hotel employees.3 Threesets of analyses are performed on each of the two samplesof employees; one on the pooled sample (including genderas an independent variable), one on the male sample andone on the female sample.

4. Data, variables and regression models

4.1. Data

The data come from the population database FD-Trygd(FDT) maintained by Statistics Norway and the Norwe-gian Social Science Data Service (NSD). NSD provided thedata, but they are not responsible for the present analysis.FDT links longitudinal individual level data from severaladministrative registers in Norway and provide extensiveinformation about, among other things, annual wage,employment history, educational background and socio-demographic characteristics. The two samples analyzed inthe study comprise all Norwegians between 18 and 67 yearswho during the period from 1994 to 2002 were employedeither in tourism enterprises (henceforth tourism employ-ees) or in hotels (henceforth hotel employees). The analysespertain to 22,032 tourism employees and 86,252 hotelemployees. However, since these individuals contribute tothe panel data with up to a maximum of 9 observations(one per year), the analyses are based on 133,653 person-years for tourism employees and 500,740 person-years forhotel employees. The two samples are unbalanced but theaverage contributions from each employee to the totalnumber of observations are 6.1 in the tourism sample and5.8 in the hotel sample.

4.2. Variables

Earnings are measured as tourism/hotel employees’annual wage, adjusted and standardized to 2002 Norwe-gian Crowns using the CPI.4 During 2002, 100,000 NOKequaled about US$12,500 or h14; 400 on average. The mainexplanatory variables are gender, education, work experi-ence, parenthood, marital status, observation year andworking hours. The control variables include immigrantstatus, number of family members, region of familyresidence, tourism sub-sector/hotel classification, parents’educational level and a dummy variable indicating if the

3More specifically, tourism or hotel employment refers to the following

ISIC or NACE codes: 63110, 63210, 63220, and 71911 (ISIC); 55101,

55102, 55110, 55120, 55210, 55220, 55230, 63301, 63302, 63303, 63304,

63305, and 63309 (NACE).4Annual wages below 10,000 NOK and above the 99th percentage of the

wage distribution were discarded. The reported analyses, thus, pertain to

more than 98% of the samples’ wage distributions.

employee completed his/her schooling the same year as theannual wage was reported. Two measures of education willbe used in the analyses: level of education and years ofschooling. Following Høgsnes et al. (2006) the variablework experience is defined as age minus 16 minus years ofschooling above 16 for men. For women the similarmeasure is age minus 16 minus years of schooling above 16minus number of children below 18 years. The reason formaking the latter qualification is the critique that tradi-tional proxy measures of work experience do not take yearsof maternity leave into account, and thereby overestimateswomen’s work experience (Hum & Simpson, 2004). Theconsequential subtraction of one year of work experienceper child might seem arbitrary at first sight, but in recentyears as much as 80% of employed women in Norway havetaken the maximum 12 months of maternity leave(Danielsen & Lappgaard, 2003). The variable parenthoodis measured as the number of children aged less than 11years in the family, whereas marital status contrastsmarried and unmarried employees. Finally, three dummiesrepresent working hours: 4–19 h per week (i.e. the referencecategory), 20–29 h per week and 30 or more hours perweek. Descriptive statistics for these variables for bothsamples are presented in Table 1. As can readily be seen,women outnumber men in both the tourism sample and inthe hotel sample; nearly 70% of the employees are women.This is in line with the findings of Lucas (2004) and Santosand Varejao (2006) and underscores the notion of tourismas ‘‘female jobs.’’

4.3. Regression models

The starting point for the multivariate analyses is theexamination of the cross-sectional and time-dimensionaldifferences in tourism employees’ annual wages. Theanalyses then proceed with assessing the similar differencesfor the hotel employees. Since the data have a panel orlongitudinal structure and a number of the independentvariables are time-invariant, so-called random effectsregression analysis is the chosen estimation technique(cf. Wooldridge, 2002). Following convention, the naturallogarithm of annual wage in NOK is the dependentvariable in the regressions.

5. Results

5.1. Results A: tourism employees

The left hand side of Table 1 suggests that tourismemployees on average earned about 204,000 NOK annuallyin the 1994–2002 period. Expressed in logs (i.e. naturallogarithms) this yearly wage corresponds to 11.975. Themodal educational category among the tourism employeesis post secondary or tertiary education (41%), whereas onlyabout 2% are college or university graduates (i.e. universitylevel, higher). Furthermore, 26% have some collegeor university background (i.e. university level, lower). The

ARTICLE IN PRESS

Table 1

Sample means and standard deviations (in parentheses) for main study variables for tourism and hotel employees in Norway, 1994–2002

Variables Tourism employees Hotel employees

All Men Women All Men Women

Annual wagea 203,522 244,867 184,515 172,383 209,919 155,572

(118,402) (134,623) (104,812) (104,596) (119,616) (92,297)

Log annual wage 11.975 12.168 11.886 11.807 12.010 11.716

(.817) (.823) (.799) (.800) (.810) (.780)

Education

Lower secondary schooling or belowb .067 .095 .055 .108 .087 .118

(.251) (.293) (.227) (.311) (.282) (.322)

Upper secondary level .237 .271 .222 .299 .301 .298

(.425) (.444) (.415) (.458) (.459) (.457)

Post secondary or tertiary level .411 .356 .436 .381 .410 .368

(.492) (.479) (.496) (.486) (.492) (.482)

University level, lower .262 .245 .270 .195 .179 .202

(.440) (.430) (.444) (.396) (.383) (.402)

University level, higher .023 .034 .018 .017 .023 .014

(.50) (.181) (.132) (.128) (.149) (.118)

Years of schooling 12.89 12.72 12.98 12.42 12.38 12.43

(2.64) (2.80) (2.56) (2.73) (2.88) (2.65)

Work experience (in years) 14.56 17.74 13.10 12.46 13.09 12.18

(11.98) (12.53) (11.42) (11.40) (10.96) (11.59)

Work experience-squared 355.60 471.87 302.16 285.31 291.29 282.64

(495.91) (548.20) (.460.17) (464.34) (446.13) (472.10)

Number of childreno11 years .423 .397 .436 .427 .358 .457

(.753) (.760) (.750) (.761) (.737) (.771)

Marital status (married ¼ 1) .358 .401 .338 .286 .266 .295

(.480) (.490) (.473) (.452) (.442) (.456)

Working hours per week

4–19 hb .121 .106 .127 .255 .241 .261

(.326) (.308) (.333) (.436) (.428) (.439)

20–29 h .074 .059 .081 .147 .095 .170

(.262) (.236) (.273) (.353) (.294) (.376)

30 or more hours .805 .834 .791 .598 .664 .569

(.396) (.371) (.406) (.490) (.472) (.495)

Gender (men ¼ 1; women ¼ 0) .314 .309

(.464) (.462)

N (observations; i.e. person-years) 133,653 42,092 91,561 500,740 154,889 345,851

N (persons) 22,032 6873 15,159 86,252 26,938 59,314

aAnnual wage is inflated to 2002 Norwegian Crowns (NOK) using the consumer price index (CPI).bReference category in the regressions in Table 2.

C. Thrane / Tourism Management 29 (2008) 514–524518

average of years of schooling is 12.89 years, whereas themean of years of work experience is 14.56 years. The mean ofnumber of children aged less than eleven years in the familyis .42. Of the tourism employees 36% were married at somepoint in time during the period studied, and 81% workedmore than 30 hours during a normal week. Table 1 alsoreveals that male and female tourism employees have distinctprofiles, both in terms of annual wages and in thedistribution on the independent variables. As regardsthe former, men received almost 33% higher wages onaverage in the period studied (12.168�11.886 ¼ .282;exp (.282)�1 ¼ .326). (When log coefficients are larger than.15, they must be adjusted by taking the antilog of thecoefficient minus 1 in order to express differences inpercentages; see Halvorsen & Palmquist, 1980.) In otherwords, women on average made about 75% of their malecounterpart’s wages on a yearly basis in the period taken as

a whole (1/(1+exp (.282)�1) ¼ .754). Relative to women,more men have only lower or upper secondary schooling. Incontrast, the reverse pattern holds for post secondaryschooling and for some college/university training(i.e. university level, lower). At the graduate level, however,men again outnumber women. A striking feature concernsthe difference in years of work experience. In this regard,men on average have nearly 5 more years than women(17.74�13.10 ¼ 4.64). For the remaining independent vari-ables (number of children, marital status, working hours) thedifferences between men and women are rather small,although it could be mentioned that more men are marriedand that the percentage working full-time is slightly higheramong men (83% versus 79%).Table 2 presents the results from the regression analyses.

Among the tourism employees the gender coefficient is.181. That is, men received on average 20% higher annual

ARTICLE IN PRESS

Table 2

Annual wage (logged) by independent variables for tourism and hotel employees in Norway, 1994–2002. Random effects regression analysis

Variables Tourism employees Hotel employees

All Men Women All Men Women

Gender (men ¼ 1; women ¼ 0) .181 .190

(.008) (.004)

Educationa

Upper secondary level .106 .072 .134 .056 .049 .058

(.017) (.027) (.022) (.007) (.014) (.008)

Post secondary or tertiary level .273 .211 .308 .261 .248 .257

(.017) (.027) (.021) (.007) (.013) (.008)

University level, lower .543 .475 .572 .544 .439 .568

(.018) (.031) (.023) (.008) (.016) (.009)

University level, higher 1.130 1.046 1.155 1.240 1.107 1.281

(.033) (.051) (.044) (.017) (.029) (.021)

Work experience (in years) .075 .091 .067 .065 .075 .061

(.001) (.002) (.001) (.001) (.001) (.001)

Work experience-squared �.002 �.002 �.001 �.001 �.002 �.001

(.000) (.000) (.000) (.000) (.000) (.000)

Number of childreno11 years �.034 .052 �.080 �.039 .057 �.080

(.005) (.008) (.006) (.002) (.004) (.003)

Marital status (married ¼ 1)b .080 .121 .068 .079 .117 .070

(.009) (.017) (.010) (.005) (.008) (.006)

Observation yearc

1995 �.012* �.016* �.007* .006* .017 .002*

(.006) (.010) (.007) (.003) (.006) (.004)

1996 .030 .026* .037 .058 .069 .054

(.006) (.011) (.008) (.004) (.007) (.004)

1997 .069 .070 .075 .103 .131 .091

(.007) (.012) (.008) (.004) (.007) (.005)

1998 .135 .146 .139 .179 .212 .165

(.007) (.012) (.009) (.004) (.007) (.005)

1999 .175 .176 .185 .215 .240 .205

(.007) (.013) (.009) (.004) (.007) (.005)

2000 .194 .196 .206 .235 .257 .227

(.007) (.013) (.009) (.004) (.008) (.005)

2001 .212 .210 .229 .257 .276 .250

(.008) (.014) (.009) (.004) (.008) (.005)

2002 .231 .228 .250 .283 .294 .280

(.008) (.014) (.010) (.004) (.008) (.005)

Working hours per weekd

20–29 h .089 .063* .103 .089 .016* .115

(.018) (.037) (.020) (.006) (.013) (.007)

30 or more hours .294 .267 .283 .296 .259 .302

(.012) (.025) (.014) (.005) (.009) (.005)

Constant 11.101 11.192 11.122 11.003 11.154 11.054

(.033) (.060) (.039) (.015) (.028) (.018)

R2 (between) .455 .417 .466 .331 .316 .310

R2 (within) .240 .230 .254 .241 .236 .250

R2 (overall) .370 .335 .377 .287 .270 .276

N (observations; i.e. person-years) 133,653 42,092 91,561 500,740 154,889 345,851

N (persons) 22,032 6,873 15,159 86,252 26,938 59,314

Note: Robust standard errors are in parentheses. The regressions also control for additional marital statuses (four dummies), immigrant status (a dummy),

number of family members, region of family residence (seven dummies), tourism sub-sector/hotel classification (seven dummies for tourism sub-sector/five

dummies for hotel classification), parents’ educational level, and a dummy variable indicating if the employee completed his/her schooling the same year as

the annual wage was reported.

*Coefficient is not statistically significant at the .01-level (two-tailed tests).aReference category: lower secondary schooling or below.bReference category: unmarried.cReference category: 1994.dReference category: 4–19h.

C. Thrane / Tourism Management 29 (2008) 514–524 519

ARTICLE IN PRESSC. Thrane / Tourism Management 29 (2008) 514–524520

wages than women during the period from 1994 to 2002,ceteris paribus (exp (.181)�1 ¼ .198). In light of both thedescriptive results of Table 1 and the previous research(Marchante et al., 2005; Santos & Varejao, 2006), thisfinding is not surprising. Furthermore, it should be notedthat the independent variables in the regression account forabout 36% of the overall gender wage gap ((.282�.181)/.282 ¼ .358).5 The result that more education is associatedwith higher annual wages, but not in a strictly linearfashion, is also in accordance with past research (Banuls &Rodrıguez, 2005; Marchante et al., 2005). A notablefinding here is that it is first and foremost a college/university experience that pays off among the tourismemployees. The return to a lower-level university educationwas 72% higher than the return to lower secondaryschooling or below (exp (.543)�1 ¼ .721). For a higher-level university experience, the analogues figure was 210%(exp (1.13)�1 ¼ 2.096). It should also be mentioned that,due to the educational dummies’ narrow confidenceintervals (results not shown), the returns to each educa-tional level are significantly higher than the returns to theeducational level below. Work experience has a curvilinearor concave relationship with annual wage, again as to beexpected (Banuls & Rodrıguez, 2005; Marchante et al.,2005; Santos & Varejao, 2006). The annual wage reachedits highest level at 23.1 years of work experience. Eachadditional child below 11 years in the family was associatedwith a 3.4% yearly wage reduction, and married employeesin general made about 8% higher annual wages thanunmarried ones. The effects of time, represented by a set ofdummies for each observation year, on annual wages arealso easily summarized. Compared with the base year 1994,the tourism employees had a positive annual wagedevelopment from 1996 onwards. Finally, the workinghour dummies, as anticipated, had noticeable positiveimpacts on the annual wage.

The R2between is a goodness-of-fit measure of the cross-

sectional variation in the dependent variable. In the pooledsample of tourism employees, thus, the multivariate modelexplains about 46% of the variation in annual wages. Incontrast, the R2

within is a similar measure of the time-dimension variation in the dependent variable; i.e. howchanges in the independent variables within tourismemployees affect changes in wages. In this latter sense,the model explains only 24% of the annual wage variation.In other words, the model does a much better job ofaccounting for wage differences between employees than itdoes accounting for wage differences as a result of changeswithin employees. The R2

overall is simply the weightedaverage of the R2

between and R2within.

5The Blinder/Oaxaca decomposition of gender differences in wages is

not carried out here for two reasons. First, this technique is developed for

(repeated) cross-sectional data, whereas the present data are longitudinal.

Second, to actually grasp the discrimination portion (if any) of the gender

wage gap, one needs matched employer–employee data.

An important observation from Table 2 is that male andfemale tourism employees have significantly different wagereturns to several of the independent variables. Fourfeatures are especially noteworthy here. First, the annualwage premiums associated with increasing education aregenerally higher for women than they are for men. A casein point is the returns to a lower-level university education,where the yearly wage premium was 77% for women (exp(.572)�1 ¼ .771) and 61% for men (exp (.475)�1 ¼ .608).For post secondary or tertiary education the analoguefigures were 36% and 24%, whereas they were 217% and185% for a higher-level university education. Takentogether these findings are all in agreement with Dough-erty’s (2005) recent review, but at odds with Banuls andRodrıguez’s (2005) results. Second, male employees experi-ence clearly higher wage returns to work experience thanfemale ones during the younger years, a result that fits wellwith the life-cycle human capital model (Ben-Porath, 1967;Polachek, 2006). In other words, men have a more distinctconcave experience–earnings profile than women. Thispattern is more easily detected in Fig. 1. Third, while beingin the childrearing phase is beneficial in terms of earningsfor men, the opposite appears to be the case for women.More specifically, each additional child below 11 years inthe family was associated with a 5% annual wage premiumfor men, whereas the corresponding figure for women wasa near 8% wage penalty. Furthermore, the female over-representation in the industry appears to account for theoverall negative association between parenthood and wagein the pooled sample. These findings are in keeping withrecent research (Høgsnes et al., 2006; Lundberg & Rose,2000; Waldfogel, 1998). Finally, the annual wage premiumassociated with marriage was slightly higher for men (12%)than it was for women (7%). Again, these results fit wellwith the limited evidence available (Dougherty, 2006;Høgsnes et al., 2006).

5.2. Results B: hotel employees

According to the right hand side of Table 1, the meanannual wage for hotel employees in the 1994–2002 periodwas about 172,000 NOK (11.807 expressed in logs). In otherwords, hotel employees on average received 18% lowerannual wages than tourism employees in the exa-mined period (11.975–11.807 ¼ .168; exp (.168)–1 ¼ .183).Regarding the hotel employees’ distribution on theindependent variables, there is little need to explicate theresults to the same level of detail as for the tourismemployees. Yet there are certain differences that deservesome comments. First, hotel employees tend to be lesseducated than tourism employees: 41% of the hotelemployees have only upper secondary or less schooling(.108+.299 ¼ .407), whereas the similar percentage fortourism employees is 30. In contrast, 29% of the tourismemployees have college/university experience while thecorresponding figure for hotel employees is 21%. Second,a considerably lower percentage of the hotel employees are

ARTICLE IN PRESS

0

0 5

Ann

ual w

age

in N

OK

400000

350000

300000

250000

200000

150000

100000

50000

Work experience in years

40353025201510

Male employee

Female employee

Fig. 1. Annual wage by work experience for tourism employees.

Note: The predictions are made from the gender-specific models in Table 2, and refer to a married full-time employed non-immigrant in 2002 with post

secondary/tertiary schooling. The remaining independent variables are set at their gender-specific sample means.

6Compared with the three tourism samples, all three measures of

goodness-of-fit are lower for the hotel samples. The reason for this is most

likely that the overall wage distribution is more condensed for the hotel

samples, as reflected in the lower standard deviations of the annual wages.

A further conjecture could be that it is Norwegian hotel employees’ higher

trade union propensity (compared with tourism employees) that brings

about their more compressed overall wage structure.

C. Thrane / Tourism Management 29 (2008) 514–524 521

working full-time (i.e. more than 30h during a normal week).Third, the hotel employees are generally somewhat lessexperienced than the tourism employees. As for the tourismemployees, male and female hotel employees also havedifferent profiles as regards the means of some of theconsidered variables. In the period studied, men onaverage had 34% higher annual wages than women(12.010�11.716 ¼ .294; exp (.294)�1 ¼ .341). That is,women earned about 75% of men’s wages on average(1/(1+exp (.294)�1) ¼ .745). Marginally more women haveonly lower secondary schooling or some college/universityexperience (i.e. university level, lower), whereas the opposite isthe case for the remaining educational categories. Withrespect to work experience there is only a very modest genderdifference, whereas male employees have slightly fewer smallchildren than their female counterparts.

According to Table 2 the gender coefficient is .190among the hotel employees. In other words, male hotelemployees received on average 21% higher annual wagesthan female hotel employees in the 1994–2002 period,ceteris paribus (exp (.190)�1 ¼ .209). As for the tourismemployees, the independent variables in the regression alsoaccount for about one-third of the overall gender wage gapbetween male and female hotel employees ((.294�.190)/.294 ¼ .354). For the educational and work experiencevariables, we again observe the expected results: moreeducation pays off and wages tend to increase withexperience at a diminishing rate. The effects of number ofchildren and marriage are also very similar to the onesobtained for the tourism employees. In contrast, the hotelemployees experienced a more positive annual wagedevelopment in the 1994–2002 period. A few gender

differences can also be noted as regards the wage returnsto the various independent variables. First, education atthe university level is more beneficial for women than formen. Second, the returns to work experience duringthe younger years are somewhat less steep for womenthan for men. Third, while having small children isbeneficial for men’s wages, women experience the exactopposite pattern. Fourth, the effect of marriage on wages isslightly stronger for men. Finally, the male hotel employ-ees’ annual wage growth was more pronounced than it wasfor their female counterparts. In other words, one of thefour groups stands out in terms of having the mostbeneficial earnings development in the period: male hotelemployees.6

In the preceding model specifications a set of educationallevel dummies represented the variable education. Table 3reports what happens if these educational dummies arereplaced with years of schooling. Other than this, themodels yielding the results reported in Table 3 include thesame set of independent variables as those reported inTable 2. The annual overall returns to one more year ofschooling are above 6% for both types of employees. Themagnitude of this effect is much greater than the returnsreported by Banuls and Rodrıguez (2005) and Marchanteet al. (2005). However, since the former study focused on

ARTICLE IN PRESS

Table 3

Annual wage (logged) by years of schooling for tourism and hotel

employees in Norway, 1994–2002. Random effects regression analysisa

Variables Tourism employees Hotel employees

All Men Women All Men Women

Years of schooling .066 .063 .065 .062 .045 .069

(.002) (.004) (.003) (.001) (.002) (.002)

Note: Robust standard errors are in parentheses. All coefficients are

statistically significant at the .01-level (two-tailed tests).aAll regressions control for the independent variables considered in the

regressions reported in Table 2, except for the educational level dummies.

C. Thrane / Tourism Management 29 (2008) 514–524522

hourly wages and the latter on monthly wages, these figuresare not necessarily comparable. There is also a substantialgender difference in these returns for hotel employees:female hotel employees experienced a 6.9% annual wagereturn to schooling, whereas the similar figure for maleswas 4.5%. The latter result provides additional support tothe notion that female tourism/hotel employees in Norwayare somewhat better compensated than similar maleemployees for their investments in education.7

6. Summary and conclusion

Tourism is an important and growing industry in theworld economy. As such, the relative lack of labor marketstudies in tourism, as opposed to studies of tourists, issomewhat puzzling. This paper has made an effort tobalance this issue more evenly by examining one importantaspect of Norwegian tourism employees’ labor marketoutcomes, namely their earnings. Set within the humancapital framework developed by Becker (1962, 1975) andMincer (1974), the empirical research questions concernedhow the human capital variables education and workexperience, as well as the socio-demographic variablesparenthood and marriage, affected tourism employee’sannual wages. Furthermore, particular attention wasdevoted to how gender and time was associated with thesewages.

In terms of annual wages in the 1994–2002 period takenas a whole, the male tourism employees made about 20%more than the female ones, ceteris paribus. This differenceis in agreement with the gender wage gap reported in anumber of previous studies. A natural question then arisesas to why is this the case. According to several scholars,

7Many of the prior Mincer-type studies of earnings differentiation are

restricted to full-time employees only. This choice is often made in order to

ensure that the sample in question is as homogenous as possible. In light of

this, the preceding analyses have also been carried out for the full-time

employees only. In general these analyses showed the same overall pattern

as the reported ones, although some of the coefficients changed marginally

in magnitude. Because of this, and the fact that the gender and employee

type variations in working hours are of interest in and of themselves, the

results for all employees are presented.

men generally receive higher earnings than women becausethey are over-represented in occupations and occupationalpositions that pay better, and/or more typically areemployed in industries, sectors, establishments or firmswith higher mean wages (Blau & Kahn, 2006; Gunderson,2006; Meyersson Milgrom & Petersen, 2006; Petersen &Morgan, 1995; Weinberger, 1998). In other words, once thewage regressions include variables controlling for thesefeatures, the gender coefficient becomes small and(perhaps) insignificant. Still, recent research also reports anon-trivial and significant gender coefficient when theabove-mentioned controls are taken into account, indicat-ing discrimination against women (Bayard et al., 2003;Datta Gupta & Rothstein, 2005; Santos & Varejao, 2006).Thus, future research on matched employer–employee datais called for in order to settle this score more definitively fortourism employees.8

The human capital variables education and workexperience are important determinants of tourism employ-ees’ annual wages. Based on the several hundreds of priorMincer regressions reported in the literature, this is also asexpected. Investments in education pay off even inNorway, a country that, in a comparative perspective, isknown for its general low returns to education (Harmon,Walker, & Westegard-Nielsen, 2001). It is also interestingto note that educational investments appear to be moreprofitable for female tourism employees compared withmale ones. This finding is contrary to those of Banuls andRodrıguez’s (2005) study of Spanish tourism employees,but in accordance with Dougherty’s (2005) more generalfindings. For work experience the results support theconcave experience–earnings relationship. That is, wagesrise steeply immediately after career entry when the humancapital investments are at their most intense, level offduring midlife, and finally decrease at older ages. However,female employees’ earnings trajectories are flatter thanmen’s (cf. Fig. 1). Part of the reason for this genderdifference is that female tourism employees marry and/orhave children at some point during the first part of theirworking life, and this interruption stalls their accumula-tion of human capital. Indeed, this is exactly what the life-cycle human capital model predicts (Ben-Porath, 1967;Polachek, 2006).Although parenthood and marriage are parts of the

explanation for the gender-specific experience–earningsrelationship, it is important to note that these choices by

themselves also are associated with annual wages. Accord-ing to the results of this study, parenthood is non-beneficialin terms of personal earnings for female tourism employ-ees, whereas the opposite is the case for males. Further-more, marriage appears to be less beneficial in terms ofwages for women. A statement by Polachek (2006, p. 109)is here worth quoting: ‘‘For men, marriage and children are

8It should be mentioned that discrimination also could occur in the

hiring process for various occupations/occupational positions, industries

and establishments.

ARTICLE IN PRESSC. Thrane / Tourism Management 29 (2008) 514–524 523

associated with more work, whereas for women, marriageand children are associated with less work.’’ Concerningthe effects of parenthood and marriage, the results of thisstudy could just as easily be summed up by simply addingto this (in italics): For men, marriage and children areassociated with more work and, thus, higher earnings,whereas for women, marriage and children are associatedwith less work and, thus, lower earnings. It should be noted,however, that the question of causality with respect to thevariables parenthood and marriage is unresolved, and thatfuture research should address this more explicitly. Thetourism employees followed in this study all experienced apositive wage development in the period from 1994 to 2002.In conclusion, this study has clearly demonstrated thathuman capital as well as socio-demographic variables andtime are important determinants of Norwegian tourismemployees’ annual wages. In addition, the study has shownthat an employee’s gender continues to play an importantrole in terms of the wage he or she receives.

References

Almond, P., & Rubery, J. (1998). The gender impact of recent European

trends in wage determination. Work, Employment and Society, 12,

675–693.

Anderson, D. J., Binder, M., & Krause, K. (2002). The motherhood wage

penalty: Which mothers pay it and why? American Economic Review,

92, 354–358.

Banuls, A. L., & Rodrıguez, A. B. R. (2005). Returns on education

in the Spanish tourism labour market. Tourism Economics, 11,

119–132.

Bayard, K., Hellerstein, J., Neumark, D., & Troske, K. (2003). New

evidence on sex segregation and sex differences in wages from

matched employee–employer data. Journal of Labor Economics, 21,

887–922.

Becker, G. S. (1962). Investment in human capital: A theoretical analysis.

Journal of Political Economy, 70, 9–49.

Becker, G. S. (1975). Human capital. New York: National Bureau of

Economic Research.

Ben-Porath, Y. (1967). The production of human capital over the life-

cycle. Journal of Political Economy, 75, 352–365.

Blau, F. D., & Kahn, L. M. (1992). The gender earnings gap: Some

international evidence. Working Paper No. 4224, Cambridge, MA:

National Bureau of Economic Research.

Blau, F. D., & Kahn, L. M. (1996). International differences in male wage

inequality: Institutions versus market forces. Journal of Political

Economy, 104, 791–837.

Blau, F. D., & Kahn, L. M. (2006). The gender pay gap: Goingy going

but not gone. In F. D. Blau, M. C. Brinton, & D. B. Grusky (Eds.),

The declining significance of gender (pp. 37–66). New York: Russell

Sage Foundation.

Blinder, A. S. (1973). Wage discrimination: Reduced form and structural

estimates. Journal of Human Resources, 8, 436–455.

Budig, M. J., & England, P. (2001). The wage penalty for motherhood.

American Sociological Review, 66, 204–225.

Church, A., & Frost, M. (2004). Tourism, the global city and the labour

market in London. Tourism Geographics, 6, 208–228.

Danielsen, K., & Lappgard, T. (2003). Tid er viktig nar barn blir født

[Time is important when children are born]. Samfunnspeilet, 5, /http://

ssb.no/samfunnsspeilet/S.

Datta Gupta, N., & Rothstein, D. S. (2005). The impact of worker and

establishment-level characteristics on male–female wage differentials:

Evidence from Danish matched employee–employer data. Labour, 19,

1–34.

Dougherty, C. (2005). Why are returns to schooling higher for women

than for men? Journal of Human Resources, 40, 969–988.

Dougherty, C. (2006). The marriage earnings premium as a distributed

fixed effect. Journal of Human Resources, 41, 433–443.

Gunderson, M. (2006). Viewpoint: Male–female wage differentials: How

can that be? Canadian Journal of Economics, 39, 1–21.

Halvorsen, R., & Palmquist, R. (1980). The interpretation of dummy

variables in semilogarithmic equations. American Economic Review,

70, 474–475.

Harmon, C., Walker, I., & Westegard-Nielsen, N. (2001). Education and

earnings in Europe—a cross country analysis of returns to education.

UK: Edward Elgar Publishing.

Heckman, J. J., Lochner, L. J., & Todd, P. E. (2003). Fifty years of Mincer

earnings regressions. Working Paper 9732, Cambridge, MA: National

Bureau of Economic Research.

Høgsnes, G., Penner, A., & Petersen, T. (2006). The Family Gap in Wages

and Careers. Memorandum no. 1, Norway: University of Oslo.

Hum, D., & Simpson, W. (2004). Reinterpreting the performance

of immigrant wages from panel data. Empirical Economics, 29,

129–147.

Korenman, S., & Neumark, D. (1992). Marriage, motherhood, and wages.

Journal of Human Resources, 27, 233–255.

Lazear, E. P. (2000). Economic imperialism. Quarterly Journal of

Economics, 115, 99–146.

Lee, C.-K., & Kang, S. (1998). Measuring earnings inequality and

median earnings in the tourism industry. Tourism Management, 19,

341–348.

Loh, E. S. (1996). Productivity differences and the marriage

wage premium for white males. Journal of Human Resources, 31,

566–589.

Lucas, R. E. (2004). Employment relations in the hospitality and tourism

industries. London: Routledge and Kegan Paul.

Lundberg, S., & Rose, E. (2000). Parenthood and the earnings of married

men and women. Labour Economics, 7, 689–710.

Marchante, A. J., Ortega, B., & Pagan, R. (2005). Educational mismatch

and wages in the hospitality sector. Tourism Economics, 11,

103–117.

Meyersson Milgrom, E. M., & Petersen, T. (2006). The glass ceiling in the

United States and Sweden: Lessons from the family-friendly corner of

the world, 1970 to 1990. In F. D. Blau, M. C. Brinton, & D. B. Grusky

(Eds.), The declining significance of gender (pp. 156–211). New York:

Russell Sage Foundation.

Mincer, J. (1958). Investment in human capital and personal income

distribution. Journal of Political Economy, 66, 281–302.

Mincer, J. (1974). Schooling, experience and earnings. New York:

Colombia University Press.

Mincer, J., & Polachek, S. (1974). Family investments in human capital:

Earnings of women. Journal of Political Economy, 82(Supp. 2),

S76–S108.

Murphy, K. M., & Welch, F. (1990). Empirical age–earnings profiles.

Journal of Labor Economics, 8, 202–229.

Oaxaca, R. (1973). Male–female wage differentials in urban labor

markets. International Economic Review, 14, 693–709.

Petersen, T., & Morgan, L. (1995). Separate and unequal: Occupation–

establishment sex segregation and the gender wage gap. American

Journal of Sociology, 101, 329–365.

Polachek, S. (2006). How the life-cycle human capital model explains why

the gender wage gap narrowed. In F. D. Blau, M. C. Brinton, & D. B.

Grusky (Eds.), The declining significance of gender. New York: Russell

Sage Foundation.

Psacharopoulos, G., & Patrinos, H. A. (2004). Returns to investment in

education: A further update. Education Economics, 12, 111–134.

Riley, M., Ladkin, A., & Szivas, E. (2002). Tourism employment. Analysis

and planning. Clevedon: Channel View.

Riley, M., & Szivas, E. (2003). Pay determination. A socioeconomic

framework. Annals of Tourism Research, 30, 446–464.

Robinson, H. (2003). Are you experienced? British evidence on age–

earnings profiles. Applied Economics, 35, 1101–1115.

ARTICLE IN PRESSC. Thrane / Tourism Management 29 (2008) 514–524524

Santos, L. D., & Varejao, J. (2006). Employment pay and discrimination in

the tourism industry. Research—Work in Progress no. 5, Faculdade de

Economia da Universidade do Porto.

Sturman, M. C. (2001). The compensation conundrum Does the

hospitality industry shortchange its employees—and itself? Cornell

Hotel and Restaurant Administration Quarterly, August, 70–76.

Waldfogel, J. (1998). Understanding the ‘‘family gap’’ in pay for women

with children. Journal of Economic Perspectives, 12, 137–156.

Weinberger, C. J. (1998). Race and gender wage gaps in the market for

recent college graduates. Industrial Relations, 37, 67–84.

Wooldridge, J. M. (2002). Econometric analysis of cross section and panel

data. Cambridge, MA: MIT Press.