ARAB WOMEN IN THE ISRAELI LABOR MARKET: CHARACTERISTICS AND POLICY PROPOSALS *

Transcription

1 Israel Economc Revew Vol., No. 2 (23), 4 ARAB WOMEN IN THE ISRAELI LABOR MARKET: CHARACTERISTICS AND POLICY PROPOSALS * ERAN YASHIV AND NITSA KASIR (KALINER) 2 Abstract Ths paper examnes the characterstcs of labor force partcpaton among Arab women n Israel. The partcpaton rates of Israel Arab women are low relatve to those of Israel Jewsh women and to women n Western countres. In a seres of emprcal tests, makng use of ntra-temporal and nter-temporal varaton, the study characterzes the key determnants of the patterns of partcpaton of these women. Hgher educaton and martal status are mportant n explanng partcpaton, as are modern atttudes and modern knowledge. A key mplcaton of the fndngs s that t s not enough to strengthen explanatory varables such as educaton so as to brng about an ncrease n partcpaton. It s also necessary to take measures that wll help overcome the dffculty n translatng explanatory varables, such as educaton, nto partcpaton. Ths dffculty may be due to frctons, such as nformatonal ssues and obstacles to moblty, and to dscrmnaton. The paper proposes a seres of polcy measures to rase partcpaton and smulates a number of future partcpaton trajectores and ther potental contrbuton to GDP.. INTRODUCTION Ths paper examnes the labor force partcpaton of Arab women n Israel. It studes the determnants of the partcpaton decson, ncludng changes that took place over tme and proposes measures for government polcy. The paper bulds on, and expands, our research on the topc that was reported n a prevous study (see Yashv and Kasr (Kalner) 2) but focuses on women only. * Ths paper was wrtten wthn the framework of employment polcy papers sponsored by the Pnhas Sapr Forum for Economc Polcy. The authors thank Ayelet Shpaer, Ruth Bouganm and Davd Elezer for ther excellent research assstance n organzng and processng the data, and the partcpants n the Sapr Forum for ther useful remarks. Eran Yashv, The Etan Berglas School of Economcs, Tel Avv Unversty yashv@post.tau.ac.l. 2 Ntsa Kasr (Kalner), Research Department, Bank of Israel, ntsa.kasr@bo.org.l.

2 2 ISRAEL ECONOMIC REVIEW The paper s structured as follows: Secton 2 presents background data. Secton 3 delneates a labor force partcpaton model. Secton 4 presents econometrc estmaton of the partcpaton equaton, usng data from the 29 Labor Force Survey. Secton 5 examnes long-term changes n the behavor of the explanatory varables and ther margnal value as estmated n ths equaton. Secton 6 examnes labor force partcpaton usng data from the 25 Socal Survey. Secton 7 proposes polcy measures on ths topc and smulates future partcpaton trajectores and ther contrbuton to GDP. Secton 8 concludes. 2. BACKGROUND DATA a. Data sources The data n ths study were taken from Labor Force Surveys (varous years) and the Socal Survey (for 25), both carred out by the Israel Central Bureau of Statstcs (CBS). The quarterly Labor Force Survey ncludes nformaton about partcpaton, employment, and unemployment, and demographcs. The survey refers to the permanent populaton of Israel aged 5 and older. In each quarter four panels comprsng 2,7 households each are surveyed. Each panel s nvestgated over two successve quarters, s removed from the sample for two successve quarters, and returns to the sample for two successve quarters. 3 The Socal Survey s an annual survey whch provdes nformaton about the lvng condtons and welfare of the adult Israel populaton and asks how ndvduals n socety perceve varous aspects of ther lves. The survey s composed of fxed topcs that are nvestgated every year, and a specfc topc whch s nvestgated n detal and changes each year. In 25, the survey ncluded an n-depth nvestgaton of labor force partcpaton, whch s the reason we chose t for ths study. The Socal Survey ncludes numerous detals that do not appear n the Labor Force Survey, ncludng nformaton on relgous characterzaton, use of varous knds of technology, and relatons wth famly members. The sample for the 25 Socal Survey conssted of 9,543 ndvduals aged 2+ who represent 4.3 mllon people n ths age group. Approxmately 5 of them were found not to belong to the survey populaton and 83 percent of the others took part n the survey, so the sample has 7,647 people n all. b. Arab women n the Israel labor market The fgures below present key stylzed facts of Arab women s partcpaton. 3 The four panels exst only n relatvely large Jewsh localtes and n several Arab localtes, such as Nazareth and Eastern Jerusalem. In the other localtes, only one or two panels exst.

3 ARAB WOMEN IN THE ISRAELI LABOR MARKET: CHARACTERISTICS AND POLICY PROPOSALS 3 Fgure Partcpaton Rate over Tme, by Sector Women, 97 29,2 7 6 Jewsh women Arab women 5 4 Pct An Arab woman was defned n as a non-jewsh woman whose father s not Jewsh and whose father s contnent of orgn s nether Europe nor Amerca. In 979 2, an Arab woman was defned as a Chrstan, Muslm, or Druze woman whose contnent of brth, and whose father s contnent of brth, was nether Europe nor Amerca. From 2 on, an Arab woman may be profled wth greater precson due to the collecton of new varables. 2 In 998, 2, and 29, a change was made n the method of extrapolatng the CBS sample. The rates n those years were also calculated accordng to the old extrapolaton method to facltate comparson between the years. Source: calculatons based on CBS Labor Force Surveys. 99 Year Fgure 2 Women Lfe Cycle Partcpaton Rates, Pct. 6 5 Jewsh women Labor Force Survey Arab women Labor Force Survey Age Source: calculatons based on the CBS Labor Force Survey, 2

4 4 ISRAEL ECONOMIC REVIEW Fgure 3a Women Lfe Cycle Partcpaton Rates, Selected Countres, U.S. Italy France Germany Israel Arabs 6 Pct Source: calculatons based on the CBS Labor Force Survey, 28; Fgure 3b Women Lfe Cycle Partcpaton Rates, Comparson wth Muslm and Arab Countres, Age Turkey Syra Iran Israel Arabs Pal. Authorty 25 Pct Source: Israel Arabs calculatons based on the CBS Labor Force Survey, 28. Iran and Syra The data for Syra are current up to 27. Palestnan Authorty: Age

5 ARAB WOMEN IN THE ISRAELI LABOR MARKET: CHARACTERISTICS AND POLICY PROPOSALS 5 Fgure 4 Women Lfe Cycle Partcpaton Rates over Tme Pct In 98 2, an Arab woman was defned as a Chrstan, Muslm, or Druze woman whose contnent of brth, and whose father s contnent of brth, was nether Europe nor Amerca. From 25 on, an Arab woman has been defned on the bass of a new and more accurate varable that was frst ncluded n the Labor Force Survey n 2. Source: calculatons based on CBS Labor Force Surveys. Age Fgure 5 Dstrbuton of Employment by Industry Arab Women, Pct Educaton Health, welfare, and socal servces Wholesale and retal trade Agrculture and manufacturng Bankng, nsurance, and busness servces Accommodaton servces and restaurants Communty, socal, personal, and other servces Transport, storage, communcatons Publc admnstraton Constructon, electrcty, and water Foregn organzatons and enttes Unknown Source: calculatons based on the CBS Labor Force Survey, 2.

6 6 ISRAEL ECONOMIC REVIEW Fgure 6 Dstrbuton of Employment by Occupatons Arab Women, Pct Lberal and techncal professons Agents, salespeople, and servce workers Whte-collar workers Academc occupatons Unsklled workers 4. Sklled workers n manufacturng, constructon, and other Source: calculatons based on the CBS Labor Force Survey, ManagersSklled workersoccupaton n agrculture unknown The graphs ndcate the followng man ponts:. The partcpaton rate of Arab women has been rsng steadly over the years, as n many Western countres. The rate doubled between 97 and 2 from percent to 2 percent but remans low. Snce ther partcpaton rate has ncreased more slowly than that of Jewsh women, the gap between the two populatons has wdened In terms of the age profle of partcpaton, Arab women have a much lower profle than Jewsh women. The partcpaton rate of Arab women n Israel s also low by Western standards. However, the patterns of Arab women s partcpaton are not materally dfferent from the norm n varous Arab and Muslm countres. 3. Over the years, the age profle among women has been trendng up, especally n the age group. 4. The data on occupatons and employment by ndustres show a large concentraton of Arab women n educaton and healthcare. 5 4 See also The Book of Arab Socety n Israel (3), 29, pp (n Hebrew). 5 For more on the partcpaton profle of Arab women, see Fchtelberg, 24.

7 ARAB WOMEN IN THE ISRAELI LABOR MARKET: CHARACTERISTICS AND POLICY PROPOSALS 7 3. THE PARTICIPATION EQUATION AND ESTIMATION There s a volumnous lterature on labor force partcpaton. 6 The followng s a basc model that appears wdely n ths lterature, as specfed, for example, n the survey paper by Blundell and Macurdy (28) whch we follow here. The ndvdual has a quas-convex utlty functon that rses wth consumpton (c) and lesure (l). The problem of the ndvdual s to maxmze utlty, as follows: () max U ( c, l, v ) s.t. l + c = y + w h ( 2) ( 3) h + l = T c h + Where U s the utlty functon and v s the vector of the ndvdual s characterstcs. These characterstcs may be observed or unobserved and may nclude cultural factors. The wage per hour s denoted w, y s ncome from other sources, T s total hours avalable to the ndvdual, and h s the number of work hours. By solvng ths problem we obtan the supply of work hours, h, as a postve functon of w, a negatve functon of y, and a functon of characterstcs vector v. * (4) h = h ( w, y, v ) + The partcpaton decson, p, s expressed by the followng: (5) p = { h * > h } * When varable h, ndvdual s desred number of work hours, exceeds the threshold h, the ndvdual partcpates ( p = ). Otherwse, she does not ( p = ). The * expressons h and h are derved from the specfcaton of the utlty functons. For dscusson of varous functonal forms, see Blundell, Macurdy, and Meghr (27), partcularly pp a. The partcpaton equaton We adopt the followng functonal forms for h and The ndvdual s desred work hours are denoted by: h * = + z δ + ln wδ + ln yδ 3 δ + ς (6) 2 * h : 6 For a detaled dscusson, see Blundell, MaCurdy, and Meghr (27) and, n partcular, Secton 2.

8 8 ISRAEL ECONOMIC REVIEW Where δ s constant; z s a vector of exogenous explanatory varables such as gender, age, and martal status; ln w s the log of the wage offered to the ndvdual; ln y s the log of the ndvdual s ncome from other (.e., non-labor) sources; and ς represents unobservable characterstcs. The parameters n the equaton are δ, δ 2, and δ 3 and the vector δ. The threshold, h, s expressed by: (7) h = γ + zγ + ln wγ 2 + ln y γ 3 + ξ Where γ s a vector of parameters that s not necessarly dentcal to δ ; γ s a constant; γ 2 and γ 3 are parameters; and ξ unobservable characterstcs. Notably, the characterstcs vector v s captured n Equatons (6) and (7) by the constants δ and γ, by the vector z, and by the unobserved terms ς and ξ. We also note that the vector z may be broken down nto two components, as follows: D C (8) = [ z z ] z, D z t ncludes educaton and demographc varables and Where varables. Therefore, the partcpaton decson, (9) p = { h * > h } C z ncludes cultural p, s descrbed by the followng equaton: Snce the wage s unobserved n the case of non-partcpants, we assume: ln w θ + z θ + z θ + ω = 2 2 () Where z 2 s a vector of exogenous varables; θ s a parameter; θ and θ 2 are vectors of parameters; and ω are unobservable characterstcs. We assume that the other ncome varable s determned as follows: ln y = ( ) + χ () z Where [ z, z2, z3 ] = z n other words, a vector of exogenous varables that ncludes both the varables mentoned above and addtonal ones. z 3 and χ are unobservable characterstcs. Therefore, by combnng Equatons (9), (), and (), we obtan the followng partcpaton equaton: (2) p { β + z β + z β + z β + u } = > Where β s a parameter; β, β 2, and β 3 are vectors of parameters; vectors z, z2, z3 are defned as above; and u are unobservable characterstcs. Ths equaton may be estmated n varous ways, dependng on the assumptons made about the dstrbuton of the unobserved characterstcs. Below we estmate t usng a Probt method.

9 ARAB WOMEN IN THE ISRAELI LABOR MARKET: CHARACTERISTICS AND POLICY PROPOSALS 9 b. The emprcal method and the data The estmaton reported below refers to equaton (2). The data are taken from the Labor Force Surveys or the 25 Socal Survey, as noted n Secton 2. above. The basc regressons use Labor Force Survey data; when cultural varables are added to the equatons, the Socal Survey s used nstead. 4. THE PARTICIPATION REGRESSION Frst we present the results of estmaton of Equaton (2) usng data from the Labor Force Survey for 29. The explanatory varables nclude the classc ones (educaton and age), demographc varables (martal status, number of chldren, type of localty), number of addtonal wage earners and relgon. The results of the regresson show that the age coeffcents yeld a profle that rses and then falls n the course of the lfe cycle, much lke the partcpaton rates n Fgures 2 and 4. Consequently, even when we use control varables, we obtan a hump-shaped partcpaton profle across the lfe-cycle. The educaton coeffcents show, as expected, that the level of educaton has postve, strong, and sgnfcant effects on partcpaton, wth 6+ years of schoolng havng a much stronger effect than 3 5 years. These fndngs reflect the fact that the lkelhood of fndng work and the potental wage both ncrease n schoolng. The regresson shows that dvorcees and never-marred women tend to partcpate more; separated women also tend to partcpate more, but the effect s not as strong. These fndngs can be explaned by the needs of the household accordng to ts composton. The presence of addtonal wage earners n the household has a postve effect on partcpaton. Ostensbly ths fndng clashes wth the prevous fndng about martal status because the presence of addtonal wage earners n the household makes t less necessary to work outsde the home (an ncome effect). However, t evdently reflects people s tendency to marry those who are smlar to them, ncludng smlarty n schoolng (assortatve matchng). 7 The number of chldren up to age 7 has a negatve effect; evdently ths reflects the negatve ncome effect of chld allowances alongsde the need for chldcare. Although the presence of young chldren usually has a negatve effect, t s not sgnfcant n the regresson for 29. (However, see below.) Women who lve n large ctes exhbt hgher partcpaton rates, evdently because they have better access to transportaton (especally mportant n a tradtonal socety) and more employment opportuntes. 8 However, lvng n Jerusalem has a negatve effect, probably because t s a proxy for afflaton wth the Palestnan economy. Fnally, relgon has a margnal effect wth beng Chrstan or Druze ncreasng the woman's lkelhood of partcpaton. 7 Varous studes n Israel that ran a partcpaton or employment regresson yelded a smlar fndng. See, for example, Flug and Kasr (Kalner), 23, An ethnc enclave has a postve effect on Muslm women and a negatve one on Chrstan women. Chrstan Arabs are more urban than Muslms and Druze, who dwell n rural enclaves. See, for example, Khattab (22).

10 ISRAEL ECONOMIC REVIEW Table Arab Women's Labor Force Partcpaton Regresson 29 Labor Force Survey Column Column 2 Varable Age group (control: 35-39) Coeffcent S.D. Coeffcent S.D ** Years of schoolng (control: -2) Martal status (control: marred) Separated.74.2 Dvorced.66.7 Wdowed Sngle.58.7 N chldren up to age 9 n famly.4.3 N chldren up to age 7 n famly Localty type (control: 5,-, populaton) Jerusalem Tel Avv.58.8 Hafa.33.5,-2, populaton Up to 5, populaton.2.9 Rural localty Other wage earners n household (control: none) Relgon (control: Muslm).35.7 Chrstan.42.7 Druze.9.7 Constant N observatons 7,3 7,3 F-statstc Prob > F.. * denotes % sgnfcance; ** denotes 5% sgnfcance ; denotes percent sgnfcance Source: 29 Labor Force Survey. Labor force partcpaton regresson: -partcpated; -dd not partcpate

11 ARAB WOMEN IN THE ISRAELI LABOR MARKET: CHARACTERISTICS AND POLICY PROPOSALS An alternatve nterpretaton of the fndngs follows the dchotomy of modern versus tradtonal Arab women. Modern Arab women share the characterstcs of contemporary Western women, ncludng advanced schoolng and hgh partcpaton. Modernty has two aspects, one relatng to culture and mentalty and the other to the use of advanced technologes ( modern knowledge ). Obvously, the two are connected. Many studes have found that modernty varables affect partcpaton n a manner dstnct from other varables. 9 The fndngs n Table also suggest a postve relaton between modernty varables among Arab women and partcpaton: there s a postve correlaton between modernty and level of schoolng whch has a very strong postve effect on partcpaton. Arab women who are dvorced or sngle (above a certan age) are more lkely not to be tradtonal. The fndngs on hgher partcpaton rates among these women are consstent wth the percepton that modern women partcpate more. Does age have an effect when combned wth the other explanatory varables? One may answer ths queston by dvdng the regressons nto age groups. The fndngs obtaned (not reported) resemble those of the regresson coverng the entre populaton of Arab women, but n dfferent ntenstes. What s the source of the dfferences between Arab women and Jewsh women? One may run the same regresson for Jewsh women and examne the dfferences. Table 2 deconstructs the dfferences n these regressons n the manner proposed by Farle (25), whch dfferentates between the effect of dfferences n the dstrbuton of the explanatory varables (between Jewsh women and Arab women) and the effect of the dfferences n the coeffcents of these varables. Table 2 shows the probabltes of partcpaton among Arab and Jewsh women, how much of the gap n probablty s explaned by dfferences n the dstrbuton of the explanatory varables, and the contrbuton of each explanatory varable to explanng the gap. The analyss shows that dfferences n the dstrbuton of the varables explan only percent of the dfference n the probablty of partcpaton between Arab women and Jewsh women:.35 of a dfference of.37 n the probablty of partcpaton. The mplcaton s that most of the dfferences are assocated wth the effect of the explanatory varables on partcpaton, as dstnct from the value of the explanatory varables. It was also found that dfferences n educaton and number of chldren explan. of the partcpaton gap. That s, f educaton and number of chldren (up to age 9 and up to age 7) were dstrbuted dentcally among Arab women and among Jewsh women, the partcpaton gap would be percentage ponts smaller. However, the effect of age and, partcularly, age 55+, acts n the opposte drecton,.e., f Arab women and the Jewsh women had the same age composton (that s, f the Arab women were older), the partcpaton rate of the former would be even lower than t s. 9 For example, Lfshtz (24) found n all her smulatons that modern women (defned as belongng to a famly n whch both spouses are perceved as equal) partcpate more than tradtonal women (those n famles where the male s the man wage earner). See also Ecksten and Lfshtz (2).

12 2 ISRAEL ECONOMIC REVIEW Table 2 Farle Decomposton, Women's Labor Force Partcpaton Regresson 29 Labor Force Survey Varable Coeffcent Age (control: 35-39) ** Years of schoolng (control: -2) Martal status (control: marred) Separated. Dvorced.3 Wdowed. Sngle. N chldren up to age 9 n famly.5 N chldren up to age 7 n famly.26 Localty type (control: 5,-, populaton) Jerusalem.2 ** Tel Avv. Hafa.,-2, populaton.2 Up to 5, populaton. Rural localty. Other wage earners n household (control: none) 2+ Total observatons = 44,3 Total Jewsh women = 37,394 Total Arab women = 6,97 Probablty of labor force partcpaton among Jewsh women: S.D Probablty of labor force partcpaton among Jewsh women: Dfference = Total explaned: * denotes % sgnfcance; ** denotes 5% sgnfcance; denotes percent sgnfcance Source: 29 Labor Force Survey. Labor force partcpaton regresson: -partcpated; -dd not partcpate

13 ARAB WOMEN IN THE ISRAELI LABOR MARKET: CHARACTERISTICS AND POLICY PROPOSALS 3 5. CHANGES OVER TIME In the prevous sectons, we noted that Israel Arab women have lower partcpaton rates than Israel Jewsh women and women n Western countres. The partcpaton regresson (partly) explans these partcpaton patterns, takng nto account educaton, martal status, chld allowances and place of resdence. Ths account s vald for the analyss of a crosssecton of women n a gven year (29),.e., usng the dfference among women at a gven pont n tme. In ths secton, we examne the dfference over tme. Frst, how dd the explanatory varables n the aforementoned regresson change over tme? Second, how dd ther margnal effect (a functon of the regresson coeffcents) change over tme? a. Changes n explanatory varables We begn by presentng graphs that show the dfferences n the explanatory varables over tme. Fgure 7a Rate of Arab Women wth Post-Secondary Schoolng, Over Tme years Arab women years Arab women 3 5 years Jewsh women 2 6+ years Jewsh women Pct Years Source: calculatons based on CBS Labor Force Surveys.

14 4 ISRAEL ECONOMIC REVIEW Fgure 7b Medan Age of Women Over Tme Age Medan age, Arab women Medan age, Jewsh women Year Source: calculatons based on CBS Labor Force Surveys. Fgure 7c Dstrbuton of Martal Status Over Tme Pct. 2 5 Dvorced Arab women Sngle Arab women Separated Arab women Dvorced Jewsh women Sngle Jewsh women Separated Jewsh women Year Source: calculatons based on CBS Labor Force Surveys.

15 ARAB WOMEN IN THE ISRAELI LABOR MARKET: CHARACTERISTICS AND POLICY PROPOSALS 5 Fgure 7d Share of Women wth Chldren up to Age Arab women No chldren up to age 9 chld up to age chldren up to age 9 4+ chldren up to age Jewsh women No chldren up to age 9 chld up to age chldren up to age 9 4+ chldren up to age P c t. 3 P c t Year Year Source: calculatons based on CBS Labor Force Surveys. Fgure 7e Share of Women wth Chldren up to Age Arab women No chldren up to age 7 chld up to age chldren up to age 7 4+ chldren up to age Jewsh women No chldren up to age 7 chld up to age chldren up to age 7 4+ chldren up to age P c t. 3 P c t Year Year Source: calculatons based on CBS Labor Force Surveys.

16 6 ISRAEL ECONOMIC REVIEW Fgure 7f Relgon 25 2 Chrstan Druze 5 Pct Year Source: calculatons based on CBS Labor Force Surveys. The graphs show a marked ncrease n the proporton of Arab women wth hgher educaton over the years, some ncrease n the medan age, a slght declne n the share of never-marred women (after an ncrease n the 97s), a decrease n the number of chldren ncludng a smaller share of large famles, a declne n the proporton of Chrstan women, and stablty n the proporton of Druze. Most of these changes explan the upward drft of the partcpaton rate over tme; the ncrease n age and the declne n the proporton of Chrstan women have the opposte effect. b. Changes n the margnal effects of the explanatory varables Over the years, the regresson coeffcents have been changng and so has the margnal effect of each varable. The graphs below depct the development of the margnal effects of the varables over tme. For elaboraton on the state of hgher educaton n Israel Arab socety, see The Book of Arab Socety n Israel: Populaton, Socety, Economy (3), pp

17 ARAB WOMEN IN THE ISRAELI LABOR MARKET: CHARACTERISTICS AND POLICY PROPOSALS 7 Fgure 8a Changes n the Margnal Effects of Schoolng.5.4 Margnal value of coeffcent years 9 years 3 5 years 6+ years Year Source: calculatons based on CBS Labor Force Surveys. Fgure 8b Changes n the Margnal Effects of Age Margnal value of coeffcent Age 8/2 24 Age Age 3 34 Age Year Source: calculatons based on CBS Labor Force Surveys.

18 8 ISRAEL ECONOMIC REVIEW Fgure 8c Changes n the Margnal Effects of Age Margnal value of coeffcent Age Age 5 54 Age Age 6 64 Age Year Source: calculatons based on CBS Labor Force Surveys. Fgure 8d Changes n the Margnal Effects of Presence of Addtonal Wage Earners.2.5 addtonal breadwnner 2 or more addtonal breadwnners Margnal value of coeffcent Year Source: calculatons based on CBS Labor Force Surveys.

19 ARAB WOMEN IN THE ISRAELI LABOR MARKET: CHARACTERISTICS AND POLICY PROPOSALS 9 Fgure 8e Changes n the Margnal Effects of Number of Chldren.5 Margnal value of coeffcent Number of chldren up to age 9 Number of chldren up to age Years Source: calculatons based on CBS Labor Force Surveys. The fgures ndcate that hgher schoolng has become more mportant over tme, especally among those who have 6+ years of schoolng and even n comparson wth those who have 3 5 years. The negatve effects of beng poorly educated ( 8 years of schoolng) have also ncreased somewhat. Over the years, the hump shape of partcpaton across the lfe cycle has become steeper (as n Fgure 5, but here t s calculated on the bass of the age coeffcents n the regresson). The effect of the number of chldren up to age 9 was negatve n most years; the effect of the number of chldren up to age 7 was usually negatve and has become somewhat more so n the past decade. The effect of postve assortatve matchng has strengthened, evdenced by the ncrease n the postve effect of havng addtonal wage earners. It should be noted though that a negatve effect, exsted at the begnnng of the sample perod. c. Changes over tme the Farle decomposton In ths secton, we examne how the dstrbuton of the explanatory varables affects the dfferental between Arab women and Jewsh women n the probablty of labor force partcpaton. For each explanatory varable, we also examne the varance over the years n the regresson coeffcents, decomposed by the method proposed by Farle (25).

20 2 ISRAEL ECONOMIC REVIEW As we explaned above, ths method reveals how much of the partcpaton gap can be traced to dfferences n the explanatory varables. In addton to ths decomposton, we ran each year s partcpaton equatons n an OLS regresson and performed a smlar decomposton usng the method proposed by Blnder (973) and Oaxaca (973). Notably, OLS estmaton s more problematc when the dependent varable s bnary, as t s here. The fndngs obtaned, however, are very smlar and both decomposton methods lead to the same conclusons. Fgures 9a and 9b present the two decompostons. Fgure 9a Farle (25) Decomposton Value of coeffcent Dfference left axs Total explaned left Share of explanaton n dfference rght axs Years. Source: calculatons based on CBS Labor Force Surveys.

21 ARAB WOMEN IN THE ISRAELI LABOR MARKET: CHARACTERISTICS AND POLICY PROPOSALS 2 Fgure 9b Blnder (973) and Oaxaca (973) Decomposton Meabn dfference of explanatory varables Constant dfference Dfference of explanatory varable coeffcents 2 Percent explaned Years Source: calculatons based on CBS Labor Force Surveys. Fgure 9a shows that whle the dfference between Arab women and Jewsh women n the probablty of partcpaton wdened consderably over tme from.25 n 97 to.4 n recent years the part of the gap that s explaned by dfferences n the dstrbuton of the explanatory varables has contracted consderably. The dfferences n the dstrbuton of the varables whch explaned about half the gap n 97 explaned less than percent n 29 accordng to the Farle (25) method. Fgure 9b ndcates that accordng to the Blnder (973) and Oaxaca (973) decomposton, the explanatory power of dfferences n the dstrbuton of the explanatory varables declned from 39.2 percent n 98 to 3.8 percent n 29. What does ths fndng mean? The followng equaton shows the decomposton of the regressons followng the OLS formula: J A J A J A J J A P P = C C + X X ) + ( ) A P J ( β β β P, are the means of the dependent varables,.e., the probablty of where J A partcpaton among Jewsh women and Arab women, respectvely; A J C, C are the constants n the Jewsh and Arab regressons, respectvely; X, X are the explanatory A varables n the Jewsh and Arab regressons, respectvely; and β, β J are vectors of coeffcents estmated n the regressons of Jewsh and Arab women, respectvely. X J

22 22 ISRAEL ECONOMIC REVIEW The decomposton yelds three components: the dfference n the constants of the regressons, the dfference n the means of the explanatory varables (multpled by the regresson coeffcents), and the dfference n the regresson coeffcents (multpled by the means of the explanatory varables). We address each component below: For the frst component, examnng the gap between the regresson constants, we fnd that ths factor provdes the man explanaton for the partcpaton dfferental. The dfference between the constants has been narrowng slghtly over tme. For the second component, the graphs above show a decrease n the explanatory power of the mean dfferental n the explanatory varables. Observng each varable separately, we fnd that the average dfferental has narrowed n some varables and wdened n others, but the decrease n ths component shows that the contracton effect was domnant. The thrd component was negatve; by ths measure, the coeffcents of the varables were hgher for Arab women than for Jewsh women. Had ths component been domnant, Arab women would have had a hgher partcpaton rate than Jewsh women. Over the years, ths effect has been weakenng,.e., the negatve spread of the coeffcents has been narrowng. What does ths decomposton teach us? The ncrease n Arab women s partcpaton rates may be traced, n the man, to the fact that they are becomng more educated, havng fewer chldren, and formng more households that have an addtonal wage earner. Harder to explan s the concurrent wdenng of the partcpaton gap n favor of Jewsh women. Accordng to the fndngs above, the man explanatory factor s the weakenng of the effect of the thrd component. The queston s how to understand ths component, especally n comparson wth the frst one, whch s domnant. One may say that the frst component reflects the dsparty between Jewsh women and Arab women n matters of cultural dfference, dscrmnaton, or frctons. For example, cultural factors, dscrmnaton by employers, and frctonal factors such as poor access to transport, lack of nformaton, lack of day care centers, etc., may lead to lower partcpaton by Arab women, ceters parbus. Ths component has always been more domnant than the thrd one, accordng to whch Arab women may have hgher coeffcents due to selectvty. Now that ths last component has weakened, the dsparty has wdened because the frst component has become even more domnant. Fgure 9c demonstrates ths n graph form. It shows the actual partcpaton rate of Jewsh and Arab women, the rate predcted by the regresson f Arab women s coeffcents were the same as those of Jewsh women, and the rate predcted f Arab women had the Jewsh women s coeffcents and also ther constant. The graph shows that the Arabs partcpaton rate n the last-mentoned case would have been 68% n 29.

23 ARAB WOMEN IN THE ISRAELI LABOR MARKET: CHARACTERISTICS AND POLICY PROPOSALS 23 Fgure 9c Smulaton of Partcpaton Rates by OLS Regressons.8 Actual partcpaton rate, Jewsh women Actual partcpaton rate, Arab women Arab women s partcpaton rate wth Jewsh women s coeffcents and constant Arab women s partcpaton rate wth Jewsh women s coeffcents Source: calculatons based on CBS Labor Force Surveys. In another noteworthy phenomenon, not all the explanatory varables converge. That s, Arab women dd not become lke Jewsh women n any varable that postvely explaned the ncrease n the employment rate. Thus, the educaton gap wdened (Fgure 7a), the share of Chrstan women n the populaton declned, the proporton of the nevermarred decreased somewhat n recent years (after rsng n the early 97s), and the number of households that had at least two wage earners dropped. The fndngs about the weakenng of the second component over tme, however, show that the explanatory power of the spread between the explanatory varables has narrowed overall. The decrease traces manly to the narrowng of the gaps as a result of the convergence of the age dstrbuton and the number of young chldren, the narrowng of educaton gaps snce the md-99s, and the ncrease n the number of Arab women lvng n households that have addtonal wage earners.

24 24 ISRAEL ECONOMIC REVIEW 6. PARTICIPATION OF ARAB WOMEN: FINDINGS FROM THE 25 SOCIAL SURVEY a. Fndngs from the partcpaton regressons The 25 Socal Survey allows us to relate to addtonal varables. In the partcpaton regressons based on ths survey, we added several explanatory varables that are unque to the survey: relgous afflaton, varables attestng to modern knowledge (command of Englsh, havng a drver s lcense, or usng a computer), and atttudes toward men and women n the labor market. Tables 3a 3c report the results, showng three regressons: one ncludng varables specfc to the Socal Survey, one comprsed of varables partcular to the Labor Force Survey, and one that ntegrates the two. Table 3a Arab Women Labor Force Partcpaton Equaton, "Culture" Varables Only--25 Socal Survey 25 Socal Survey Varable Coeffcent Relgon (control: Muslm) Chrstan -. Druze.46 Relgosty (control: not relgous) Very relgous -.28 Relgous -.44 Not so relgous -.7 Profcent n Englsh.4 Has drver's lcense.94 Uses computer.4 Atttudes Man and woman retre at same age.3 Both work.38 In famly wth young chldren: One parent should work less.73 Both parents should work less.6 Both parents should work as usual.39 Constant -2.2 N observatons 532. F-statstc 2.22 Prob > F. * denotes % sgnfcance; ** denotes 5% sgnfcance; denotes percent sgnfcance Source: 25 Socal Survey. Labor force partcpaton regresson: -partcpated; -dd not partcpate S.D

25 ARAB WOMEN IN THE ISRAELI LABOR MARKET: CHARACTERISTICS AND POLICY PROPOSALS 25 Table 3b Arab Women Labor Force Partcpaton Equaton, "General" Varables Only 25 Socal Survey Varable Coeffcent Age (control: 35-39) * * Years of schoolng (control: -2) ** Martal status (control: marred) Dvorced.7 Wdowed.4 Sngle.58 N chldren (control: ) ** ** * ** Other wage earners n household (control: none).49 ** Localty type (control: 5.-, populaton Jerusalem -.6 ** Hafa -.33 Tel Avv. * 2,-5,populaton. Rural localty -.6 Satsfacton wth publc transt (control: do not use publc transt) Very satsfed -.6 Satsfed.4 Not so satsfed -.66 ** Not satsfed at all.7 Constant -.39 N observatons 53. F-statstc 4.6 Prob > F. * denotes % sgnfcance; ** denotes 5% sgnfcance; denotes percent sgnfcance Source: 25 Socal Survey. Labor force partcpaton regresson: -partcpated; -dd not partcpate S.D

26 26 ISRAEL ECONOMIC REVIEW Table 3c Arab Women Labor Force Partcpaton Regresson, Most Varables 25 Socal Survey Varable Age (control: 35-39) Coeffcent S.D ** Receves beneft Martal status (control: marred) Dvorced Wdowed.4.57 Sngle N chldren (control: ) ** Relgon (control: Muslm) Chrstan..29 Druze.48 **.24 Profcent n Englsh.38 **.8 Has drver's lcense.88.8 Uses computer Atttudes.36 *.9 Man and woman retre at same age.27.7 Both work In famly wth young chldren:.58 **.23 One parent should work less.73.8 Both parents should work less.97 **.42 Both parents should work as usual.36 *.9 Constant N observatons 53 F-statstc 5.42 Prob > F. * denotes % sgnfcance; ** denotes 5% sgnfcance; denotes percent sgnfcance Source: 25 Socal Survey. Labor force partcpaton regresson: -partcpated; -dd not partcpate The tables ndcate that the results for the ordnary varables correspond to those of the Labor Force Survey.

27 ARAB WOMEN IN THE ISRAELI LABOR MARKET: CHARACTERISTICS AND POLICY PROPOSALS 27 The fndngs show that women who have modern atttudes and knowledge are more nclned to partcpate n the labor market than others. (See also Kng, Naon, and Volda- Tzadk, 29.) Ths fndng s consstent wth the results obtaned by Sperngs and Smts (2), who examned the partcpaton of Arab women n Egypt, Jordan, Syra, Morocco, and Tunsa. Somewhat to our surprse, the regresson based only on modern atttudes and knowledge yelded an explanaton not much less powerful than that elcted by the ordnary regresson. In ths regresson, modern knowledge and atttudes have a postve and sgnfcant effect. Relgous afflaton, however, was not found to be sgnfcant. Another varable examned was satsfacton wth publc transportaton. We ncluded t due to the common argument that partcpaton s mpeded by dffcultes related to publc transportaton. Ths varable, too, was not found to be sgnfcant. b. Farle decomposton Below we compare Jewsh women wth Arab women usng the decomposton proposed by Farle (25), descrbed above. Table 3d Arab and Jewsh Women Labor Force Partcpaton Equaton "Culture" Varables Varable Relgon (control: Muslm) Coeffcent S.D. Chrstan Druze Relgosty (control: not relgous) (omtted). Very relgous (omtted). Relgous Not so relgous.29 *.6 Profcent n Englsh Has drver's lcense Uses computer Atttudes.3.72 Man and woman retre at same age Both work In famly wth young chldren:.6.22 One parent should work less Both parents should work less.3.9 Both parents should work as usual Total observatons = 3,838 Total, Jewsh women = 3,36 Total, Arab women = Probablty of labor force partcpaton among Jewsh women = Probablty of labor force partcpaton among Arab women = Dfference: Total explaned: * denotes % sgnfcance; ** denotes 5% sgnfcance; denotes percent sgnfcance Source: 25 Socal Survey. Labor force partcpaton regresson: -partcpated; -dd not partcpate

28 28 ISRAEL ECONOMIC REVIEW Table 3d Arab and Jewsh Women Labor Force Partcpaton Equaton "Culture" Varables Varable Relgon (control: Muslm) Chrstan Druze Relgosty (control: not relgous) Very relgous Relgous Not so relgous Profcent n Englsh Has drver's lcense Uses computer Atttudes Man and woman retre at same age Both work In famly wth young chldren: One parent should work less Both parents should work less Both parents should work as usual Total observatons = 3,838 Total, Jewsh women = 3,36 Total, Arab women = 532 Coeffcent -.46 (omtted) (omtted) * Probablty of labor force partcpaton among Jewsh women = Probablty of labor force partcpaton among Arab women = Dfference: Total explaned: * denotes % sgnfcance; ** denotes 5% sgnfcance; denotes percent sgnfcance Source: 25 Socal Survey. Labor force partcpaton regresson: -partcpated; -dd not partcpate S.D

29 ARAB WOMEN IN THE ISRAELI LABOR MARKET: CHARACTERISTICS AND POLICY PROPOSALS 29 Table 3e Arab and Jewsh Women Labor Force Partcpaton Regresson, Farle Decomposton General Varables Varable Age (control: 35-39) Years of schoolng (control: -2) Martal status (control: marred) Dvorced Wdowed Sngle N chldren (control: ) Other wage earners n household (control: none) 2+ Localty type (control: 5.-, populaton Jerusalem Hafa Tel Avv 2,-5,populaton Rural localty Satsfacton wth publc transt (control: do not use publc transt) Very satsfed Satsfed Not so satsfed Not satsfed at all Total observatons = 3,838 Total, Jewsh women = 3,36 Total, Arab women = 532 Coeffcent (omtted) ** ** ** ** ** ** ** * * S.D Probablty of labor force partcpaton among Jewsh women = Probablty of labor force partcpaton among Arab women = Dfference: Total explaned: * denotes % sgnfcance; ** denotes 5% sgnfcance; denotes percent sgnfcance Source: 25 Socal Survey. Labor force partcpaton regresson: -partcpated; -dd not partcpate

30 3 ISRAEL ECONOMIC REVIEW Table 3f Arab and Jewsh Women Labor Force Partcpaton Regresson, Farle Decomposton Most Varables Varable Coeffcent Age (control: 35-39) * ** Years of schoolng (control: -2) ** Receves beneft -.72 Martal status (control: marred) Dvorced.23 Wdowed.6 Sngle. N chldren (control: ) * 4.22 ** 5.9 ** 6.66 * Relgon (control: Muslm) Chrstan -.33 Druze (omtted) Profcent n Englsh -.87 * Has drver's lcense.63 Uses computer.67 Atttudes Man and woman retre at same age.32 Both work.9 In famly wth young chldren: One parent should work less.268 Both parents should work less.7 Both parents should work as usual.26 Total observatons = 3,838 Total, Jewsh women = 3,36 Total, Arab women = 532 Probablty of labor force partcpaton among Jewsh women = Probablty of labor force partcpaton among Arab women = Dfference: Total explaned: * denotes % sgnfcance; ** denotes 5% sgnfcance; denotes percent sgnfcance Source: 25 Socal Survey. Labor force partcpaton regresson: -partcpated; -dd not partcpate S.D The analyss shows that n the Socal Survey, as n the fndngs elcted by the regressons based on the Labor Force Surveys, the dfferent dstrbuton of the standard

31 ARAB WOMEN IN THE ISRAELI LABOR MARKET: CHARACTERISTICS AND POLICY PROPOSALS 3 explanatory varables such as educaton, age, martal status, number of chldren, and type of localty (Table 3e) explan only a relatvely small part of the dfferental n partcpaton of Arab and Jewsh women. However, by runnng a regresson on varous cultural varables, ncludng subjectve ones, we found that they explan much (more than 6 percent) of the dfferental n partcpaton rates. Ths outcome corresponds to other studes (Bertrand and Mullanathan, 2) that found subjectve varables hghly effectve n explanng behavoral dfferences among ndvduals. 7. POLICY IMPLICATIONS The labor force partcpaton rate of Arab women n Israel stands at around 2 percent, markedly below that of the rest of the country s populaton and one of the lowest rates n the world. Ths s one of the most mportant challenges that the Israel economy faces n helpng the Arab sector ntegrate nto the economy. Ths stuaton has many mplcatons for both Arab socety and the economy at large. They nclude mparment of the Arab sector s economc development, the persstence of poverty and ntersectoral economc gaps, the perpetuaton of feelngs of estrangement from Israel, and non-fulfllment of the growth potental of the economy generally. Together they underscore the need for a government polcy that wll help Arab women fnd ther place n the labor market. Ths secton outlnes a seres of polcy measures derved from the foregong analyss and performs smulatons of Arab women s partcpaton n the labor force and contrbuton to GDP. a. Polcy Proposals The pcture mpled by the fndngs n the prevous sectons s one of partcularly low partcpaton among Arab women. Furthermore, the analyss reveals much varaton n partcpaton rates. Women who are modern n terms of educaton, martal status, number of chldren, and profcency n advanced sklls, such as Englsh and computer use, tend to partcpate n the labor market. The low partcpaton problem, as stated, focuses on poorly educated women; those lackng profcency n modern sklls; those wth a large number of chldren, especally young ones; those who lve n areas that have problems n accessng publc transportaton; and those who have tradtonal atttudes towards women s work and men s work. The study shows, however, that t s not enough to strengthen explanatory varables such as educaton to nduce an ncrease n partcpaton, because varous frctonal factors mpede the occupatonal ntegraton of Arab women even f ther characterstcs are dentcal to those of Jewsh women. Consequently, polcy measures should attempt to mtgate these frctons. In vew of the fndngs, a range of polcy measures to ncrease Arab women s labor force partcpaton should be appled measures that wll enhance educaton and sklls; reduce the costs of gong out to work; act to change socal norms relatng to women workng outsde the home; mprove publc transport; and mprove physcal access to workplaces.

32 32 ISRAEL ECONOMIC REVIEW In ths secton, we propose a seres of polcy measures that tackle the obstacles to partcpaton that our study found. The measures address dfferent tme horzons; alongsde those of near-term effect, t s mportant to nclude long-term measures whch wll change partcpaton patterns n a sustanable way. (For elaboraton, see Yashv and Kasr (Kalner) 23.). Local employment matchng centers varous NGOs already run these centers n several localtes but they are small n scale. Such centers may provde nformaton, connect employers wth workers, and provde transport servces to work. By dong so, they may help to overcome two sgnfcant hurdles culture and transportaton whch are often nterrelated. 2. Vocatonal tranng the sort that wll equp women wth advanced sklls n consderaton of ther cultural characterstcs and potental employers needs. 3. Subsdzaton of chldcare nvokng classcal polcy measures of reducng costs of gong to work: subsdzng daycare and afternoon care centers, caregvers, etc. 4. Promotng the physcal accessblty of workplaces by nvestng n approprate transport nfrastructure and vehcles a crucal problem among Arabs n Israel s related to geographc concentraton and transport dffcultes. Infrastructure n many Arabs localtes s undeveloped, prmarly n terms of publc transportaton. When t s hard to get to work, t s no wonder that partcpaton and employment are low, especally among women. The dffculty at ssue, whch relates to both nter-urban and ntra-urban travel, relates to poor road nfrastructure and severe lack of means of transport. What s needed, then, s to enhance physcal access by nvestng n approprate transport nfrastructure and vehcles. In the near term, a government subsdy for potental workers, employers, and transport companes may be helpful. In the longer term, these nfrastructures need sgnfcant mprovement to brng down the hgh cost of gong to work. It s also mportant to nvest n brngng workplaces to the exstng centers. Encouragng frms to stuate themselves n geographc areas accessble to Arabs may do much to promote Arab employment generally and that of Arab women partcularly. 5. Occupatonal gudance n school preparng currcular materals at the hgh school level to enhance awareness of labor-force partcpaton and women s rghts, among both women and men. The ntegraton of local leadershp nto the program wll mprove the chances of success. 6. Subsdzng educaton and hgher studes n the long term, the soluton to ncreasng women s partcpaton s rooted n the educaton system. Basc and advanced studes are fundamental and hghly sgnfcant n all matters related to the performance of the labor market: the partcpaton decson, occupatonal choce, productvty on the job, etc. More should be nvested at all levels of the Arab sector s educaton system. 7. Legslaton aganst dscrmnaton and tougher enforcement experence abroad shows that developed countres tend to have both ant-dscrmnaton laws and strateges geared to make the publc more aware of dscrmnaton, the laws aganst t, and the rghts of ndvduals who fall vctm to t. It s mportant for Israel to broaden legslaton For addtonal polcy measures, see also Jabarn (27) and Shtewy (28).

33 ARAB WOMEN IN THE ISRAELI LABOR MARKET: CHARACTERISTICS AND POLICY PROPOSALS 33 that relates to employment dscrmnaton aganst Arabs and to strengthen the enforcement of ant-dscrmnaton laws n general. b. Smulaton of partcpaton rates and contrbutons to GDP How much utlty should we expect the pro-employment measures to delver? It should be borne n mnd that the partcpaton rates of Arab women have been rsng persstently over tme, even wthout specfc polcy measures that target Arab women explctly. What we can do, then, s perform smulatons of reasonable future scenaros. We need to remember, however, that any exercse of ths knd has an nherent element of uncertanty. Ths uncertanty exsts because changes n fertlty, educaton, and cultural and socal patterns, as well as economc ncentves, all affect populaton ncrease and growth n partcpaton, meanng that the uncertanty exsts at many levels. The followng basc scenaro assumes that the ncrease n the past four decades wll contnue,.e., t ndcates what may be expected f the exstng trajectory perssts. Two alternatve scenaros credt the measures lsted above wth favorable effects. Thus, accordng to these scenaros, the proposed measures wll enhance partcpaton enough so that the rate of ncrease n Arab women s employment wll exceed the exstng trajectory. As t s mpossble to quantfy the exact effect of each of the measures proposed above, we chose two scenaros that specfy, smply, a future ncrease that s.5 or.7 tmes greater than the ncrease that the current trajectory projects. To put matters nto perspectve, we compared these scenaros wth the target of Arab women s partcpaton n 22, ncluded n the statement of recommendatons of the Employment Polcy Commttee (June 2) apponted by the Mnster of Industry, Trade, and Labor. The smulaton s based on two alternatve forecasts for the rate of ncrease n women s partcpaton n the labor market (one from the CBS and one we calculated ourselves); t assumes that f these women were hred, ther per-capta output would equal the average output per worker n Israel wth some adjustments (see below). Fgure a plots the ncrease n the populaton of Arab women aged 5+. Accordng to the CBS projecton ( mddle scenaro ), the annual growth rate s forecast to be 3.5 percent between 2 and 25, 3. percent n 25 2, 2.6 percent n 22 25, and 2.3 percent n Accordng to our forecast, the growth rates are 3.3 percent n 2 5, 3.4 percent n 25 2, 3.4 percent n 22 25, and 2.8 percent n Fgure b charts the ncrease n the number of women partcpants on the bass of two alternatve forecasts, both predcated on the actual number of partcpants n 97 2 and a projecton of ths number nto the future. Techncally, Alternatve A was created by a regresson on the number of partcpants on a thrd-order tme polynomal, and Alternatve B was created by the same regresson n whch the dependent varable s the number of partcpants smoothed by a Hodrck-Prescott flter. Fgure c shows the women s partcpaton rate,.e., the predcted number of partcpants dvded by the predcted populaton. It s based on the forecasts n Fgures a and b. Accordngly, t has four alternatves (not all of whch appear n the graph) n accordance wth the forecasts n the numerator (number of partcpants, two alternatves)