Chinese Business Review Volume 10, Number 11, November 2011 (Serial Number 101) David Publishing David Publishing Company www.davidpublishing.com Publication Information: Chinese Business Review is published monthly in hard copy (ISSN1537-1506) and online by David Publishing Company located at 1840 Industrial Drive, Suite 160, Libertyville, Illinois 60048, USA Aims and Scope: Chinese Business Review, a monthly professional academic journal, covers all sorts of researches on Economic Research, Management Theory and Practice, Experts Forum, Macro or Micro Analysis, Economical Studies of Theory and Practice, Finance and Finance Management, Strategic Management, and Human Resource Management, and other latest findings and achievements from experts and scholars all over the world Editorial Board Members: Moses N Kiggundu (Canada) Polyxeni Moira (Greece) Iltae Kim (Korea) Sorinel C PUŞNEANU (Romania) ZHU Lixing (Hong Kong) Jehovaness Aikaeli (Tanzania) Ajetomobi, Joshua Olusegun (Nigeria) LI Kui-Wai (Hong Kong) Shelly SHEN (China) Chris TIAN (China) Ruby LI (China) Manuscripts and correspondence are invited for publication You can submit your papers via Web Submission, or E-mail to economists@davidpublishing.com, china4review@hotmail.com Submission guidelines and Web Submission system are available at http://www.davidpublishing.com Editorial Office: 1840 Industrial Drive, Suite 160, Libertyville, Illinois 60048 Tel: 1-847-281-9862 Fax: 1-847-281-9855 E-mail: economists@davidpublishing.com Copyright©2011 by David Publishing Company and individual contributors All rights reserved David Publishing Company holds the exclusive copyright of all the contents of this journal In accordance with the international convention, no part of this journal may be reproduced or transmitted by any media or publishing organs (including various websites) without the written permission of the copyright holder Otherwise, any conduct would be considered as the violation of the copyright The contents of this journal are available for any citation, however, all the citations should be clearly indicated with the title of this journal, serial number and the name of the author Abstracted / Indexed in: Database of EBSCO, Massachusetts, USA Ulrich’s Periodicals Directory ProQuest/CSA Social Science Collection, Public Affairs Information Service (PAIS), USA Summon Serials Solutions Chinese Database of CEPS, Airiti Inc & OCLC Chinese Scientific Journals Database, VIP Corporation, Chongqing, P R China Subscription Information: Print $450 Online $300 Print and Online $560 David Publishing Company 1840 Industrial Drive, Suite 160, Libertyville, Illinois 60048 Tel: 1-847-281-9862 Fax: 1-847-281-9855 E-mail: order@davidpublishing.com Chinese Business Review Volume 10, Number 11, November 2011 (Serial Number 101) Contents Financial Forum Effectiveness of the Global Banking System in 2010: A Data Envelopment Analysis Approach 961 Ngo Dang-Thanh Application of Pareto Distribution in Wage Models 974 Diana Bílková The Chaotic Monopoly Price Growth Model 985 Vesna D Jablanovic Marketing Analysis of the Relationship Between Perceived Security and Customer Trust and Loyalty in Online Shopping 990 Nihan Özgüven Industrial Economics Growth Potential and Profitability Analysis of Insurance Companies in the Republic of Serbia 998 Dragana Ikonić, Nina Arsić, Snežana Milošević Regional Economics Sustainable Consumption and Production in the Baltic Sea Region 1009 Janis Brizga, Dzintra Atstaja, Dzineta Dimante Enterprise Management Motifs and Impediments for the Harmonization of Accounting Regulations for Small and Medium-Sized Companies in the EU 1021 Tamara Cirkveni Change of Management Values in Estonian Business Life in 2007-2009 Anu Virovere, Mari Meel, Eneken Titov 1028 Public Economics The Impact of Tax Policies on Taxpayers Budget in Terms of Risk, Sensitivity and Volatility 1043 Boloş Marcel Ioan, Otgon Cristian Ioan, Pop R zvan Valentin Interactions Between Knowledge Sharing and Organizational Citizenship Behavior 1061 Yavuz Demirel, Zeliha Seçkin, Mehmet Faruk Özçınar Social Economics Enhancing Organization’s Performance Through Effective Vision and Mission 1071 Ben E Akpoyomare Oghojafor, Olufemi O Olayemi, Patrick S Okonji, James U Okolie Determinants of Female Employment Rate in the European Union Irena Spasenoska, Merale Fetahu-Vehapi 1076 Chinese Business Review, ISSN 1537-1506 November 2011, Vol 10, No 11, 961-973 Effectiveness of the Global Banking System in 2010: A Data Envelopment Analysis Approach∗ Ngo Dang-Thanh University of Economics and Business (Vietnam National University), Hanoi, Vietnam Massey University, Palmerston North, New Zealand The current crisis has revealed the weaknesses of the global financial in general and its banking system in particular, and puts forward a requirement for assessing the effectiveness and stability of the banking sectors across countries Based on available data from 64 countries over the world, the author tried to evaluate the effectiveness of the banking sectors in those countries through the view point of the data envelopment analysis approach to define how the global banking systems is under the effect of the current crisis Findings from the research showed that banking systems in advanced economies are still more effective than in developing countries Moreover, it explained the effect of the current financial crisis, the role of public finance (and the government), and the development of the (privately) commercial banks to the effectiveness of the banking sectors The research also explained some determinants that can affect the effectiveness of the banking system, including inflation, bank concentration, and level of economic development Keywords: data envelopment analysis, effectiveness, efficiency, banking, cross countries Introduction Because of the important role of the banking and financial system in the rapid development of new industrial economies (NIEs) in the 1960s-1970s, there were renewed interests in the relationship between financial and economic growth Schumpeter (1911) argued that the role of financial intermediaries in savings mobilization, projects evaluation and selection, risk management, entrepreneurs monitor, and facilitating transactions is important to technological innovation and economic growth Following this argument, many other leading economists continuing emphasized the positively essential role of the financial sector in economic development, including Goldsmith (1969), Shaw (1973), McKinnon (1973), King and Levine (1993a, 1993b) Banks are the core of the financial system They accept deposits from savers and lend them to borrowers ∗ Acknowledgement: The author would like to offer special thanks to Professor David Tripe at Centre for Banking studies, Massey University, New Zealand for his supports, encouragement and useful comments The author also thanks participants at the 18th Annual Global Finance Conference in Bangkok, Thailand, April 2011 for their constructive comments and feedback to improve the quality of the paper The usual disclaimer applies Ngo Dang-Thanh, Ph.D candidate, Lecturer, Faculty of Political Economy, University of Economics and Business (Vietnam National University), Centre for Banking Studies, Massey University Correspondence concerning this article should be addressed to Ngo Dang-Thanh, Faculty of Political Economy, University of Economics and Business (Vietnam National University) E-mail: ndthanhf@yahoo.com 962 EFFECTIVENESS OF THE GLOBAL BANKING SYSTEM IN 2010 They hold liquid reserves which allowing predictable withdrawal demand They issue liabilities which are more liquid than the deposits They also reduce (or some times eliminate) the need of self-finance (Bencivenga & Smith, 1991, p 195) Banks hold an important role within the financial system, and to some certain level, researching the banking system therefore means researching the financial system Started from the bankruptcy of the Northern Rock Bank in the UK (2008, February), however, the global financial crisis and its heavily impacts have put researchers and policy makers under the requirement of re-assessment and re-evaluation the stability and performance of the global financial and banking system1 A firm is effective when it reaches its target outputs Similarly, a banking system is defined as effectiveness if it can fulfill its missions of providing banking services and monitoring the stability of the system Meanwhile, if banking systems are set under similar conditions of macro- and micro-economic, the level of outcomes that a banking system can provide (in term of services and stability) is indeed its efficiency In this sense, the problem of calculating effectiveness of banking systems all over the world becomes the problem of evaluating its efficiency with a (dummy) similar and equal input This research is trying to define the effectiveness of the global banking system in 2010 through analysing cross-country data observed from 64 countries, using the data envelopment analysis (DEA) approach The remainder of this paper is organized as follows Section gives some reviews on efficiency and effectiveness evaluation in the banking sector using DEA approach Section explains the methodologies and technical will be applied in the research Section shows empirical results and section concludes Literature Review To evaluate the efficiency of a set of firms (or banks), the most popular approaches are ratio analysis, parametric analysis and nonparametric analysis (the latter two methods belongs to the X-efficiency approach) While ratio analysis focuses on ratios between two variables (of inputs or outputs) to define the productivity and efficiency, X-efficiency analysis evaluates the efficiency of a bank through a multi-variables aspect DEA is a popular nonparametric method applied in evaluating efficiency in finance and banking area After Farrell (1957) laid the foundation for a new approach in evaluating efficiency and productivity at micro-level, Charnes, Cooper and Rhodes (1978) and then Banker, Charnes and Cooper (1984) developed the CCR and BCC-DEA model, respectively, to evaluate the (relative) efficiencies of the researched decision making units (DMUs) Since then, DEA was increasingly applied in efficiency evaluation, especially in social sciences2 There are a limited number of researches using DEA to examine banking performance at cross-country level A study in 1997 showed that out of 130 studies on banking performance and efficiency, only six were focused on comparing the efficiency level of banking systems across countries (Berger & Humphrey, 1997, pp 182-184) As shown in Table 1, all three DEA studies were using small sample data at institutional (bank) level to define the benchmark frontier, hence, the global banking system was left untouched In the 2000s, further studies which used common frontier approach were developed by add in the model According to Science Direct, since 2010, there are more than 2,200 journal articles regarding banking performance after the crisis of 2007-2008 (Retrieved December 20, 2010, from http://www.sciencedirect.com) Recent study of Avkiran (2010) showed that there are more than 170 articles using DEA as a main methodology to analyse the efficiency of banks and banks branches, including Sherman and Gold (1985), Peristiani (1997), Schaffnit, Rosen and Paradi (1997), and Pastor, Knox Lovell and Tulkens (2006) 963 EFFECTIVENESS OF THE GLOBAL BANKING SYSTEM IN 2010 some environmental/controllable variables such as banking market conditions or market structure and regulation (Kwan, 2003; Lozano-Vivas, Pastor, & Hasan, 2001; Maudos, Pastor, Perez, & Quesada, 2002; Sathye, 2005) However, as they are also mainly focused on institutional level data while macro-environment is different from country to country, they ignored that banks which are efficient in this country may not performance well if they run as foreign-owned banks in other countries (Berger, 2007, p 125) Hence, while trying to examine the whole banking systems across countries, this study attempts to overcome the above problem Table Studies on Banking Performance at Cross-Country Level (Prior to 1997) Authors (date) Berg, Forsund, Hjalmarsson, & Suominen (1993) Fecher & Pestieau (1993) Bergendahl (1995) Ruthenberg & Elias (1996) Bukh, Berg, & Forsund (1995) J Pastor, Perez, & Quesada (1997) Method used Countries included Institution Data envelopment analysis Norway, Sweden, Finland Bank Distribution free approach Mixed optimal strategy Thick frontier approach Data envelopment analysis Data envelopment analysis 11 OECD countries Norway, Sweden, Finland, Denmark 15 developed countries Norway, Sweden, Finland, Denmark 08 developed countries Financial service Bank Bank Bank Bank Note Source: Berger and Humphrey (1997) As DEA evaluates the efficiency of each DMU based on the optimal multipliers (or weights) of inputs and outputs factors, it allows us to examine the effectiveness of a banking system by looking at the achievements of the banking sector, including both quantity (assets, deposits, credits, etc.) and quality (overhead cost, nonperforming loans, frequency of bank crises, etc.) factors of commercial banks in the economy3 They are chosen following 122 variables represent the stability of the global financial system (WEF, 2010, Appendix A) However, since DEA treats those factors dynamically (meaning each country can have its own preference on them), to be understandable in evaluating and comparing the effectiveness of the banking systems between countries, a common preference (or common set of weights) for the above analyzed factors is required Therefore, in this research, the DEA model will be divided into three stages, in which the first stage conducts a dynamic DEA model (DSW model) to define the relatively efficiencies of the banking systems from these 64 countries; the second stage examines the determinants affecting that efficiencies (Tobit model); and the third stage defines the common set of weights for those analyzed factors (CSW model) in order to conduct the final banking effectiveness scores Technical Methodologies In the first step, DSW model is produced to calculate the maximum effectiveness scores that each country can achieve with the observed (achievement) factors Mahlberg and Obersteiner (2001) and Depotis (2004) developed an input-oriented DEA-like model which treats all factors as outputs, while input is a dummy variable (values equal to for all countries) Therefore, the DSW model in this research is in fact a constant-returns-to-scale (CRS) and input-oriented DEA model For an evaluated country j0-th, its efficiency score (DSWj0) can be expressed by the following non-negative linear problem: It is important to notice that these factors are outcomes that a banking system is aiming for; hence, the DEA model in this paper will use them all as output variables 964 EFFECTIVENESS OF THE GLOBAL BANKING SYSTEM IN 2010 DSWj0 = max ∑u y ∑v x m mj (1) k kj Subject to: ∑u m ymj ≤ ∑ vk xkj , ≤ j ≤ n ∑v x = ∑u = k kj m um ≥ xj = {all original input values are assumed to be equal to 1} where: um: weight of m-th output factor; vk: weight of k-th input factor; xkj: k-th input of j-th country, k = 1; ymj: m-th output of j-th country; n: number of countries; m: number of factors Due to the fact that some countries can have the same scores in this DSW model, a super efficiency DEA model (Zhu, 2001) is also ran to determine the ranking order of the researched countries, makes it easier to compare the effectiveness’s of the banking systems between countries In the next step, a Tobit regression (for more details, see Tobin, 1958) is used to determine the factors affecting the country’s banking efficiencies (Tobit model) Since the CSW scores are bounded between to 1, non-censored regression models could be biased (Fethi & Pasiouras, 2010), while Tobit regression is justify as in equation (2) Variables used in this model are ones that mainly related to the financial efficient of a banking system at micro-level and are expressed in Table EF = + 1*CONC + 2*ROA + 3*ROE + 4*CIR + 5*INF + 6*CTA + 7*NIM + 8*CII + 9*GROUP (2) Table Variables of the Tobit Model Variables EF CONC ROA ROE CIR INF CTA NIM CII GROUP Definition CSW-DEA scores Bank concentration (assets of three largest banks as a share of assets of all commercial banks) Bank’s average return on assets (Net income/Total assets) Bank’s average return on equity (Net income/Total equity) Bank’s cost to income ratio (Total costs as a share of total income of all commercial banks) Inflation, consumer prices (annual %) Bank’s capital to assets ratio (ratio of bank capital and reserves to total assets) Net interest margin of banks (value of bank’s net interest revenue as a share of its interest-bearing assets) Depth of credit information index (measures rules and practices affecting the coverage, scope and accessibility of credit information) Dummy variable of income group (equals to if country belongs to lower income, if middle income, and if high income group) EFFECTIVENESS OF THE GLOBAL BANKING SYSTEM IN 2010 965 The last step is to define the optimal common set of weights which should be used for compare and ranking countries based on their banking systems’ effectiveness It is done by applying the CSW model It is believed that the efficient frontier found in the DSW model in the first step is the “best practice frontier” (Grosskopf & Valdmanis, 1987; Schaffnit, Rosen, & Paradi, 1997); hence, the optimal common weight set will be the one that get every countries’ performances closest to that frontier There are several ways to define that common set of weights is based on this idea While imposing bounds for factor weights, Roll and Golany (1993) found out that the common set of weights can be defined by maximizing the average efficiency of all DMUs or maximizing the number of efficient DMUs Kao and Hung (2005) applied a compromise solution approach to minimize the total squared distances between the optimal objective values (found by DEA) and the common weighted values (found by using common set of weights) Jahanshahloo, Memariani, Lotfi and Rezai (2005) applied the multiple objective programming approach to simultaneously maximize the performance scores to get it closes to the “best practice frontier” Liu and Peng (2008) applied the common weights analysis to minimize the vertical and horizontal virtual gaps between the benchmark line (slope equals to 1.0, or performance scores equal to 1.0) and the coordinate of common weighted DMUs In this paper, we modified the model of Kao and Hung (2005) into a minimum distance efficiencies model, in which the common set of weights can be defined as the one minimizing the total distances between optimal efficiencies (DSW scores) and common weighted scores (CSW scores) of all DMUs, under the condition that each DMU’s efficiency cannot exceed its DSW efficiency4 To understand the role of each factor in CSW scores, another condition was added where the total sum of weights is equal to (or 100%) The country’s banking effectiveness scores will be constructed based on that CSW scores and findings from the super efficiency DEA results in the previous step This CSW model can be expressed as a non-negatively linear problem as follows: ( ∑ e*j − e j ) (3) Subject to: e*j = DSWj ej = ∑u y ∑v x m mj k kj ,1≤j≤n e j ≤ e*j ∑v x = ∑u = k kj m um ≥ 0.015 xj = {all original input values are assumed to be equal to 1} where: um: weight of m-th factor; ymj: m-th factor of j-th country; This constrain makes these distances non-negative, hence, they can be used directly rather than the squared distances Mahlberg and Obersteiner (2001) found that restriction weights with lower bound of 0.01 steered a middle course between too strong predetermination and too large flexibility DETERMINANTS OF FEMALE EMPLOYMENT RATE IN THE EUROPEAN UNION 1077 by 2010 and to increase the percentage of women in employment from an average of 51% in 2000 to more than 60% by 2010 The new goal required member states to consider setting national targets for an increased employment rate, which it was argued by enlarging the labor force would reinforce the sustainability of social protection systems In addition to the 2010 Lisbon targets, the Stockholm Europen Council of 2001 set intermediate targets for employment rates in the EU in 2005 of 67% overall and 57% for women It also set a new target for raising the average EU employment rate for older men and women (aged 55-64) to 50% by 2010 (Retrieved from http://www.europa.eu, summary of legislations) Progress of Total and Female Employment Rate After the Lisbon Strategy The progress towards the Lisbon employment targets of overall employment (70%), female employment (60%) and employment of older man and women (50%) can be seen in Table Table Progress of Total and Female Employment Rate From 2000 to 2009 in Percentages 1997 Employment rate % overall 70% (2010 target, Lisbon summit) 60.7 Employment rate 55-64 years old % overall 50% (2010 target, Stockholm summit) 36.4 Female employment rate 15-64 years old 60% (2010, Lisbon summit) 50.8 GDP growth percentage change in previous year Base line scenario of 3% per annum, Lisbon summit 2.5 2000 63.4 37.5 54.1 3.9 2001 64.1 38.4 55.0 1.9 2002 64.2 39.8 55.6 1.2 2003 64.5 41.5 56.2 1.2 2004 64.8 42.3 57.0 2.3 2005 65.4 44.2 57.8 1.8 2006 66.2 45.3 58.8 2.9 2007 67.0 46.5 59.7 2.7 2008 67.3 47.4 60.4 0.6 2009 64.6 46.0 58.6 2.5 EU Benchmark Notes The employment rate is calculated by dividing the number of persons aged 15 to 64 in employment by the total population of the same age group The indicator is based on the EU Labour Force Survey The survey covers the entire population living in private households and excludes those in collective households such as boarding houses, halls of residence and hospitals Employed population consists of those persons who during the reference week did any work for pay or profit for at least one hour, or were not working but had jobs from which they were temporarily absent Source: EUROSTAT—European Commission Statistics First three years of the decade after the launch of the Lisbon strategy were characterized with only a moderate growth in employment The slowdown which began in the first half of 2001 and saw growth reached a standstill by the last quarter of 2002, followed by only a very moderate recovery over the course of 2003 With nearly zero growth in 2003 the progress towards the Lisbon 2010 target of 70% overall employment had came to a standstill By 2003, it became clear that the EU was going to miss the intermediate employment rate target for 2005 although the employment rate target for women still remained in reach In 2006, 13 member states have meet the 2010 employment target for women including for the first time Cyprus, Germany, Latvia and Lithuania Since 2000, large increases have been achieved in Cyprus, Estonia, Greece, Latvia and Italy where rates have risen by around percentage points and Spain with 12 percentage points Greece and Italy were still far from the 1078 DETERMINANTS OF FEMALE EMPLOYMENT RATE IN THE EUROPEAN UNION target Total employment continued to expand in the EU in 2008 However, the rise was not uniform with respect to gender, age and type of employment The EU employment rate, i.e., the share of the population aged 15-64 years (the working-age population) in employment, amounted to 65.9% in 2008, up 0.5 percentage points compared with 2007 The growth rate for female employment was almost three times that for male employment In 2008, the employment rate of women aged 15-64 amounted to 59% , while for man it was almost 73% In 2009, the employment rate for persons aged 15-64 was above 70% while in all member states the male employment rate was higher than the female employment in 2009 except for Lithuania where the female rate was percentage point higher for men Estonia and Finland have recorded the smallest difference between male and female employment rates while Malta, Greece and Italy recorded the greatest The aim of the empirical testing in this study is to capture the common institutional determinants of female employment rate by employing the generalized method of moments (GMM) proposed by Arellano-Bond (1991) and Arellano-Bover (1995)/Blundell-Bond (1998) We have considered dynamic model estimation as a more appropriate approach because of the possibility to capture the costly sluggish adjustment of employment as well as the effect of persistence captured by adding a lagged dependent variable Moreover, we were also interested whether there will be a change in the sings in front of the coefficients especially for fertility rate and maternity leave which traditionally have been taken as factors that negatively influence female employment rate The contribution of this analysis can be seen in a manner that highlights the relationship and significance of the institutional variable towards improvement of the social and welfare system across countries members of the European Union as the backbone for supporting female in perusing carriers and higher employment The study is organized as follows In the next section, we provide a brief yet detail insight to literature review for each variable included in the research Empirical specification diagnostics and results are given in the third section followed by the fourth where we discuss our findings and conclusion Literature Review and Variable Description In general there are two main approaches employed across studies of female employment across countries The first approach examines the level of female employment itself, focusing on the labour force participation rate and the number of hours worked such as part-time to full-time category The second approach explores the determinants of female employment While the determinants selected for analysis vary from study to study, they may include one or more of the following three types of variables: (1) individual level (micro variables, such as number of children in household); (2) institutional level (macro) variables, such as the size of the welfare state; and (3) a combination of individual and institutional level variables (Warnecke, 2008) While single-country level studies use micro-level variables, macro-level variables focus on family policies as diverse social, political, institutional and cultural constraints of the average female participation rate In the following we have provided a brief literature review by analyzing the institutional level (macro) variables used in this study Maternity Leave Parental leave policies support new parents usually in two complementary ways: by guaranteeing job-protected leave and by offering financial support during that leave The aim of benefits is to subsidize the child care provided by the mother while job protection aims to ensure continuity of women’s careers in the labor market Most of the literature has found that more generous parental leave mandates tends to delay women’s DETERMINANTS OF FEMALE EMPLOYMENT RATE IN THE EUROPEAN UNION 1079 return to work However, evidence of the relationship between duration of leave and women’s labor market outcomes is mixed Job protected leave mandates are expected to increase women’s employment and earnings by encouraging job continuity after birth Yet, prolonged periods of absence from the workplace might lead to loss of specific and general human capital and weaker labor market prospects after returning to work (Voicu & Hielke, 2003) Hence previous employers, while obliged to re-employ mothers when they return to work after the baby break, may either remunerate them relatively worser than their colleagues or may dismiss or layoff re-entered women with a higher probability as soon as the job protection period upon re-entry has run out (Lalive, Schlosser, & Zweimuller, 2009) Ruhm (1998) compares employment rates and wages of men and women using panel data from European countries, and finds that longer leave mandates are associated with higher female employment but lower relative wages Ejrnaes and Kunze (2006) investigate the role of parental leave on the family wage gap using administrative data for Germany and exploiting exogenous variation in the length of parental leave generated by policy changes in the German system The authors found that longer parental leave duration leads to detrimental effects on employment and wages In contrast, Schonberg and Ludsteck (2008) study the same reforms and found only minor effects on employment rates and a mixed effect on wages In our analysis maternity leave is defined as the number of paid weeks a woman is entitled in case of normal birth (International Social Comparison Database) Child Care Facilities Theories about child care (e.g., Andersson, Duvander, & Hank, 2004; Rindfuss & Brewster, 1996) include four dimensions of child care: availability, quality, cost, and acceptability In our case, we are testing the impact of the availability upon female employment rate Availability is the degree to which a family has ready access to the needed child care—This might include not only convenient geographical location but also the availability of slots for the right age range and the right time of day (Rachel, 2001) Our measure of availability is the participation rate of years old children in pre-school education This indicator presents the percentage of the years old who are enrolled in education-oriented pre-primary institutions which provide education-oriented care for young children They can either be schools or non-school settings, which generally come under authorities or ministries other than those responsible for education As such it is a measure of utilization rather than capacity When estimating the child care effect on female employment rate, we have considered possible simultaneity that gives rise to endogeneity in women’s decision to work and in the decision to use institutional child care According to Coneus, Goeggel and Muehler (2009) the decision to use child care outside the home is strongly connected to mothers’ decision to work after child birth and vice versa They provide evidence on the determinants of institutional child care use addressing the endogeneity of mothers’ labour supply by applying an instrumental variable approach Based on the German Socioeconomic Panel from 1989 to 2006 they have shown the children have a higher probability to attend institutional care if their mothers increase their actual weekly working time On the other hand, limited availability in the form of limited slots and hours a facility can remain open limits compatibility with the mother’s working hours Kreynfeld and Hank (2000) have found that due to the very limited opening hours mothers using child care may not even be able to work part time and must seek additional forms of child care, which are rarely available Greater availability of child care it is associated 1080 DETERMINANTS OF FEMALE EMPLOYMENT RATE IN THE EUROPEAN UNION positively with female’s employment (Van Dijk & Siegers, 1996) Therefore, a positive sign is expected in front of the coefficient (Eurostat Database) Fertility Rate The fertility rate is defined as the mean number of live children born to a women during her lifetime Traditionally, a high fertility rate has a negative impact upon female employment However, recent studies have shown that the current relationship between female employment rate and fertility rate is somewhat ambiguous and has been undergoing constant change in recent years Due to institutional changes in form of subsidies, child care programs, possibility of part-time working and unemployment rates might have changed the sign in past decades (DelBoca & Saures, 2007) The correlation between fertility and female employment across developed countries was negative, significant and quite stable during the 1970s and up to the early 1980s However, by the late 1980s, the correlation had become positive and equally significant (Ahn & Mira, 2002) The reversal of the sign occurred simultaneously with the emergence of a high and persistent unemployment rate, increasing dispersion in the availability of part-time jobs, child care availability and job protection provided by paid maternity leave Where the unemployment rate is high, women are less likely to leave the labor market, knowing that it may be more difficult to reenter later due to the scarcity of jobs (Del Boca, 2003) Additionally, when there is greater insecurity in the labor market, parents may be more reluctant to have children because of the fear of not having enough income to support their potential children (Del Bono, 2002) The greater availability of part-time job opportunities within a country reduces the opportunity cost of having children, as mothers will be less likely to give up their jobs to raise their children In a carefully conducted empirical study based on provincial-level Italian panel data, Del Boca (2002) documents that availability of child care and part-time work increases both the probability of working and having a child Many researchers have suggested the increasing availability of market child care as a possible explanation for the recent fertility upswing in some developed countries Moreover, countries with longer maternity leave programs have significantly higher fertility rates than countries with shorter maternity leave policies Based on a cross-sectional time-series data for the European Union, DiCioccio and Wunnava (2008) found that neither female education nor increased employment were significant in determining fertility and vice versa When estimating the fertility effect on female employment, it is important to recognize the mutual dependence between the labour supply of married women and fertility, i.e., the endogeneity in either life cycle models or static models of female labour supply Even though children can be exogenous to the hours of work decision for married women, according to Xie (1997), children are endogenous to the female participation decision where children under six have dramatic negative impact on female employment However, based on the findings that government regulations and job protection has mitigated the effect of fertility upon employment rate, we might expect downsizing of the negative effect upon employment or even an insignificancy of the coefficient (World Bank Database) College Education In order to estimate the effect of education upon female employment, the percentage of graduated females with tertiary education is used, irrespective of fields of education such as mathematics, science, computing, engineering, manufacturing and construction Greater levels of educational attainment theoretically afford DETERMINANTS OF FEMALE EMPLOYMENT RATE IN THE EUROPEAN UNION 1081 females a greater range of employment possibilities and a greater choice of superior (more “desirable”) employment, e.g., higher paying jobs, thereby creating an incentive to enter the labour force Therefore we can expect a positive relationship between the percentage of female population with college education and female employment (Gerner & Zick, 1983; Rextorat, 1990; Miller & Xiao, 1999; World Bank Database) Female Unemployment Rate The female unemployed rate is female unemployed persons as a percentage of the female labour force based on International Labour Office definition Unemployed persons comprise persons aged 15 to 74 who are without work and have been actively seeking work This variable captures the labour market conditions and the expectation of obtaining gainful employment Unfavorable market conditions, i.e., a high female unemployment rate negatively influences the female employment rate, therefore we can expect a negative sign in front of the coefficient (Eurostat Database) Female Part-Time Employment Female part-time employment is defined as the percentage of total female employment The rise in female participation has occurred hand-in-hand with an increase in the part-time rate in many countries While the causality is not clear between the decision to participate into the labour market and the choice of working part-time, the rising proportion of women willing to join the labour markets, mostly explained by rising levels of education, contributed to the development of part-time employment Sociological and cultural reasons, such as the separation of tasks within the household and the family model, combined with institutional reasons (e.g., the lack of childcare facilities) explain in part why women are more inclined to work part-time than men For instance, a “male breadwinner” model of family encourages women to work part-time rather than full-time (Fagan & O’Reilly, 1998) According to Budelmeyer, Mourre and Ward (2008), part-time work creates an opportunity for women to combine taking care of their children with market work Therefore a positive sign is expected (Eurostat Database) Growth of GDP per Capita This is defined as the annual percentage growth per capita According to the Neoclassical Endogenous Growth theory, the rate of growth is endogenous in a sense that is driven by the rate of growth of the labour force i.e., employment and the technological change (Setterfield, 2009) At the same time, a higher employment rate implies an unambiguous increase in GDP per capita with no negative implications for long-run productivity growth in the existing workforce (Carone, Denis, Morrow, Mourre, & Roger, 2006) The growth rate in labour productivity is the most important determinant of the growth of GDP per capita as it accounts for at least half of GDP per capita growth in most OECD countries Greater labour utilization is the factor that can make an important contribution to GDP growth by providing a significant boost to the annual growth It is expected a positive sign in front of the coefficient (World Bank Database) describes the relative judgment of oneself in comparison with others Overprecision is “excessive certainty regarding the accuracy of one’s beliefs” (Moore & Healy, 2008, p 4) Overconfidence has been observed in social judgments, self-predictions, and professional predictions, in retrospective as well as prospective judgments (Allwood & Granhag, 1999; Dunning, Griffin, Milojkovic, & Ross, 1990; Lichtenstein & Fischhoff, 1977; Paese & Feuer, 1991; Vallone, Griffin, Lin, & Ross, 1990; Von Winterfeldt & Edwards, 1986) With regard to confidence judgments about achieving future goals, 1082 DETERMINANTS OF FEMALE EMPLOYMENT RATE IN THE EUROPEAN UNION this calibration deficit implies that decision makers systematically overestimate their capacity to reach stated goals Empirical Specification Descriptive Statistics and Model Specification In this chapter, we test the significance of selected institutional variables on a county’s female employment rate by conducting a first-order dynamic panel estimation Dynamic panel estimation has been used since static modeling of an employment equation may lead to problems According to Lachenmaier and Rottmann (2007), the high cost of hiring and firing are a well-known argument for costly employment adjustment, especially in European economies If a firm faces these high costs, the actual employment will deviate from the equilibrium level in the short run The short-run dynamics compound the influences from adjustment costs, expectation formation and decision processes Therefore, a dynamic panel data model is considered in order to model the sluggish adjustment Dynamic panel models estimate the effects on some observed outcome of other variables of interest, which may be exogenous or potentially endogenous, conditional on both unobserved individual heterogeneity and one or more lags of the dependent variable1 In our case apart from the dependent variable, female employment rate, three individual variables were taken as endogenous variable: fertility rate, GDP per capita and childcare facilities To get consistent estimates in the presence of lags of the dependent variable we employ the generalized method of moments (GMM)2 proposed by Arellano-Bond (1991) and Arellano-Bover (1995)/Blundell-Bond (1998) introduced also by Roodman (2006) Applying GMM we account for the potential endogeneity arising from the lagged dependent variable as well as three above mentioned variables Using the appropriate instruments for the endogenous variables one can overcome the endogeneity problem although using lags two and deeper for the endogenous variable in the GMM-style and all regressors included in the RE increases the number of instruments and too many instruments “can overfit endogenous variables” (Roodman, 2006, p 13) According to Roodman (2006) dynamic panel estimation is a proper approach for situations with few time periods, say ≤ 15, and many cross-section units where the number of units is far greater than the number of time periods However, in our case we are going to use a panel data with relatively long time period and considerably small number of cross-sections, i.e., 25 out of 27 EU countries for 13 years starting from 1995 In panels where T is large, the dynamic panel bias becomes small and a more straightforward fixed effect estimator works However, large bias has been found even for T = 30 and our data has a considerably smaller time dimension than this At the same time, the number of instruments in difference and system GMM tends to explode with T If N is small, the cluster-robust standard errors and the Arellano-Bond autocorrelation test may become variables (Geoff Pugh, The basic characteristics of the linear dynamic panel model are displayed in the following equation: Yit = Yi,t-1 + ( i + 竺it) It is a first-order dynamic panel model, because the explanatory variables on the right-hand side include the first lag of the dependent variable (Yi,t-1) where the group-specific random effect ( i) control for all unobserved effects on the dependent variable that are unique to the country and not vary over time, i.e., captures specific ignorance about country i and an error that varies all over countries and time (竺it) capturing the general ignorance of the determinants of Yit GMM is a general method of estimating population parameters from a data sample GMM assumes population conditions expressed in terms of expectations, i.e., E(竺t, xt) = which is a restriction on the covariance between the error term and the independent variable known as conditions GMM dynamic panel estimation allows use of set of instruments per variable within the data which give a great possibility for resolving endogeneity problems within the model DETERMINANTS OF FEMALE EMPLOYMENT RATE IN THE EUROPEAN UNION 1083 2009) Despite the expected problems, we continue to investigate dynamic model out of economic reasons explained at the beginning of the chapter Hence this chapter should be seen in terms of illustrating an approach that may be more appropriate for data set with a larger cross-section dimension Female employment ratei,t = f(em_ratei,t-1, matleavei,t, chcarei,t, fertratei,t, educ i,t unemrate i,t part_emi,t, GDPi,t, time dummy, error term) The chosen variables in the model given above have been considered in the literature as possible determinants of the female employment rate The dependent variable is the female employment rate which is females in employment as a percentage of the total female population over 15 years age and up to 64 years age (Eurostat Database) In order to obtain more valid result we have included time dummy variables in the model which will capture and place the effect of universal shocks (business cycle effects, demand shock, etc.) from the idiosyncratic error term in to the systematic part of the model According to Roodman (2006), contemporaneous correlation, as a result of universal time-related shocks, can cause cross-individual correlation in the error term which may give biased estimates and the addition of such time dummies may lessen or remove such correlation The abbreviation used in the regressions as well as brief summary statistics are given in Table Table Descriptive Statistics Variables Abbreviation Maternity leave matleave Child care facilities chcare Fertility rate fertrate College education educ Female unemployment rate unemrate Part-time employment rate part-em Standard deviation Mean value Min 8.50 19.41 13 19.63 78.07 29.7 Max 52 100 0.24 1.48 1.09 10.56 25.89 2.8 54.2 2.13 4.75 9.23 2.2 30.8 16.14 23.43 2.7 75.2 31.6 74.3 -4.58 12.23 Female employment rate em_rate 9.40 55.86 Growth in GDP per capita GDP 2.37 3.47 Empirical Results In order to estimate the significance of female employment rate determinates, we have considered dynamic model estimation as a more appropriate approach since dynamic estimation gives us the possibility to model the costly sluggish adjustments of employment According to Peseran, Smith and Im (1996), pervasive slope heterogeneity is often evidence in panel-time series for groups such as countries, regions, industries and firms In such panels inclusion of lagged dependent variable in conventional FE and RE model could lead to biased and inconsistent estimates (Peseran & Smith, 1995) Secondly, in RE estimation, the lagged dependent variable is correlated with the compound error term Because the compound error term has a time invariant component, it influences the dependent variable in each period and hence, must be correlated with lagged values of the dependent variable which conflicts with the basic assumption of linear regression (Creen, 2003) In our case, the error component regression model controls for unobservable characteristics such as tradition, employer preferences, i.e., stereotyping, prejudice, work-life balance policies and so on According to the literature, there is a high possibility for these to be correlated with 1084 DETERMINANTS OF FEMALE EMPLOYMENT RATE IN THE EUROPEAN UNION some of the independent variables in the employment rate, e.g., in highly traditional counties, where the bread-winner model is dominant there is higher possibility of a higher fertility rate and less possibility of supporting the female partner to enter the working market and to enroll tertiary education There are two partial solutions for situations where we have slope heterogeneity and an error component correlation One is a group—mean regression which involves reducing panel to a cross section and the second is estimating purely static panel, both eliminating dynamics However, omitting dynamics and estimating a static model in presence of a dynamic relationship entail serious misspecification as well as bias and inconsistent estimates (Bond, 2002a) The importance of modeling dynamics has been strongly emphasized by Green (2008), according to whom, inclusion of lagged variable brings entire history in the right-hand-side of the equation, so that any measured influence is conditional on this history where any impact of the independent variables represents effect of new information Therefore, in order to get consistent estimates in presence of lags of the dependent variable we employ the generalized method of moments (GMM) proposed by Arellano-Bond (1991) and Arellano-Bover (1995)/Blundell-Bond (1998), and developed in Roodman (2006) In the GMM approach, apart from the dependent variable three other variables were classified as endogenous: childcare, fertility rate and GDP per capita However, we have to be aware that classifying these variables as endogenous makes the use of this estimation procedure problematic given our data set since applying more endogenous variables create more instruments, thereby causing bias and inefficient estimates When analyzing cross-sectional data with slightly longer time series, difference estimation is considered as a more suitable approach However, according to the obtained results when comparing both difference and system estimation, in our case the system approach appears as more appropriate Examining the diagnostic statistics (m1 + m2 statistics as well Sargan test), in the system estimated model there is 1st-order serial correlation and there is No 2nd order serial correlation while Sargan test p = 0.2050 On the other hand, diagnostic testing of the Arellano and Bond Difference estimation, indicated weak instruments since Sarnan test p = 1.000 while m1 + m2 statistics suggested presence of 1st order serial correlation well as 2nd order serial correlation On the bases of diagnostic testing, the system estimated model is going to be used as it is preferred over the difference model The obtained results are given in Table When we look at statistics of significance in the difference estimated model, all of the coefficients are statistically, individually insignificant at the conventional 5% critical level Compared to the difference estimated model, the system estimated model has the same statistics as significant, with the exception for three variables: em_rate(lagged), GDP and unemrate Before the interpretation of the system GMM estimation, two tests for instrumental validity have been used: (1) test for first and second order serial correlation among the residuals (m1 and m2 statistics); and (2) the Sargan test of over-identifying restrictions Arrelano and Bond’s (1991) GMM estimation require E[ 竺it, 竺i,t-2] = 0, i.e., no second-order serial correlation in the error term of the first differenced equation, where m2 statistics test the maintained hypothesis (Ho) in the equation above The m1 statistics has a subsidiary role by providing information on the robustness of m2 statistics The m2 statistics is unreliable, i.e., it may fail to reject, if the error term in levels follows a random walk Thus, if there is first order serial correlation in the first differenced error term where < p < 1, the random walk in the first order errors is excluded Therefore, the m1 and m2 statistics require first order serial correlation and No second order serial correlation According to the first way of testing instrumental validity, i.e., test for first and second order of serial DETERMINANTS OF FEMALE EMPLOYMENT RATE IN THE EUROPEAN UNION 1085 correlation among the residuals, in all cases the t-statistics for m2 we accept the null of No 2nd order autocorrelation in the differenced error terms At the same time we reject the null of m1 statistics for 1st order autocorrelation in the differenced error terms Table Comparison Between System and Difference Estimated Models Models Variables Dependent variable: Em_rate Independent variables: Constant Standard Errors; t-statistics Em_rate(lagged) Standard Errors; t-statistics chcare Standard Errors; t-statistics fertrate Standard Errors; t-statistics matleave Standard Errors; t-statistics educ Standard Errors; t-statistics unemrate Standard Errors; t-statistics Part_em Standard Errors; t-statistics GDP Standard Errors; t-statistics m1 pr > z m2 pr > z Sargan/Hansen test Prob > chi2 Wald test Prob > chi2 If TS > CV Reject Ho: the independent variables are jointly zero System dynamic panel-data estimation 11.51 (2.595); (4.43) 0.84 (0.031); (27.28) -0.003 (0.01); (-0.32) -0.05 (0.883); (-0.06) -0.05 (0.027); (-1.72) 0.006 (0.025); (0.39) -0.20 (0.039); (-5.16) 0.02 (0.015); (1.63) 0.20 (0.0054); (3.72) -2.28 (0.20) 1.15 (0.25) 130.41 (0.2050) 2,345.45 (0.000) Difference dynamic panel-data estimation 45.95 (30.72); (1.50) 0.57 (0.177); (3.22) 0.05 (0.039); (0.17) 2.43 (1.811); (1.34) -1.41 (1.36); (-1.04) 0.03 (0.036); (0.87) -0.11 (0.14); (-0.81) -0.07 (0.17); (-0.39) 0.15 (0.23); (0.64) -0.54 (0.59) 0.19 (0.85) 7.888 (1.000) 5,062.50 (0.000) Table Interpretation of Diagnostic Tests for Arrelano and Bover System GMM Models Arellano-Bover dynamic panel-data estimation Arellano-Bover dynamic panel-data estimation Diagnostic tests 1st lag in levels 2nd lag in level and 1st lag difference Number of instruments m1 Pr > z m2 Pr > z Sargan/Hansen test Prob > chi2 183(Max) 2.3167 (0.0205) 0.1061 (0.2687) 181.61 (0.1928) 86(Min) -1.9278 (0.0539) 1.6056 (0.1084) 7.88 (1.000) Arellano-Bover dynamic panel-data estimation 3th lag in level and 2nd lag differences (1 more instrument) 111 2.271 (1.1937) 0.2326 (0.0231) 116.7 (0.0563) Arellano-Bover dynamic panel-Data estimation 4th lag in level and 3th lag differences (2 more instruments) 135 -2.2837 (0.0224) 1.1487 (0.2507) 130.41 (0.2050) 1086 DETERMINANTS OF FEMALE EMPLOYMENT RATE IN THE EUROPEAN UNION Concerning the Sargan/Hansen test, as second test for instrument validity, too low and two high p-values can be indicative of weak instruments (Roodman, 2007) Also there is the problem of “to many” instruments where the Sargan test grows weaker the more instruments were tested and we unable to reject the null of instrument validity In this case, p values obtained in most of the cases were above the apparently very high rule of thumb a threshold of p = 0.25 According to the statistics presented in the table only the forth case of estimation provided p value near to the rule of thumb suggested by Rodman, p = 0.2050 Even we accept the null of valid instruments we still had to deal with the problem of two many instruments given that there are only 23 groups in our data set As discussed too many instruments can overfit endogenous variables and fail to expunge their endogenous components In our case the number of instruments is massively over the number of cross-section There are two options in trying to deal with this problem: limiting the lags used in the GMM-style instruments or using command for collapsing instruments available in xtabond2 In this analysis the second approach has been conducted The number of instruments are reported in Table 4, first row There is no clear guidance from the literature on how many instruments are “too many” (Roodman, 2009), although > xtabond2 < does give a warning when the number of instruments is larger than the number of cross-sectional units One of the ways to limit the instrument count is by collapsing them, i.e., creating instruments for each variable only Namely, when we “collapse” an instrument set, we create not a whole matrix of instruments but a single column vector of instruments, which means that there is only one instrument for all time periods (Pugh, 2004) At the same time there has been a growing evidence that that panel data is likely to exhibit cross-sectional dependence which may arise due to spatial dependencies, economic distances, common shocks thereby causing errors to be “correlated across the entire cross section” (Sarfidis et al., 2006) The evidence of 2nd—no order serial correlation might imply possibility of no heterogeneous error cross sectional dependence Table The Difference in Hansen Test ( C-statistics) C-statistics Chi2 P-values Ho: instrument validity 2 Hansen test of over-identified restriction Chi = 4.52 Prob > chi = 1.000 Not rejected Difference in Hansen test of exogeneity of instruments Chi2 = 5.17 Prob > chi2 = 0.819 Not rejected Comparison of both tests Chi2 = 0.65 Prob > chi2 = 1.000 Not rejected Note The p-values given above were compared to the conservative threshold suggested by Roodman (2007 and 2009) which is p = 0.25 According to the statistics presented in Table 5, in our preferred model the system GMM instruments for levels are valid, in which case we can accept the “steady-state” assumption required for system estimation and there is no undue problem with cross-sectional dependence However, the number of instruments still remained high and above the number of cross-sectional groups, i.e., 48 Despite the fact that we did not have 1st order correlation and have 2nd order correlation, at the same time valid instruments (Sargan/Hansen) we cannot say that this is a sensible model The number of instruments remained high which overfit the number of instruments and bias the results At the conventional 5% critical value, almost all of the coefficients were statistically individually insignificant Apart from the coefficient of the lagged dependent variable and two other variables, i.e., tertiary education and GDP growth are significant at 10% level DETERMINANTS OF FEMALE EMPLOYMENT RATE IN THE EUROPEAN UNION 1087 All of the signs in front of the coefficients are as expected, i.e., there is high level of persistency between lagged dependent variable and dependent variable in level as well as a negative relationship between female employment rate and increase in female unemployment, long maternity leave and high fertility rate Table Interpretation of the System Model Regressors Coefficient with robust SE Economic interpretation Employment rate (lagged) 0.96* Child care facilities 0.07 Fertility rate -7.55 Maternity leave -0.0013 Education 0.11** Female unemployment rate -0.04 Part-time employment 0.05 GDP 0.23* Constant term 3.71 On average, the female employment rate in current period is estimated to be 0.96% of the employment rate in the last period (t-1), ceteris paribus (high level of persistency) On average, 1% point increase of years old in pre-school education in current period, gives 0.07% point increase in the percentage of female employment ceteris paribus On average, increase average of birth rates in current period, will decrease the employment rate by 7.55%, ceteris paribus On average, week increase in maternity leave in the current period will give 0.0013% point decrease in the percentage of employment rate, ceteris paribus On average, 1% point increase in the female tertiary graduates will give 0.11% point increase in the female employment rate, ceteris paribus On average, 1% point increase in the female unemployment rate will give 0.04% point decrease in the female employment rate, ceteris paribus On average, 1% point increase in the female part-time employment will give 0.05% point increase in the female employment rate, ceteris paribus On average, 1% point increase in the GDP growth will give 0.23% point increase in the female employment rate, ceteris paribus The constant term has no theoretical meaning Notes * Significant at 10% level; ** Significant at 5% level; Coefficients without asterisk are statistically insignificant Conclusion In this paper, we have investigated the question what determines female employment rates in the European Union where a sample of 27 countries has been analyzed over a time period of 14 years from 1995 Because of the costly sluggish adjustments of employment we have specified a dynamic model where system GMM model appeared as more suitable compared to the difference model At the same, we have to be aware that our data set has a considerable lack of cross-sections Moreover, the number of instruments were increasing enormously over the number of observation even though we tried to reduce them Nevertheless, results from our study have been supported by vast number of empirical research in the literature At the same time, the choice of variables incorporated in the model was based on theory and empirical investigations where the chosen variables were suggested as possible determinants of female employment rate The main findings can be summarized as follows: The coefficient estimators for tertiary education and GDP growth per capita were found to be significant at 10 percent level while lagged female employment rate had coefficient estimator significant at the percent level The significance as well as the positive sign in front of these coefficients are as expected According to our results, the current employment rate is highly influenced by the previous year employment rate Past research on female employment has repeatedly shown that persistence is an important aspect of the female employment, 1088 DETERMINANTS OF FEMALE EMPLOYMENT RATE IN THE EUROPEAN UNION while first-order dependence the most important factor in explaining persistence in female labor supply behavior (Heckman & Willis, 1977; Nakamura & Nakamura, 1985; Eckstein & Wolpin, 1989) Persistence in employment may be due to state dependence which arises from human capital accumulation or the costs of searching for job which in turn may be influenced by social policies such as employment regulations and the availability of child care Tertiary education raises the utility of working full-time and lowers the utilities associate with part-time work and no work Our results have indicated that 1% point increase in the female tertiary graduates will give 0.11% point increase in the female employment rate, other things being constant, while a 1% point increase in the GDP growth will give 0.23% point increase in the female employment rate, ceteris paribus Contrary to our expectations, the rest of coefficients are statistically individually insignificant at both 5% and 10% level in our estimates Maternity leave that guarantees a (post-leave) right to return to work is an important component of family policies However, even though many analyses have revealed that job-protected leave can increase the time mothers spend at home with their infants and also the likelihood they return to their pre birth employer, women’s employment opportunities decline with the time away from work Our analysis results suggest that the length of maternity leave has negative impact on female employment where for every one week increase in the length on maternity leave, female employment rate is expect to decrease by 0.013% On the other hand, high part-time employment rate has a positive impact on employment For every percentage point increase in the percent of working women who work part-time, female employment rate is expected to increase by 0.05 High level of education, sociological and cultural reasons as well as increase in the fertility rate and number of children per family, have increased the need for working part-time instead of working full-time so as to reconcile professional and family life Limited availability of part-time employment and the limited availability of affordable child care services increase the costs of working for mothers, making it difficult to participate in the labour market Other social factors seem to negatively contribute to countries’ employment rates The influence of current labour market conditions and the responsiveness of women to those conditions are relatively strong Women’s responses to labour market conditions are normally found to be sensitive to changes in the unemployment rate Namely, high unemployment within a country has a negative effect upon female employment rates Our results suggest that a percentage point increase in a country’s unemployment rate will decrease its employment rate by 0.04 Moreover, women in their mid-20s facing a tight labor market and worsening economic conditions (i.e., high unemployment) tend to restrict their fertility below their ideal level, even though this phenomenon seems to be much weaker if they were employed in a stable public sector job, (Adsera, 2006) Concerning fertility rate effect on female employment our analysis has revealed a negative relationship, as expected Namely, a percentage increase in the average births, will decrease the employment rate by 7.55%, other things being equal However, many researchers have indicated that the female employment rate and fertility decisions are both affected by similar forces The decisions to work and have a child are positively influenced by the available supply of public child care as well as the availability of part time jobs Many researchers have indicated that by increasing the flexibility of employment relationships, more women would find it attractive to enter into the market As suggested by feminist observers (e.g., Folbre, 1997, 2001), countries that facilitated combining the worker and mother roles have higher fertility and higher female employment rates DETERMINANTS OF FEMALE EMPLOYMENT RATE IN THE EUROPEAN UNION 1089 where a major institutional influence is the availability and acceptability of child care centers According to our research, on average, 1% point increase of years old in pre-school education in current period, gives 0.07% point increase in the percentage of female employment References Adnett, N (1996) European labour markets—Analysis and policies Addison Wesley Longman Lim Andersson, G., Duvander, A Z., & Hank, K (2004) Do child-care characteristics influence continued child bearing in Sweden? An investigation of the quantity, quality and price dimension Journal of European Social Policy, 14, 407-418 Arellano, M., & Bover, O (1991) Another look at the instrumental variable estimation of error-components models Journal of Econometrics, 68, 29-51 Bond, S (2002a) Dynamic panel models: A guide to micro data methods and practices Institute for Fiscal Studies/Department of Economics, UCL, CEMMAP (Centre for Microdata Methods and practices) Working Paper CWPO9/02 Borjas, J G (2005) Labour economics (3rd ed.) McGraw-Hill/Irwin Bosch, G., Rubery, J., & Lehndorff, S (2007) European employment models under pressure to change International Labour Review, 146(3-4) Buddelmeyer, H., Mourre, G., & Ward, M (2008) Why Europeans work part-time? A cross-country panel analysis, European Central Bank-Eurosystem, Working paper series No 872/February 2008 Retrieved from http://ssrn.com/abstract_id=1091120 Carone, G., Denis, C., Morrow, M K., Mourre, G., & Roger, W (2006) Long-term labour productivity and GDP projections for the EU25 member states: A production function framework European economy European Commission directorate—General for economics and financial affairs, Economic papers Carrasco, R (2001) Binary choice with binary endogenous regressors in panel data: Estimating the effect of fertility in female labour participation Journal of Business Economic Statistics, 19, 385-394 Cebula, J R., & Coombs, K C (2007) Recent evidence of factors influencing the female labour force participation rate Journal of Labor Research, 29, 272 -284 Conesus, K., Goeggel, K., & Muehler, G (2009) Maternal employment and child care decision (Oxford economic papers April, 61, S1, pi172-i188, 17p, charts, graphs) De Vilota, Paloma, & Ignacio, F ( 2001) The impact of tax/benefit system on women’s work Madrid: European Community (Employment, Social Afares and Equal Opportunities), Retrieved from http://www.europa.eu.int/comm/employment social/equ opp/women work.pdf Del Boca, D., Locatelli, M., & Vuri, D (2004, January) Child care choices by Italian households (IZA Discussion Paper No 983) Del Boca, D., & Robert, M S (2007) Life cycle employment and fertility across institutional environments (IZA Discussion Paper) Dingeldey, I (2000) International comparison of tax systems and their impact on the work-family balancing North RhineWestphalia, Germany: Institute for Work and Technology Retrieved from http://www iatge.de/aktuell/veroeff/am/dingel100.pdf Edin, P A., & Gustavsson, M ( 2001) Time out of work and skill depreciation Mimeo Uppsala University Ehrenberg, G R., & Smith, S R ( 2003) Modern labour economics—Theory and public policy (8th ed.) Prentice Hall Elliot, F R ( 1991) Labour economics—A comparative text McGraw-Hill Book Company Espin-Andersen, C (1990) The three worlds of welfare capitalism Cambridge: Polity Press EUROSTAT (n.d.) European Commistion Statistics Retrieved from http://www.epp.eurostat.ec.europa.eu European Council (2000) Lisbon European Council 23-24 March 2000 Presidency conclution European parlament 2000 Retrived from http://www.europa.eu/legislation_summaries/ Falzone, J S (2000, November) Labour market decision of married women: With emphasis on “part-time employment” International advances in economic Research, 6(4), 662-671 Green, W ( 2002) Limdep Version 8.0: Econometrics modeling guide, Plainviwe NY: Econometric Software, Inc Heckman, J J., & Macurdy, E T (1980) A life cycle model of female labour supply Review of Economic Studies, XLVII, 47-74 Hotchkiss, J L (2006) Changes in behavioral and characteristics determination of female labor force participation, 1975-2005 Federal Reserve Bank of Atlanta Economic Review, 91(1) Lachenmaier, S., & Rottmann, H (2007) Effects of innovation on employment: A dynamic panel analysis CESifo Working Paper No 2015 Category 9, Industrial Organization 1090 DETERMINANTS OF FEMALE EMPLOYMENT RATE IN THE EUROPEAN UNION Lalive R., Schlosser, A., & Zweimuller, J (2009) How employment protection and parental leave benefits affect mothers’ post-birth careers? Retrieved from http://www.iza.org/en/papers/253_24112009.pd Mallier, T A., & Rosser, J M (1987) Women and the economy—Comparative study of Britain and USA Macmillian Press Ltd Martin, J., & Roberts, C (1980) Women and employment a lifetime perspective Department of Employment Office of Population Census and Surveys Michaud, P., & Tatsiramos, K (2008) Fertility and female employment dynamics in Europe: The effect of using alternative econometric modeling assumptions Paper funded by the NIA RAND Center for the study of aging (P30AG012815) and the NCHD funded RAND Population Center (R24HD050906) OECD (1990) Employment outlook Paris Geoff, P ( 2004) The GMM estimation of dynamic panel models: An intuitive explanation of the principles Staffordshire University Business School Peseran, M., & Smith, R (1995) Estimating long-run relationship from dynamic heterogeneous panels Journal of Econometrics, 68, 79-113 Peseran, M., Smith, R., & Kyung, S (1996) Dynamic linear model for heterogeneous panel (Chap 8) In L Matyas, & P Svestre (Eds.), The econometrics of panel data: Hand book of theory and applications (2nd ed.) Dordrecht: Kluwer Academic Publishers Pugh G (2009) The GMM estimation of dynamic panel models: An intuitive explanation of the principles Staffordshire University Business School, September 2004, updated March 2006, March 2008, September 2008 and March 2009 Rachel, A G., Lingsay, P., & Chase-Lamsdale (2001, May) Availability of child care in United States: A description and analysis of data sources Demography, 38(2) Roodman, D (2009) How to xtabond2: An introduction to difference and system GMM in Stata The Stata Journal, 9(1) Ruhm, C J (1998) The economic consequences of parental leave mandates: Lessons from Europe The Quarterly Journal of Economics, 113(1), 285-317 Setterfield, M (2009) Neoclassical growth theory and heterodox Groeth theory: Opportunities for and obstacles to greater engagement Department of Economics Trinity College Hartford, CT 06106 USA Srafidis, V., Yamagata, T., & Robertson, R (2006) A test of cross-section dependence for linear dynamic panel model with regressors Faculty of Economics, Univesrsity of Cambridge Troske, R K., & Alexandru, V (2009) The effect of children on the level of labour—Market involvement of married women: What is the role of education? (IZA-Institute for the study of labour, Discussion Paper No 4070, March 2009) Verbeek, M (2000) A Guide to modern econometrics Chisester: John Wiley Voicu, A., & Hielke, B (2003) Children and women’s participation dynamics: Transitory and long-term effects (IZA Discussion paper) Retrieved from http://www.iza.org/en/webcontent/events/transatlantic/papers_2003/voicu_buddelmeyer.pdf Warnecke, L T ( 2008) Women as wives, mothers and workers: How welfare eligibility requirements influence women’s labor force participation—A case study of Spain Journal of economic issues, XLII(4)