Elana Costa E Slva, Aldna Correa, Alexandra Braga, Vtor Braga A Multnomal Logstcs Model for Perceptons on Entrepreneurshp ELIANA COSTA E SILVA, ALDINA CORREIA, ALEXANDRA BRAGA AND VITOR BRAGA School of Technology and Management - ESTG Polytechnc of Porto Rua do Curral - Margarde, 4610-156 Felgueras PORTUGAL eos@estg.pp.pt, ac@estg.pp.pt, abraga@estg.pp.pt, vbraga@estg.pp.pt Abstract: - In ths paper we present a multnomal logstcs regresson to model the experts perceptons about the Entrepreneurshp Framework Condtons, usng the most recently avalable data from GEM,.e., NES 2013. The expert s type s descrbed by a nomnal varable wth fve categores,.e.: entrepreneur ; nvestor, fnancer, banker ; polcy maker ; busness and support servces provder ; and educator, teacher, entrepreneurshp researcher. The multnomal logstc regresson model presented an overall percentage correctness of 54.1%. The results show that the odds of an experts beng an entrepreneur over beng an educator, teacher, entrepreneurshp researcher ncreases wth the ncrease n the percepton of the people workng for government agences are competent and effectve n supportng new and growng frms. The same results were found for the odds of beng nvestor, fnancer, and banker and polcy maker. Furthermore, the odds of beng a polcy maker over beng an educator, teacher, entrepreneurshp researcher ncreases wth the ncrease of the percepton of the markets for busness-to-busness goods and servces change dramatcally from year to year. The same effect s observed for busness and support servces provder. Addtonally, the odds of beng a busness and support servces provder also ncreases wth the ncrease of the ant-trust legslaton s effectve and well enforced. Key-Words: - Entrepreneural Framework Condtons, Global Entrepreneurshp Montor, Multnomal Logstcs Regresson 1 Introducton One of the major drvers of economc growth s entrepreneurshp [1]. There has been a growng nterest n promotng entrepreneural actvty both on publc and prvate ntatves [2], [3]. In an effort to fll the lack of comparable nternatonal data on entrepreneurshp, the Global Entrepreneurshp Montor (GEM) began n 1999, as a jont project between Babson College (USA) and London Busness School (UK). Snce then GEM has been the source of comparable data across a large varety of countres on atttudes toward entrepreneurshp, startup and establshed busness actvtes, and aspratons of entrepreneurs for ther busnesses [3]. The data collected by GEM has presented a fast growth. The number of countres ncluded n the GEM data collecton ncrease from 10 countres n 1999 to more than 100 countres n 2016. In fact, accordng to the 2016 GEM Global Report, GEM covers 69.2% of the world s populaton and 84.9% of the world s Gross Domestc Product (GDP). The two man sources of prmary data of the GEM project are: () the Adult Populaton Survey (APS), whch provdes standardzed data on entrepreneural actvtes and atttudes wthn each country; and () the Natonal Expert Survey (NES), whch nvestgates the natonal framework condtons for entrepreneurshp by means of standardzed questonnares. Furthermore, GEM-based scentfc publcatons have presented a large ncrease n a wder range of academc journals [3], n partcular usng APS. By usng emprcal data, the GEM project ams for: (a) assessng the level of entrepreneural actvty across countres, (b) understandng how entrepreneural actvty vares over tme, and (c) understandng why some countres are more entrepreneural than others. Explorng the relatonshp between entrepreneural actvty and economc growth, and dentfy whch publc polces boost entrepreneurshp s also one of the ams of GEM. Accordng to the GEM methodology the Entrepreneural Framework Condtons (EFCs) are condtons that enhance (or hnde) new busness creaton. Therefore, E-ISSN: 2224-2899 1 Volume 15, 2018
Elana Costa E Slva, Aldna Correa, Alexandra Braga, Vtor Braga entrepreneurshp dynamcs can be lnked to condtons these condtons [6]. Accordng to Bergmann, Mueller, & Schrettle [5], the data provded by GEM offers the possblty to pose research questons that could not have been addressed before. The data s avalable at the webpage of the GEM project. For Álvarez, Urbano, & Amorós, 87% of the artcles use APS data, whle only 3% use the NES nformaton, and 10% use both nformaton sources[2]. It s clear that the nformaton provded by NES s an untapped resource for future publcatons. Also, snce NES ndvdual level fles are not commonly used for wrtng GEM reports, they are best ndcated to carry out academc research. Addtonally, 42% of the emprcal studes use logt, probt and tobt models, 29% use multple lnear regresson analyss assocated wth the macro level, 13% use panel data, and 16% use other technques. Addtonally, the dependent varables are mostly related to entrepreneural actvtes n general (59%), ndcators of entrepreneural aspratons (14%), female entrepreneurshp (10%), by studes that use dependent varables related to economc ssues, especally growth and economc development (5%). Furthermore 5% of the research artcles attempt to explan perceptons of opportuntes and motvatons to become an entrepreneur, whle 7% use fnancal aspects as the dependent varable [2]. For further nformaton on the statstcal technques used see [5]. Sánchez-Escobedo et al. analyzed the stuaton and development of research, usng data from the GEM n the perod from 1999 to 2015, n entrepreneurshp from a gender perspectve [8]. Global APS s used n 32.5% of the research papers publsh n journals of entrepreneurshp ncluded n the Web of Scence (WOS), whle 37.5% use Natonal APS, 20% use APS and NES and 10% use APS and secondary sources, such as, OECD, World Bank, US Census, Hertage Foundaton, Encyclopeda Brtannca, BBVA Foundaton. In terms of the Statstcal technque of analyss, 20% of the works use descrptve analyss, 52.50% use descrptve analyses and logstc regresson, whle the remanng use other technques such as Panel data, ANOVA, MANOVA, Bootstrappng, Structural equatons. In ths paper the 2013 NES data provded by GEM s analyzed and a multnomal logstcs regresson approach s followed wth the objectve of analyzng the effects of dfferent type of entrepreneurshp expert specalzaton on the perceptons about the Entrepreneural Framework Condtons (EFCs). 2 Dataset and Model Estmaton The data used n the present work was collected from the GEM webste and concerns the ndvdual NES 2013 dataset. The NES survey collects data concernng the perceptons on entrepreneurshp of experts n dfferent countres. From survey to survey there are dfferences n terms of the varables consdered. However, the country s Entrepreneural Framework Condtons (EFCs), namely: A - Fnancng for Entrepreneurs; B - Governmental Polces; C - Governmental Programs; D - Entrepreneural Educaton and Tranng; E - R&D Transfer; F - Commercal and Professonal Infrastructure; G - Internal Market Openness; H - Physcal and Servces Infrastructure; I - Socal And Cultural Norms. are the same from survey to survey. In the present work, as response varable, the categorcal varable SPE s used. Ths varable characterzes the expert type and was defne usng the dummy varables SPE1, SPE2, SPE3, SPE4 and SPE5 (for detals see [6], [7]). The SPE varable has fve categores: 1=entrepreneur; 2=nvestor, fnancer, banker; 3=polcy maker; 4=busness and support servces provder; and 5=educator, teacher, entrepreneurshp researcher. As ndependent varables all NES varables ncluded n the EFCs, from A1 to I05 are used, n a total of 52 varables. Followng the common procedure adopted when workng wth real-world data, such as GEM data, the ndependent ordnal (lkert lke scale) NES varables are consder as contnuous. See Table 7 for the varable descrpton. Snce the observatons are ndependent, the dependent varable SPE s measured at the nomnal level wth fve mutually exclusve and exhaustve categores and snce there s no multcollnearty (the Varance Inflaton Factor s <3), a multnomal logstcs regresson approach s adequate for ths analyss. From a total of 2636 observatons, 1227 are vald. The mssng values correspond to the number of observatons n the dataset where data are mssng from the response varable or from any of the predctor varables. For the majorty of these observatons the experts are Entrepreneur ; 14.3% are Polcy makers ; 12.7% are Busness and support servces provder ; 10.2% are Investor, E-ISSN: 2224-2899 2 Volume 15, 2018
Elana Costa E Slva, Aldna Correa, Alexandra Braga, Vtor Braga fnancer, banker ; and 9.1% are Educator, teacher, entrepreneurshp researcher (see Table 1). Table 1: Case Processng Summary N Margnal Percentage 1= Entrepreneur 659 53.7% 2= Investor, fnancer, banker 125 10.2% 3= Polcy maker 175 14.3% SPE 4= Busness and support servces provder 156 12.7% 5= Educator, teacher, entrepreneurshp researcher 112 9.1% Vald 1227 100.0% Mssng 1409 Total 2636 Subpopulaton 1225 a a. The dependent varable has only one value observed n 1223 (99.8%) subpopulatons. A backward stepwse method, mplemented at IBM SPSS verson 23, was used to estmate the multnomal logstc regresson model, and chsquare for varables entry and removal, based on the lkelhood rato test, were consdered. Table 2 presents model fttng nformaton. By ncludng the predctor varables, the Fnal model mproves when compared to the Intercept Only model as can be seen by the Akake Informaton Crteron (AIC), Bayesan Informaton Crteron (BIC) and log lkelhood. The lkelhood of the model s used to test of whether all predctors regresson coeffcents n the model are smultaneously zero. From the Lkelhood rato Ch-Square statstc a p- value <0.0001 was obtan, therefore, at least one of the regresson coeffcents n the model s not equal to zero,.e. the model s sgnfcant. Note that the degrees of freedom of the ch-square dstrbuton s the number of predctors n the model. Table 2: Model Fttng Informaton Lkelhood Rato Model Fttng Crtera Tests Model AIC BIC -2 Log Lkelhood Ch- Square df p-value Intercept 3256.85 3277.30 Only 3248.85 Fnal 3207.93 3473.77 3103.93 144.92 48 0.000 Table 3 suggests that the model presented a good ft to the data snce Pearson ch-square p-value s 0.374>0.05 and the Devance ch-square statstc presents a p-value>0.05. Table 3: Goodness-of-Ft and Pseudo R-Square Ch- Square df p-value Cox Snell and 0.111 Pearson 4879.03 4848 0.374 Nagelkerke 0.120 Devance 3101.16 4848 1.000 McFadden 0.045 The pseudo R-Square s presented n Table 3. Table 4 presents the Lkelhood rato tests and shows that from all the ndependent varables consdered only the followng are statstcally sgnfcant, for a sgnfcant level of 10%: A03 In my country, there are suffcent government subsdes avalable for new and growng frms ; A04, there s suffcent fundng avalable from prvate ndvduals (other than founders) for new and growng frms ; B02, the support for new and growng frms s a hgh prorty for polcy at the natonal government level ; B05, the amount of taxes s NOT a burden for new and growng frms ; B06, taxes and other government regulatons are appled to new and growng frms n a predctable and consstent way ; C03, there are an adequate number of government programs for new and growng busnesses ; C04, the people workng for government agences are competent and effectve n supportng new and growng frms ; E02, new and growng frms have just as much access to new research and technology as large, establshed frms ; E03, new and growng frms can afford the latest technology ; G02, the markets for busness-to-busness goods and servces change dramatcally from year to year ; G04, the new and growng frms can afford the cost of market entry ; G06, the ant-trust legslaton s effectve and well enforced. Recall that the underlyng null hypothess s that the parameters of that varable are zero. Table 4 also presents the overall effect of each varable. The parameter estmates for the multnomal logstc regresson model are depcted n Table 5. The E-ISSN: 2224-2899 3 Volume 15, 2018
Elana Costa E Slva, Aldna Correa, Alexandra Braga, Vtor Braga parameter estmates are relatve to the reference category Educator, teacher, entrepreneurshp researcher. The sgnfcant coeffcents to a 10% sgnfcance level are n bold,.e., those for whch the p-value of Wald test s <0.10. Table 4: Lkelhood Rato Tests Lkelhood Rato Model Fttng Crtera Tests AIC of Reduced Model BIC of Reduced Model -2 Log Lkelhoo d of Reduced Model Ch- p- Effect Square df value Intercept 3291.06 3536.46 3195.06 91.13 4 0.000 A03 3208.92 3454.32 3112.92 8.99 4 0.061 A04 3212.32 3457.71 3116.32 12.39 4 0.015 B02 3215.48 3460.87 3119.48 15.55 4 0.004 B05 3217.58 3462.97 3121.58 17.65 4 0.001 B06 3210.52 3455.91 3114.52 10.59 4 0.032 C03 3210.20 3455.59 3114.20 10.27 4 0.036 C04 3218.70 3464.09 3122.70 18.77 4 0.001 E02 3208.30 3453.69 3112.30 8.37 4 0.079 E03 3214.17 3459.56 3118.17 14.24 4 0.007 G02 3211.38 3456.77 3115.38 11.45 4 0.022 G04 3210.67 3456.06 3114.67 10.74 4 0.030 G06 3212.83 3458.22 3116.83 12.90 4 0.012 The estmaton of a multnomal logstc model requres the computatonal of M-1 equatons where M s the number of categores of the response varable. In ths case, snce the SPE has fve categores, a total of four models are computed. Snce the last category ( Educator, teacher, entrepreneurshp researcher ) was used as reference, the models are: P( SPE m) log m m1a03 m2a04 m3b02 P( SPE 5) B05 B06 C03 m4 C04 E02 E03 m7 G02 m10 m5 m8 G04 m11 m6 m9 G06, m12 (1) where m=1,2,3,4. A total of four log odds were predcted relatvely to the reference category. For 1= Entrepreneur NES Experts the followng nterpretatons can be done from Table 5: an ncrease of one unt of B05 s assocated to a decrease of 0.174 n the relatve log odds of beng Entrepreneur versus Educator, teacher, entrepreneurshp researcher, whle keepng the other varables constants. ths nverse relaton s also present n C03 and E03 wth 0.219 and 0.388, respectvely, n the relatve log odds of beng Entrepreneur versus Educator, teacher, entrepreneurshp researcher ; on the other hand, for an ncrease of one unt of C04 there s an ncrease of 0.296 n the relatve log odds of beng Entrepreneur versus Educator, teacher, entrepreneurshp researcher ; the odds of beng a Entrepreneur over beng a Educator, teacher, entrepreneurshp researcher decreases n 16% wth an ncrease of one unt n B05, 20% wth an ncrease of one unt n C03, and 32% wth an ncrease of one unt n E03; the odds of beng a Entrepreneur over beng a Educator, teacher, entrepreneurshp researcher ncreases n 35% wth an ncrease of one unt n C04. For 2= Investor, fnancer, banker NES Experts, from Table 5: an ncrease of one unt of C03 s assocated to a decrease of 0.257 n the relatve log odds of beng Investor, fnancer, banker versus Educator, teacher, entrepreneurshp researcher, whle keepng the other varables constants. whle, for an ncrease of one unt of C04 there s an ncrease of 0.348 n the relatve log odds of beng Investor, fnancer, banker versus Educator, teacher, entrepreneurshp researcher. For 3= Polcy maker NEs Experts, from Table 5: an ncrease of one unt of C03 (and E05) s assocated to a decrease of 0.384 (and 0.501) n the relatve log odds of beng Polcy maker maker versus Educator, teacher, entrepreneurshp researcher, whle keepng the other varables constants. whle, for an ncrease of one unt of C04 (and G02) there s an ncrease of 0.488 (and 0.214) n the relatve log odds of beng Polcy maker versus Educator, teacher, entrepreneurshp researcher. Fnally for 4= Busness and support servces provder NEs Experts, from Table 5: an ncrease of one unt of E03 (and G04) s assocated to a decrease of 0.285 (and 0.400) n the relatve log odds of beng Busness and support servces provder versus Educator, teacher, entrepreneurshp researcher, whle keepng the other varables constant. whle, for an ncrease of one unt of G04 (and G06) there s an ncrease of 0.277 (and 0.309) n the relatve log odds of beng Busness and support servces provder versus Educator, teacher, entrepreneurshp researcher. E-ISSN: 2224-2899 4 Volume 15, 2018
Elana Costa E Slva, Aldna Correa, Alexandra Braga, Vtor Braga Table 5: Parameter Estmates 95% Confdence Interval for Exp(B) SPE B Std. Error Wald df p-value Exp(B) Lower Bound Upper Bound 1= Intercept 2.719 0.500 29.599 1 0.000 Entrepreneur A03-0.083 0.111 0.550 1 0.458 0.921 0.740 1.145 A04 0.114 0.101 1.270 1 0.260 1.121 0.919 1.367 B02-0.121 0.102 1.427 1 0.232 0.886 0.726 1.081 B05-0.174 0.095 3.373 1 0.066 0.840 0.698 1.012 B06 0.133 0.102 1.716 1 0.190 1.143 0.936 1.395 C03-0.219 0.117 3.477 1 0.062 0.804 0.638 1.011 C04 0.296 0.115 6.648 1 0.010 1.345 1.074 1.685 E02 0.161 0.118 1.863 1 0.172 1.175 0.932 1.481 E03-0.388 0.119 10.607 1 0.001 0.679 0.537 0.857 G02 0.050 0.097 0.268 1 0.605 1.051 0.869 1.272 G04-0.201 0.122 2.703 1 0.100 0.818 0.643 1.039 2= Investor, fnancer, banker 3= Polcy maker 4= Busness and support servces provder G06 0.039 0.104 0.140 1 0.708 1.040 0.848 1.276 Intercept 0.315 0.629 0.251 1 0.616 A03 0.083 0.142 0.342 1 0.559 1.086 0.823 1.433 A04-0.020 0.129 0.024 1 0.877 0.980 0.761 1.263 B02-0.109 0.129 0.719 1 0.396 0.897 0.697 1.154 B05 0.194 0.119 2.647 1 0.104 1.214 0.961 1.532 B06-0.170 0.132 1.666 1 0.197 0.844 0.651 1.092 C03-0.257 0.149 2.974 1 0.085 0.774 0.578 1.036 C04 0.348 0.144 5.830 1 0.016 1.416 1.068 1.878 E02-0.150 0.154 0.961 1 0.327 0.860 0.637 1.162 E03-0.250 0.154 2.645 1 0.104 0.779 0.577 1.053 G02 0.047 0.122 0.147 1 0.702 1.048 0.825 1.332 G04-0.069 0.155 0.197 1 0.657 0.933 0.689 1.265 G06 0.245 0.132 3.466 1 0.063 1.277 0.987 1.653 Intercept -0.165 0.603 0.075 1 0.784 A03 0.141 0.132 1.146 1 0.284 1.152 0.889 1.491 A04-0.186 0.123 2.298 1 0.130 0.830 0.653 1.056 B02 0.158 0.121 1.697 1 0.193 1.171 0.924 1.484 B05-0.077 0.113 0.470 1 0.493 0.926 0.742 1.154 B06 0.094 0.120 0.617 1 0.432 1.099 0.869 1.390 C03-0.384 0.141 7.438 1 0.006 0.681 0.517 0.898 C04 0.488 0.136 12.901 1 0.000 1.630 1.248 2.127 E02 0.080 0.140 0.325 1 0.569 1.083 0.823 1.425 E03-0.501 0.147 11.663 1 0.001 0.606 0.454 0.808 G02 0.214 0.115 3.469 1 0.063 1.239 0.989 1.552 G04-0.053 0.145 0.136 1 0.713 0.948 0.713 1.260 G06 0.130 0.123 1.113 1 0.291 1.139 0.894 1.450 Intercept 0.106 0.618 0.029 1 0.864 A03-0.146 0.135 1.175 1 0.278 0.864 0.663 1.125 A04 0.009 0.123 0.005 1 0.942 1.009 0.793 1.284 B02 0.107 0.122 0.766 1 0.381 1.113 0.876 1.414 B05-0.079 0.116 0.463 1 0.496 0.924 0.737 1.159 B06 0.020 0.123 0.026 1 0.871 1.020 0.801 1.299 C03-0.064 0.142 0.204 1 0.651 00.938 0.710 1.239 C04 0.084 0.139 0.367 1 0.544 1.088 0.829 1.427 E02 0.148 0.141 1.103 1 0.294 1.159 0.880 1.527 E03-0.285 0.145 3.863 1 0.049 0.752 0.565 0.999 G02 0.277 0.118 5.568 1 0.018 1.320 1.048 1.661 G04-0.400 0.150 7.084 1 0.008 0.670 0.500 0.900 G06 0.309 0.125 6.074 1 0.014 1.361 1.065 1.740 E-ISSN: 2224-2899 5 Volume 15, 2018
Elana Costa E Slva, Aldna Correa, Alexandra Braga, Vtor Braga 1= Entrepren eur Table 6: Classfcaton Predcted 2= Investor, fnancer, banker 3= Polcy maker 4= Busness and support servces provder 5= Educator, teacher, entrepreneurshp researcher Percent Correct Observed 1= Entrepreneur 650 1 3 2 3 98.6% 2= Investor, fnancer, banker 122 1 2 0 0 0.8% 3= Polcy maker 166 0 8 1 0 4.6% 4= Busness and support servces provder 150 0 0 4 2 2.6% 5= Educator, teacher, entrepreneurshp researcher 110 1 0 0 1 0.9% Overall Percentage 97.6% 0.2% 1.1% 0.6% 0.5% 54.1% In my country, Table 7: Varables descrpton A01 there s suffcent equty fundng avalable for new E02 and growng frms there s suffcent debt fundng avalable for new and A02 E03 growng frms there are suffcent government subsdes avalable A03 E04 for new and growng frms there s suffcent fundng avalable from prvate A04 ndvduals (other than founders) for new and E05 growng frms A05 there s suffcent venture captalst fundng E06 avalable for new and growng frms ) there s suffcent fundng avalable through ntal A06 F01 publc offerngs (IPOs) for new and growng frms B01 Government polces (e g, publc procurement) F02 consstently favor new frms B02 the support for new and growng frms s a hgh F03 prorty for polcy at the natonal government level B03 the support for new and growng frms s a hgh F04 prorty for polcy at the local government level B04 new frms can get most of the requred permts and F05 lcenses n about a week the amount of taxes s NOT a burden for new and B05 G01 growng frms taxes and other government regulatons are appled B06 to new and growng frms n a predctable and consstent way B07 C01 C02 C03 C04 copng wth government bureaucracy, regulatons, and lcensng requrements t s not unduly dffcult for new and growng frms a wde range of government assstance for new and growng frms can be obtaned through contact wth a sngle agency In my country, new and growng frms have just as much access to new research and technology as large, establshed frms new and growng frms can afford the latest technology there are adequate government subsdes for new and growng frms to acqure new technology the scence and technology base effcently supports the creaton of world-class new technology-based ventures n at least one area there s good support avalable for engneers and scentsts to have ther deas commercalzed through new and growng frms there are enough subcontractors, supplers, and consultants to support new and growng frms new and growng frms can afford the cost of usng subcontractors, supplers, and consultants t s easy for new and growng frms to get good subcontractors, supplers, and consultants t s easy for new and growng frms to get good, professonal legal and accountng servces t s easy for new and growng frms to get good bankng servces (checkng accounts, foregn exchange transactons, letters of credt, and the lke) the markets for consumer goods and servces change dramatcally from year to year G02 the markets for busness-to-busness goods and servces change dramatcally from year to year G03 new and growng frms can easly enter new markets G04 the new and growng frms can afford the cost of market entry scence parks and busness ncubators provde new and growng frms can enter markets wthout G05 effectve support for new and growng frms beng unfarly blocked by establshed frms there are an adequate number of government the ant-trust legslaton s effectve and well G06 programs for new and growng busnesses enforced the people workng for government agences are the physcal nfrastructure (roads, utltes, competent and effectve n supportng new and H01 communcatons, waste dsposal) provdes good growng frms support for new and growng frms E-ISSN: 2224-2899 6 Volume 15, 2018
Elana Costa E Slva, Aldna Correa, Alexandra Braga, Vtor Braga C05 almost anyone who needs help from a government program for a new or growng busness can fnd H02 what they need C06 Government programs amed at supportng new and H03 growng frms are effectve teachng n prmary and secondary educaton D01 encourages creatvty, self-suffcency, and personal ntatve D02 D03 D04 D05 D06 E01 teachng n prmary and secondary educaton provdes adequate nstructon n market economc prncples teachng n prmary and secondary educaton provdes adequate attenton to entrepreneurshp and new frm creaton Colleges and unverstes provde good and adequate I02 preparaton for startng up and growng new frms the level of busness and management educaton provde good and adequate preparaton for startng I03 up and growng new frms the vocatonal, professonal, and contnung educaton systems provde good and adequate preparaton for startng up and growng new frms t s not too expensve for a new or growng frm to get good access to communcatons (phone, Internet, etc ) a new or growng frm can get good access to communcatons (telephone, nternet, etc ) n about a week H04 new and growng frms can afford the cost of basc utltes (gas, water, electrcty, sewer) H05 new or growng frms can get good access to utltes (gas, water, electrcty, sewer) n about a month the natonal culture s hghly supportve of I01 ndvdual success acheved through own personal efforts I04 new technology, scence, and other knowledge are effcently transferred from unverstes and publc I05 research centers to new and growng frms the natonal culture emphaszes self-suffcency, autonomy, and personal ntatve the natonal culture encourages entrepreneural rsktakng the natonal culture encourages creatvty and nnovatveness the natonal culture emphaszes the responsblty that the ndvdual (rather than the collectve) has n managng hs or her own lfe Consderng Z m A03 m m1 m2 m4 A04 B02 C04 E02 E03 m7 G02 m10 m3 B05 B06 C03 m5 m8 G04 m11 m6 m9 G06, m12 (2) The probablty of an experts beng an m=1 "entrepreneur", 2 "nvestor, fnancer, banker", 3 "polcy maker", 4 "busness and support servces provder", P exp( Zm ) m) 1 exp( Z ( SPE 4 k 1 k, ) (3) and beng 5 educator, teacher, entrepreneurshp researcher P exp( Zm ) 5) 1 exp( Z ( SPE 4 k 1 k, ) (4) Table 6 shows the classfcaton results obtaned used the multnomal logt regresson model. The overall percentage correct s 54.1%. The model was able to predct correctly 650 (or 98.6%) of 1= Entrepreneur, eght (4.6%) 3= Polcy maker, four (2.6%) 4= Busness and support servces provder, and only one 2= Investor, fnancer, banker and 5= Educator, teacher, entrepreneurshp researcher, correspondng to 0.8% and 0.9%, respectvely. 3 Conclusons A multnomal logstcs regresson model for analyzng the dfferent type of entrepreneurshp expert specalzaton on the perceptons about the Entrepreneural Framework Condtons (EFCs) s made usng 2013 ndvdual NES data provded by GEM. The type of expert was descrbed by a nomnal varable wth fve categores: 1= entrepreneur ; 2= nvestor, fnancer, banker ; 3= polcy maker ; 4= busness and support servces provder ; and 5= educator, teacher, entrepreneurshp researcher. Usng the last category as the reference, the results show that: (a) an ncrease of one B05 the amount of taxes s NOT a burden for new and growng frms, C03 there are an adequate number of government programs for new and growng busnesses and E03 new and growng frms can afford the latest technology yelds a decrease n the relatve log odds of beng entrepreneur when compared to educator, teacher, entrepreneurshp researcher whle C04 the people workng for government agences are competent and effectve n supportng E-ISSN: 2224-2899 7 Volume 15, 2018
Elana Costa E Slva, Aldna Correa, Alexandra Braga, Vtor Braga new and growng frms yelds a sgnfcant ncrease; (b) an ncrease of one unt of C03 there are an adequate number of government programs for new and growng busnesses s assocated to a decrease n the relatve log odds of beng nvestor, fnancer, banker versus educator, teacher, entrepreneurshp researcher, whle, for an ncrease of one unt of C04 the people workng for government agences are competent and effectve n supportng new and growng frms there s an ncrease; (c) an ncrease of one unt of C03 there are an adequate number of government programs for new and growng busnesses and E05 the scence and technology base effcently supports the creaton of world-class new technology-based ventures n at least one area s assocated to a decrease n the relatve log odds of beng polcy maker versus educator, teacher, entrepreneurshp researcher, whle, for an ncrease of one unt of C04 the people workng for government agences are competent and effectve n supportng new and growng frms and G02 the markets for busness-to-busness goods and servces change dramatcally from year to year there s an ncrease; (d) an ncrease of one unt of E03 new and growng frms can afford the latest technology and G04 the new and growng frms can afford the cost of market entry s assocated to a decrease n the relatve log odds of beng busness and support servces provder versus educator, teacher, entrepreneurshp researcher, whle, for an ncrease of one unt of G04 the new and growng frms can afford the cost of market entry and G06 the anttrust legslaton s effectve and well enforced there s an ncrease. The odds of an experts beng an entrepreneur over beng an educator, teacher, entrepreneurshp researcher ncreases wth the ncrease n the percepton of the people workng for government agences are competent and effectve n supportng new and growng frms. The same occurs for the odds of beng nvestor, fnancer, banker and polcy maker. The odds of beng a polcy maker over beng an educator, teacher, entrepreneurshp researcher ncreases wth the ncrease of the percepton of the markets for busness-to-busness goods and servces change dramatcally from year to year. The same effect s observed for busness and support servces provder. The odds of beng a busness and support servces provder also ncreases wth the ncrease of the ant-trust legslaton s effectve and well enforced. In terms of the correctness of the classfcaton, the multnomal logt regresson model presented dsplays an overall percentage correctness of 54.1%. References: [1] Reynolds, P., Bosma, N., Auto, E., Hunt, S., De Bono, N., Servas, I., & Chn, N., Global entrepreneurshp montor: Data collecton desgn and mplementaton 1998 2003, Small Busness Economcs, Vol.24, No.3, 2005, pp. 205-231. [2] Álvarez, C., Urbano, D., & Amorós, J. E., GEM research: achevements and challenges, Small Busness Economcs, Vol.42, No.3, 2014, pp. 445-465. [3] Bosma, N. S., & Leve, J., Global Entrepreneurshp Montor 2009 Executve Report, 2010. [4] Bosma, N., The Global Entrepreneurshp Montor (GEM) and ts mpact on entrepreneurshp research, Foundatons and Trends n Entrepreneurshp, Vol.9, No.2, 2013, pp. 143-248. [5] Bergmann, H., Mueller, S., & Schrettle, T., The use of global entrepreneurshp montor data n academc research: A crtcal nventory and future potentals. Internatonal Journal of Entrepreneural Venturng, Vol.6, No.3, 2014, pp. 242-276. [6] Correa, A., Costa e Slva, E., Lopes, I. C., & Braga, A., MANOVA for dstngushng experts perceptons about entrepreneurshp usng NES data from GEM, In T. E. Smos, Z. Kalogratou, & T. Monovasls (Eds.), AIP Conference Proceedngs,. AIP Publshng Vol.1790, No.1, 2016, p. 140002. [7] Correa, A., Costa e Slva, E., Lopes, I. C., Braga, A. & Braga, V., Experts perceptons on the Entrepreneural Framework Condtons, In T. E. Smos, Z. Kalogratou, & T. Monovasls (Eds.), AIP Conference Proceedngs,. AIP Publshng Vol.1790, No.1, 2017, pp. 140002. [8] Sánchez-Escobedo, M. C., Fernández-Portllo, A., Díaz-Casero, J. C., & Hernández-Mogollón, R., Research n entrepreneurshp usng GEM data. Approach to the state of affars n gender studes, European Journal of Management and Busness Economcs, Vol.25, No.3, 2016, pp. 150-160. E-ISSN: 2224-2899 8 Volume 15, 2018