,

The area to the left of z value in normal distribution curve measures the probability that a variable distributed normally is located to the left to this point. It serves for positive outcome probability measurement. The area under the curve is calculated by the definite integral

,

Such cumulated distribution function (CDF) if known as probit function (Hill, Griffiths & Lim, 2011).

Binary outcome regressions estimate the probability of the positive outcome (affirmative answer to the question) as a function of an independent normal distributed variable.

,

Whenever is defined as follows



,

As denoted by Gujarati (2004), Z (latent variable used in probit regression) has a certain critical (threshold) value, say Z*; whenever exceeds the critical value, the outcome is positive. Contrary, in the opposite case, the outcome is negative (for example, a person is not involved in early-stage entrepreneurial activities).

Hence, the probability p that y takes the value 1 in the case of one regressor, denoting a positive answer, is expressed by a probit function so as

,

In this expression Ф(z) is the probit function (Hill, Griffiths & Lim, 2011).

To determine whether relationship exists between individual's desire to be involved in early-stage entrepreneurial activity of any kind (opportunity or necessity) and the fact of knowing a person who started his or her own business in the past two years as well to establish the essence of the relationship between respondent's household size and his or her career choice several probit models were built.

Stata software used in this research uses the maximum likelihood estimation. The maximum likelihood method is optimal both for linear and non-linear models. The method gives researchers an opportunity to calculate t-statistics and construct confidence intervals (Stock & Watson, 2011). An interesting characteristic of the probit regression is that its coefficients may not be directly interpreted and the effect of similar rise (by absolute value) of a certain variable may have different influence on the predicted variable. That may be a model's drawback, because the coefficients may not be easily interpreted unlike doing so in linear regression. The only possible interpretation of the obtained coefficients is that negative value says that the increase in predictor is followed by a decrease in the predicted variable.

However, for the purposes of measuring the scale of influence the concept of marginal effects was introduced. It is necessary to calculate them to interpret the significance of the independent variables used in the model (Hasebe, 2013). Non-linear nature of the probit function observed by the linear influence of an independent variable on a latent variable and simultaneous non-linear effect on the predicted probability (Baum, 2006). Marginal effects help examine a one-unit change in x (independent variable) on the probability of affirmative answer (y=1) by examining the derivative

,

Both t and are normal PDF assessed at (Hill, Griffiths & Lim, 2011).

It was observed above that the rate of changes in probability depends on the density function of normally distributed variable Z; Z, in turn, is the function of all regressors included in the model. Due to its complexity, all regressors affect the probability predicted as well as the value of .

All in all, nineteen regressions were constructed. Each model included motivation factors mentioned by academics in past research papers and several control variables, unchanged throughout the experiment. Such socio-demographic characteristics as gender, age, education attainment, an annual income of the entire household (in thirds), country of residence were introduced. All motivating factors are binary variables because they were coded as an agreement (y=1) or disagreement (y=0). Estimated probability is then compared to an optimal threshold point (i.e. 0.5) to decide which group an individual may be referred to (i.e. nascent entrepreneur or not) (Hill, Griffiths & Lim, 2011).

Following independent variables were tested for significance: priority of similar standard of living in the country of residence, consideration of business as a favourable career choice, high level and status attributed to successful entrepreneurs, frequent appearance of successful business stories in the public media, personal knowledge of people who started business in the past two years, good opportunity perception to set up a company in the area a person lives, required skills necessary to establish an enterprise perception, considering fear of failure a barrier to starting a company ("Entrepreneurial Attitudes, Perceptions, Intentions", 2019).

Several combinations of variables (motivating factors and control variables) groups were used to examine their effect on both genders as well as their role in making a decision to enter early-stage entrepreneurship, opportunity early-stage entrepreneurship, and necessity early-stage entrepreneurship. As noticed above, total early-stage entrepreneurial activities include nascent entrepreneurs and owner-managers of a new business (a company established less than 3.5 years ago).

Dataset was initially processed to get rid of missing values and undefined variables. To conduct the analysis Stata software was used. Stata is chosen for the work because it serves several opportunities at once.

To begin with, any research work demands thorough preparation of data themselves. It helps sort data and makes them suitable for model building, schemes and figures creation. Stata is one of the most efficient and effective tools for such operations. Surely, Stata contains a large number of tools for econometric analysis. Stata has options of using its menu, in-built commands, and ones for graph and figures creation. It also has a creeping line for do-files insertion (algorithms written in advance). What is more, the statistical programme is constantly updated, new features are added, users create their own packages for the programme and share with other users. Lastly, Stata is compatible with several operating systems: Linux, Unix, Windows, Mac OS (Baum, 2006). Comparing Stata with other different software packages, it was concluded to use this particular one because R with all its virtues such as flexibility, open code and a presence of powerful analytical tool does not seem appropriate. In most cases this package orients on special tasks related to the financial sector.

Prior to running probit regressions, a correlation matrix is drawn to show a correlation between every pair of independent variables. An additional pre-estimation step is the average difference, mean, and deviation calculations. Student's t-test compares two averages and identifies their significance. The t-test is used for estimating means difference of every independent variable. Mean and deviation are measures for the entire set of independent variables (both dummies and continuous ones).

As GEM pays particular attention to the external environment and social attitude towards business as a career choice, dummy variables for all noted countries were included to estimate if entrepreneurs' and employees' location matters in work-related decisions.

After running every model modification, ROC curves and scalar- and matrix-valued statistics are received with the help of estat command for all probit variations. ROC curve is a multifunctional instrument for test evaluation diagnostics. It also evaluates the specificity and sensitivity of the model. Last but not least, the ROC curve defines an optimal threshold (cut-off) point to balance the sensitivity and specificity of the regression. Cut-off determines the frequency of type I and type II errors. The curve demonstrates the dependence of true positive rate on the corresponded false positive rate. At the same time, sensitivity is a true positive classification rate - TPR); specificity (true negative classification rate - TNR). The area under the curve (AUC) estimates how well the model categorizes observations of two classes (i.e. individuals involved in TEA and others). The resulting area can take values from 0.5 to 1. The value closer to 1 proves a better model's prediction results. An area under the ROC curve higher than 0.8 says that the accuracy is good. Otherwise, it needs additional components (e.g. data visualization techniques). More detailed results are given in classification tables.

Hence, each model is run in order to estimate the type of effect (positive or negative) coupled with the marginal effect of every regressor included. Coefficients' significance and their marginal effects are confronted. Finally, with the help of ROC curve (including its AUC) and matrix received by estab command in Stata software probit binary choice regressions' sensitivity, specificity, and prediction accuracy are measured and compared. Marginal effects and coefficients' signs serve the ground for testing hypotheses and answering the research question. All used commands in Stata for data analyzing are presented in the do-file (Appendix 1).



4. Description of the results

In the following section the results of the research are presented. The first stage was to create the table with descriptive statistics. The table consists of independent variables, which were selected as the most reliable and appropriate for the research from literature review. They are the following: the age of the respondents in years; the age of the respondent after its transformation in logarithmic form; the gender of the respondents; the level of education; household income; the number of members, participating in the household of the respondent; the personal knowledge of early-stages entrepreneurs; good opportunities for starting a business where the entrepreneur lives; the fear of failure as a preventing factor to start the business; respondent's preferences for similar standards of living in the country, where they would like to operate; the consideration of people to start a business as a desirable career choice; the idea that people who are successful starting the business have a high level of status and respect; the number of stories in different public media about successful new business and the level of difficulty to start the business in the country, where they live.

These dependent variables were described and presented in several ways, such as mean values, standard deviation and the differences between variable values using Student's t-test and pr-test commands and the equality of proportions of the selected variable according to the group in which they are identified.

Student's t-test may be processed by Stata software in several cases which makes it usage wider: t-test could be used for one sample, two samples, and paired observations. Using t-test a researcher could also compare the means of two samples unequal in size. In our case, an independent group t-test was used. Stata calculates the mean of each value of the independent variable (i.e. Total Early-stage Entrepreneurial Activities). The output also includes standard errors and deviations for each level and the difference of means. The statistical package estimates t-statistic (mean difference divided by the standard error), two-tailed p-value and one-tailed p-value for the alternative hypothesis. If two-tailed p-value is less than б (0.05), then the mean is statistically significant different from zero. The test helps realize if there is the difference of means and considered a necessary step of any quantitative research.

The values are presented for independent variable Total Early-Staged Entrepreneurs (further in the text TEA). The first meaning are for the positive value of the TEA variable that contains respondents, who are involved in the entrepreneurship process and the second is for those, who are not involved in this process.

The column “Mean value” includes the mean value for the variable for both TEA and the column “St. deviation” is for the standard deviation of the variable. T-test and Pr-test were conducted to compare the difference in the means for continuous and categorical types from the variables of two groups. Thus, just two variables that describe the level of education and the preferences for similar standards of living in the country are not statistically significant at the level б=0.05 different from zero.

This particularly tells that the mean difference is not statistically different from zero. The results of the descriptive statistics are presented in the Table 1. It must be said that the major part of independent variables was occupied by categorical indicators coded in GEM Adult Population Survey.

Further, to examine the women motivation entrepreneurial factors and to define the difference between them according to the opportunity or necessity driven type of TEA, it was decided to provide the descriptive statistics for the values of the variable “Age”, as it is presented in the sample. The variable is one of the few numeric variables, according to the codebook. The results are represented in the Table 2 below and conclude the statistics for both male and female groups. On average, for 71,951 non-missing observations (individuals) the value of age was equal to approximately 34.5 years.



Table 1 Descriptive statistics of the variable “Age”

Variable	Obs	Mean	Std. Dev	Min	Max
Age	71,951	39.49974	13.60607	18	99

*Source: generated with the usage of Stata software

The results are also represented graphically in order to see the distribution of the variable and compare it to the normal distribution (Figure 1).

From the Figure 1 it is seemed that the distribution of the “Age” variable is not close to the normal. That is the reason to lead it to the logarithm function. After monotonic transformation values could be used in regression building process.

Table 2 Descriptive statistics of the variables used in the regressions

Variables	Early-Stage Entrepreneurs (TEA==1)	Non-Early-Stage Entrepreneurs (TEA==0)	PR-test/T-test
	Mean value	St. deviation	Mean value	St. deviation	Difference
Current age (in years)	36.78632	11.49806	39.93381	13.87461	3.147494 ***
Current age (ln)	16.3018	2.550448	16.88641	2.970185	.5846133 ***
Gender	.4237958	.4941826	.48665	.4998258	.0628542 ***
Education	910.6	579.1016	906.0166	567.6753	-4.583404
Household Income	26876.54	32175.22	20846.79	30203.42	-6029.754 ***
Number of household members	4.231503	2.361723	3.846847	2.584551	-.384656 ***
Personal knowledge of early-stage entrepreneurs	.6512524	.4765969	.3845097	.4864832	-.2667427 ***
Good opportunity for starting a business where they live	.6315029	.4824204	.3701705	.4828541	-.2613324 ***
Having the required knowledge and skills to start a business	.8474952	.3595268	.4892497	.4998885	-.3582455 ***
Fear of failure as a preventing factor (fear of failure)	.300289	.4584057	.4263174	.4945451	.1260284 ***
Preference for similar standards of living in the country	.6142582	.4867935	.6185478	.4857471	.0042896
People consider starting a business a desirable career choice	.6736994	.4688813	.6071189	.4883948	-.0665805 ***
Those successful at starting a business have a high level of status and respect	.704817	.456147	.6461634	.4781631	-.0586536 ***
Frequent stories in the public media about successful new businesses	.6759152	.4680544	.5962623	.4906501	-.0796529 ***
Easy to start a business in a country	.4669557	.4989309	.3613087	.4803837	-.105647 ***

*** Difference of means is significant at the level б=0.05 for two groups (early-stage entrepreneurs and non-early-stage entrepreneurs) comparison of means tests, calculated in Stata software

Figure 1. The distribution of “Age” variable

Thus, Figure 2 represents the distribution of the variable “Age” taken with logarithm function and line of normal distribution. The changes are observable and the action of taking the logarithm is justified.



Figure 2. The distribution of “Age” variable taken with logarithm function

The next stage was to examine the variable “The number of household members”, the received values are represented in the Table 3.

Table 3 Descriptive statistics of the variable “The number of household members”*

Variable	Obs	Mean	Std. Dev	Min	Max
Household Size	71,951	3.90399	2.403104	-2	90

*Source: generated with the usage of Stata software

As we see, there are inappropriately coded values with negative values. A necessary step is to omit these value because they may not be included in any of probit models run. What is also important, generally a person surveyed with two other people (household size, according to GEM, is coded as the number of people living in one house including the respondent). Standard deviation shows the spread of data. The closer it is to zero, the more reliable the results are.

To check the ability to use this continuous variable in the model building, it is necessary to see its distribution (Figure 3).



Figure 3. The distribution of “The number of household members” variable

It seems that the distribution of the variable “The number of household member” is close to the normal. Then, the variable should not be changed and can be included in regressions used. What is more, all other statistical operations can take place.

Furthermore, to discover the relationship between variables that were selected from the full dataset after literature review examination and variables that may provide the answer to the question: “Are respondents involved in early-stage entrepreneurship or not”, correlation matrixes for both genders: for the female group and for the male group were generated. However, the matrix shows no difference between the case of both genders and the case when genders are separated, that is the reason to present and describe just three matrixes. Mentioned results are advocated by the binary nature of gender variable: it may take only one of two values (either 1 or 0).

The received matrixes do not include the variable “Country” (Appendix 2). For better and easier understanding and comparing the Table 4 was created, which includes motivating factors and respondent's involvement in the entrepreneurship process. The Table 4 represents the results for both gender united. The values at the intersection of variables show the power and significance of the relationship between these variables. The closer it is to zero, the tighter the bond between two variables is. “Country” is the only indicator that determines an individual's place of residence. Some countries could be more or less correlated due to their belonging to one stage of economic development, region, or cultural cluster.

Commonly, the table shows that none of the variables has great and significant correlation with the respondent's involvement in entrepreneurship (p<.001).

In this way, the output of the Table 4 shows that the current age and fear of business failure has a negative correlation with TEA variables. The other following three variables can be highlighted as the most influencing: personal knowledge of somebody, who is involved in the entrepreneurship; good opportunities in the area, where respondent lives; and the presence of required skills and knowledges. The coefficients of these variables cannot be entitled to be significant by testing solely the correlation coefficients. However, comparing with other factors their influence is observable to a greater degree.

Moreover, from the table it can be assumed that there are some differences in coefficients concerning gender and education between entrepreneurship as an opportunity, and as a necessity-driven choice. The gender for both total early-stage entrepreneurial activities and opportunity-driven ones has a negative correlation, while positive with necessity. As female gender is coded as “1” and male gender as “0”, negative correlation between gender variable and opportunity TEA and TEA, in general, says that women have fewer chances to select these professional directions. On the contrary, necessity total-early-stage entrepreneurial activities have a positive association with the female gender. It confirms the ideas of Terell and Troilo (2010) that women less frequently than men hold such crucial work values as “respect” and “achievement” that influence the TEA rate. What is more, women who hold the mentioned values less frequently enter entrepreneurial labor force than men with similar qualities.



Table 4 The coefficients of correlation matrixes, both genders, no country

Variables	TEAyy	TEAyyOPP	TEAyyNEC
TEAyy, TEAyyOPP, TEAyyNEC	1.000	1.000	1.000
Current age (ln)	-0.071	-0.068	-0.018
Gender	-0.044	-0.052	0.003
Education	0.003	0.031	-0.047
Household income	0.069	0.094	-0.023
Number of household members	0.056	0.045	0.032
Personal knowledge of early-stage entrepreneurs	0.190	0.172	0.070
Good opportunity for starting a business where they live	0.187	0.173	0.066
Having the required knowledge and skills to start a business	0.253	0.218	0.110
Fear of failure as a preventing factor	-0.090	-0.086	-0.025
Preference for similar standards of living in the country	-0.003	-0.006	0.006
People consider starting a business a desirable career choice	0.048	0.033	0.036
Those successful at starting a business have a high level of status and respect	0.043	0.037	0.021
Frequent stories in the public media about successful new businesses	0.057	0.050	0.025
Easy to start a business in a country	0.077	0.074	0.021

*Source: correlation matrix generated with the usage of Stata software

It is aligned with the ideas of Walker and Webster (2006) and the Global Entrepreneurship Monitor that female entrepreneurs were generally “pushed” than “pulled” towards employer's activities. Although the growth rate of the female labor force is higher than that of men, there still be a gap between two gender groups in terms of their presence in the total labor force. The same situation with the preferences for similar standards of living in the country. Obviously, the negative institutional perception of entrepreneurs disregards it as a career option. Generally, in people's minds entrepreneurs seem to be wealthier and more successful that employees working for an enterprise. Acs (2010) confirms that society's culture and values can influence both the attractiveness perception of entrepreneurship and can be a barrier to setting up a SME.

From the other hand, the level of education has the negative correlation just with necessity entrepreneurship. As mentioned above, necessity entrepreneurs are usually less educated and skilled in comparison with opportunity entrepreneurs, which was described in the work by Mersha, Sriram, & Hailu (2010). The rest of the variables do not declare any noticeable correlations: all of them have not significant or rather small correlations.

The next stage to evaluate the factors, which motivate women to become entrepreneurs, was the creation of several probit models. All in all, 18 models were created. Basically, they are divided into 3 parts with different factors included and conditions indicated. Each part consists of 6 models, in which dependent variables are being changed among each other.

It means that the first three models describe the predictions for the fact of being involved in early-stage entrepreneurship for both female and male gender, ignoring the country, where the respondents live. First model makes a prediction for the Total Early-Stage Entrepreneurial Activities, the second one for the Opportunity Early-Stage Entrepreneurial Activities and the last one for the Necessity Early-Stage Entrepreneurial Activities. The following three models also make a prediction for both genders; however, the country is included in this case; the dependent variables are the same. The second part of the models includes only the representatives of female gender group. The first three models do not consider the country influencing and the rest of this part do. Finally, the last part includes only the male group of the respondents. In the beginning the country is not included in the probit-regression and the last three models do consider the impact of the country.

The received coefficients from the probit-regression cannot be interpreted directly in terms of influence scale. What is more, their comparison between each other is meaningless. They solely provide general understanding presenting the sign of relationships (negative or positive). Positive sign tells that the positive outcome is more likely. Vice versa, negative sign demonstrates lowered likelihood. That is why, judgments for changes brought by regressors is interpreted with the help of marginal effects calculated in Stata. They are listed for each model in the Appendix 3 of this paper.

In all built models the variable “Age” is a negative indicator for involving into entrepreneurship, it means that the older the person is, the lower probability of him choosing entrepreneurship is. However, some models demonstrate the trend of absent influence of the age variable. It is an arguable effect: some scholars still believe that age is not a barrier; others conclude that older people (especially women) choose entrepreneurship is a way to survive.

In those models where total necessity early-stage entrepreneurship is a dependent variable, age is not a statistically significant factor. Fear of failure demonstrates the largest effect on the choice of entrepreneurship at early-stage for both genders. Chances of opportunity entrepreneurship to be chosen (be equal to one) are also lower when a categorical variable “fear of failure” changes from zero (no fear as a barrier) to one (fear of failure may be a barrier). It confirms the fact that nobody is going to start own business if she or he has an immense fear of future failure. Fear of failure frequently retains one from starting a company. In the context of female entrepreneurship at early stages the probability to be involved in any type of entrepreneurship changes by 2.7%, 2.2% and 0.4% for TEA, opportunity TEA, and necessity TEA respectively. The introduction of dummy variables for countries in this case does not significantly influence the results.

Such education level as “some secondary” can be a motivating factor for opportunity entrepreneurship. Such types as secondary and post-secondary represent negative results for involving in entrepreneurship. It happens as less-skilled people are rarely capable of the red tape of entrepreneurship and experience a large number of difficulties at first stages (Mersha, Sriram, & Hailu, 2010). Any educational attainment reduces the probability of women entering entrepreneurial activities, which is a surprising result.

However, as national context could not be ignored, coming up with the results from the model with dummies for counties, we see that for women education is driver towards entrepreneurship, especially graduated experience. The marginal effect of graduated experience is 4.2% for total early-stage entrepreneurial opportunities for women. Men are also rather “pulled” than “pushed” to entrepreneurship by education as a construct. The marginal effects of post-secondary and graduated experience are the largest (3.5% and 4.5%).

In general, such factors as graduated experienced level of education, the high level of income in the household (which may be a source for capital), fact that those who successful at starting a business have a high level of status and respect, the frequent news in the media about successful new businesses and the easy beginning the entrepreneurship in the country show that these factors are important and motivate for opportunity entrepreneurship for both genders. Nevertheless, the influence of individuals education makes a noticeable difference for men. For instance, once an individual graduates, the probability of his involvement in opportunity TEA rises by 4.5%. And it changes by 3.5% when he receives post-secondary education.

The number of members in the household and the preferences for similar standards of living in the country are not influencing factors just in the case of single gender (female) and in the case of country presence. Only in the case of probit for opportunity TEA prediction for men, a similar standard of living makes his choice to enter this type of career lower by 0.5%. A high level of status and respect in the country are not prominent in models with countries included.

The rest of the factors: personal knowledge of early-stage entrepreneurs, good opportunity for starting a business in the location where the respondents live and the presence of required knowledges and skills, and considering the entrepreneurship as a desirable career choice positively influence on the desire to be involved in entrepreneurship of both types opportunity and necessity.

For instance, in models taking into account the country and measuring total early-stage entrepreneurial activities marginal effect of the variable explaining if a woman knows a nascent entrepreneur is 0.06. For those models measuring opportunity TEA and necessity TEA marginal effects are 0.046 and 0.014. For opportunity entrepreneurship, whenever a categorical variable (knowing others involved in TEA) changes its value from 0 (knows no other early-stage entrepreneurs) to 1 (knows some) the probability to be involved in opportunity TEA (TEA=1) increases by 4.6%. For men this factor is even more influential: marginal effect for TEA is 7.4%. Knowing others, as expected, rather “pull” (marginal effect equal to 0.056) than “push” them (0.018).

Additionally, it must be said that in models focusing separately on women and men success of others and frequent stories in media are not pivotal. Easiness of starting company matters and positively influence the probability of people setting up their SMEs. However, though the coefficient for necessity entrepreneurship and easiness to start a company is not statistically significant, it has a negative sign. It leads to the idea that a person is not “pushed” to entrepreneurship. On the contrary, he or she makes an informed decision.

The coefficients give the predicted probability of different variables influencing on engagement to start the business. And the overall picture is the following (Appendix 3).

The tables show six alternative models testing which factors are the most significant for starting the own business at the level * p<0.05; ** p<0.01. The analysis represents that there is a little difference between women and men motivating factors to begin their own business. As we see there are some factors that almost equally force women to start the business both opportunity and necessity driven. There are graduated level of education, knowledge of somebody, who has already chosen early-stage entrepreneurship, good opportunities to start the business in the areas where they live and the presence of the required knowledges and skills for business dealing. The level of income influences the desire to become or not to become an entrepreneur. Despite that, it is difficult to define as an opportunity or necessity reason.

The reasons to become a necessity entrepreneur for women are secondary and post-secondary education. Fear of fail influences just opportunity entrepreneurship. The rest of the variables are not significantly important to be influential in this context.

In all models observing some country specifics no country shows the significant coefficients across models. In some cases, it can be suggested that woman who is living in the particular country can from time to time change the reason to become an entrepreneur as a reaction to changes in society and politics.

The last stage to estimate the quality of existed models was to build the ROC-curves for each model. The area under ROC curve differs among the models and this difference can be seen in the following table (Table 5).

As this method can allow to determine the quantitative value of the informative reliability between several approaches, in this particular case several models, which include different dependent and independent variables. It can be assumed that the AUC of these models are acceptable, because the lowest value is 0,7003 and the highest 0,8090. The acceptable value of area under curve is more than 0,7 and less then 0,8.

So the best performance of different ROC-curve is by the model that includes only female gender and does not ignore country, the value is 0,8090 (Figure 4). The good indicator means AUC > 0,8, while the worst performance by the model with male representatives that does not include country and the dependent variable is involving in entrepreneurship as a necessity, the value of area under curve is 0,7003 (Figure 5).

In particular ROC-curve demonstrates the relationship between sensitivity of the model and the value equal to the difference between “1” and specificity. The command estat in Stata depicts a detailed result of model's characteristics. The resulting Table 6 is presented below for the model including only women and predicting opportunity TEA. Country is included in the model.

As could be judged by sensitivity and specificity analysis (partially presented in Table 6), only a few women were involved in entrepreneurship. Obviously, specificity of the model is quite high (99.97%) Specificity is the proportion of negative outcomes (a woman is not involved in opportunity TEA) correctly classified as negative. As there were only 18 entrepreneurs involved in TEA and of female gender, sensitivity of the model is rather low (0.3%). A larger sample of positive outcomes is needed. Moreover, if necessity TEA is null, it means that means than an individual could be either unemployed, work for an enterprise or he or she may be an owner of the mature company, which went through early stages.

Table 5 The values of area under the curve across built models

Model	Dependend Variable	Area Under Curve
Both genders ignoring country	TEAyy	0,7633
	TEAyyOPP	0,7740
	TEAyyNEC	0,7164
Both gender and country	TEAyy	0,7899
	TEAyyOPP	0,7980
	TEAyyNEC	0,7566
Female ignoring country	TEAyy	0,7735
	TEAyyOPP	0,7829
	TEAyyNEC	0,7343
Female and country	TEAyy	0,8027
	TEAyyOPP	0,8090
	TEAyyNEC	0,7796
Men ignoring country	TEAyy	0,7524
	TEAyyOPP	0,7635
	TEAyyNEC	0,7003
Men and country	TEAyy	0,7808
	TEAyyOPP	0,7881
	TEAyyNEC	0,7475

*Source: generated with the usage of ROC analysis in Stata software



Figure 4. The ROC-curve for the model, which contains answers of female representatives, including the country

Figure 5. The ROC-curve for the model, which contains answers of male representatives, ignoring the country

Table 6 Measures of the predictive model: sensitivity and specificity (opportunity TEA)

Classified	True	Total
	D	~D
+	9	9	18
-	2994	31151	34145
Total	3003	31160	34163

Similar operation could be carried out for models predicting total early-stage entrepreneurial activities (Table 7).



Table 7 Measures of the predictive model: sensitivity and specificity (TEA)

Classified	True	Total
	D	~D
+	9	9	18
-	2994	31151	34145
Total	3003	31160	34163

In this case sensitivity is 6.46% and specificity 98.97%. The model better predict negative values. All in all, 85.55% of cases are predicted correctly.

The findings on motivating factors to start the business as an opportunity-driven or necessity-driven for women from different countries identify some common outcomes as well as some special features. In both ways to start the business, knowledge of somebody, who is already involved in the entrepreneurial activities is very important for women (4.6% for opportunity TEA and 1.4% for necessity TEA). It does not matter whether those people close friends, members of their families or other people, who create an external entrepreneurial environment. The regression findings show that the probability to choose the entrepreneurship as an opportunity carrier is higher than the probability to choose the entrepreneurship as a necessity because of the knowledge of such people This fact allows to accept the first hypothesis that was created in the section “Theoretical Foundation“.

The hypothesis 2 consisted of two parts such as the increasing of number of household members is the necessity-driven factor for women to become an entrepreneur, and the second part is that this factor relates in the most cases to the men. In reality, this factor influences on the male group more then on the female group, however, it is impossible to say that this factor forces women to start the business, because it is not statistically significant. There are no sufficient data to accept this hypothesis.



Conclusion

Using the Global Individual Level Data from Global Entrepreneurship Monitor that includes the answers of different respondents across 45 countries. In this research paper the analysis of the motivating factors for women to be involved in entrepreneurial activities was conducted. Driving factors are separated into two groups. The first one is the opportunity-driven factors. Factors of this set provides a woman an idea to start the business because it is her own desire and dream. The second group is the factors that force to be an entrepreneur. Usually, it is external factors that makes them choose entrepreneurial career. In this research socio-demographic factors, gender stereotypes, inner thoughts of women and aspects of the external environment were selected to check their importance and impact on entrepreneurial involvement and participation in start-ups. The impact of every motivating factor always varies because there are no identically similar entrepreneurs that will be affected by these factors in the same way.

Thus, there have been several studies which differentiate women motivating factors to start the business. Initially scrutinized researches and findings became the basis for this particular study.

The findings of this research show that some factors can be of opportunity and necessity nature at the same time. The tested factors in this particular research are the following: age, level of education (can take different values: some secondary, secondary, post-secondary, graduated experience), level of income (can take different values: middle and high), the number of household members, personal knowledge of early-stage entrepreneurs, presence of the required skills and knowledges, fear of failure as a barrier, the similar standards of living in the country and the level of difficulty to start the business in the country, the entrepreneurship as a desirable career choice, deserving status and respect, the presence of news about business activities in the public media.

Hereby, the older the women are the more probable her involvement in early-stage entrepreneurial activities by necessity is. As it was mentioned in the literature foundation section above, the younger the women are, the riskier, more open, faster and more adventurous they are. It means that she sees such a career as an opportunity.

The higher level of education raises the possibility to choose the entrepreneurship as an opportunity-driven choice, for the necessity entrepreneurial activities at early stages education, as a motivator, affects oppositely. However, it is worth to mention that similar conclusions are made relatively to the male group.

The higher the household income is the more probable woman's consideration to choose entrepreneurship as an opportunity-driven career is. Such case can be explained by the fact that the nature of any start-up and new enterprises is risky. People with low or middle income in most cases are not ready to experience unexpected adventures knowing how risky they are. From this point, it is important to mention that people who are ready to hold all possible risks, can be more enthusiastic and required skills, needed to be successful in business. Similar skills are the reasons to choose this sphere as an opportunity to perform self-realization.

Surely, good opportunities, which can include government support and strong entrepreneurial atmosphere “pull” women to business sphere. The fear of future failure is the most repulsive factor for any kind of entrepreneurship. For example, if a woman does not want to work anymore for a certain company and has a feeling of failure, she never dares open her own company. Nevertheless, necessity entrepreneurship could be chosen, other things being equal. The preference to a similar standard of living is a “push” motivating factor. Any entrepreneurial activity follows the aim of profit maximization. People do not like those who stand out and could afford more than an average citizen can. As a result, negative pressure dominates. Other people's consideration that entrepreneurship is something good and is achievable only by clever and ambition people forces women to see the entrepreneurship as a way to prove her own significance and importance in life, what can be called a necessity-driven motive. The idea that those who are successful at starting the business have a high level of status and respect expresses an opportunity-driven factor. Many scientific pieces women's emotionality. Receiving status and respect in society allows them to start the business. The theory of gender role and stereotypes fits perfectly to this idea, because women, constantly performing only household tasks, may think that they are not capable of anything better. Doing business gives them an opportunity to believe in themselves, get status and respect. Stories about successful businesses motivate women as well. The feeling that people around make something to their taste and get profit from if, inspires women to try out themselves in the business too. The last observed factor in this study was the difficulty to start the business level in the country. Obviously, a low level of difficulty assures women in receiving assistance whenever they need.

Policy-makers, governmental structures, future necessity-driven entrepreneurs and the mass media can be interested in the results of this research. The understanding of driving factors nature is the first step to increase the rate of entrepreneurship in the world. Policy-makers and governmental structures are interested to raise the economies of countries and create better conditions for the population, increasing standards of living, for instance. It is important to discover these motivating factors among the population asking what reasons or events can spring a desire to start a company (Iakovleva, Nikolaeva, 2019).

The implementations of the findings can be the following. Opportunity-driven factors are the motives which should be multiplied by policy-makers and governmental structures mentioned above. Their main tasks are to create initiatives and perspectives for female entrepreneurs and to develop honest and comfortable field of competition. Such methods cannot only increase a female entrepreneurship rate but make it equal to the one of men. Women need support and recognition of their skills and knowledge. To get a virtue representatives of authorities should create more creative and entrepreneurial atmosphere which will help them easily start small and medium companies. As a woman can be more emotional a man, the atmosphere is a crucial factor. Communication, access to knowledge and efficient networking help increase the level of involvement in the entrepreneurial processes. Women with intention can use the findings if they want to transform their necessity-driven factors to opportunity ones. It is significant because opportunity entrepreneurial activities find their continuance in larger firms developed from start-ups.

There is no sense to believe that bringing fine entrepreneurial will make women reluctant to continue working for an enterprise. Moreover, it does not mean that work conditions should be decreased to make women leave and choose entrepreneurship. All improvements in the business sphere will give more opportunities and alternatives not just for the female part of the population, but for the population in general.

Today it is very important to talk about something trendy. People are keen on doing things they think well of; they share their admiration with friends via social networks. Currently, there are a lot of sources of information and possibilities to work as an entrepreneur. For these reasons public media can make a contribution by giving a word, posting facts, news and information, offering events and promoting conference for those who are interested in business.

All in all, it is important to focus that opportunity or necessity motivating factors are really close to each other and correlate in most cases. This statement can be a limitation of this research because it is very challenging to distinguish two groups. Moreover, the analyzed data are sourced from the open source. The data are used in many scientific papers bringing similar results. The Global Entrepreneurship Monitor is a relevant and appropriate way to learn and observe the entrepreneurship all over the world. However, to improve the research other sources of data or methods of the analyses can be added. Furthermore, the study can be improved adding the industry of operating. It is interesting to explore how earnings depend on the type of entrepreneurial activities. Women prefer to choose industries close to their interests and simultaneously less profitable than typically "male" industries. It could be the reason to start, for instance, a business in technological, energy or oil sphere just because of higher income promised. In such industries, the level of managerial experience need is higher. They take more time and may interrupt work-life balance beloved by women. It means that there are no other ways to spend time with a family, have hobbies and dedicate time to herself. So, it is interesting to investigate the contribution of industry on opportunity or necessity entrepreneurship. Similar research approach could be used for conducting a study in developing and developed countries where female activities in “male” industries are possible.

From the other hand, research can be improved by adding expert opinions and case studies. It is possible to conduct in-depth interviews and investigate how people understand opportunity and necessity factors to start a business, how they distinguish two groups of motives, what ways they believe motivational factor could be emphasized and what can make it real. The representatives may be of both genders to compare the results. It is wise to perform a study in this way because the most relevant female motives may lack sense for men.

There is an urgent need to give a chance to similar researches to be developed. That is because a lot of women today are trying to be at the one level with men, create and develop meaningful projects for the population and be confident in the result's virtue.

entrepreneurship motivation prompting



Reference list

1. Aмcs, Z. J., Szerb, L., & Lloyd, A. (2017). The Global Entrepreneurship Index. Washington, D.C.: The Global Entrepreneurship and Development Institute. Retrieved from THEGEDI: https://thegedi.org/research/gedi-index.

2. Ahl, H. (2006). Why Research on Women Entrepreneurs Needs New Directions. Entrepreneurship Theory and Practice, 30(5), 595-621. doi: 10.1111/j.1540-6520.2006.00138.x.

3. Aidis, R., Weeks, J. (2016). Mapping the gendered ecosystem: The evolution of measurement tools for comparative high-impact female entrepreneur development. International Journal of Gender and Entrepreneurship, Vol. 8 Issue: 4, 330-352.

4. Azoriмn, J. M., & Cameron, R. (2010). The Application of Mixed Methods in Organisational Research: A Literature Review. Electronic Journal of Business Research Methods, 95-105.

5. Baycan Levent, T., Masurel, E., & Nijkamp, P. (2003). Diversity in entrepreneurship: ethnic and female roles in urban economic life. International journal of social economics, 30(11), 1131-1161.

6. Berndt, T. J. (1986). Children's Comments about Their Friendships. In M. Perlmutter (Ed.), Minnesota Symposia in Child Development: Cognitive Perspectives on Children's Social and Behavioral Development, Vol. 18, 189-212.

7. Bianchi, M., Parisi, V., & Salvatore, R. (2016). Female entrepreneurs: motivations and constraints. An Italian regional study. International Journal Of Gender And Entrepreneurship, 8(3), 198-220. doi: 10.1108/ijge-08-2015-0029.

8. Bosma, N., & Kelley, D. (2019). Global Entrepreneurship Monitor 2018/2019 Global Report.

9. Bui, H. T., Kuan, A., & Chu, T. T. (2018). Female entrepreneurship in patriarchal society: motivation and challenges. Journal of Small Business & Entrepreneurship, 1-19.

10. Bygrave, W. D., Zacharakis A. (2010). "Chapter 1 - The Entrepreneurial Process".?The Portable MBA in Entrepreneurship, Fourth Edition.?John Wiley & Sons.

11. Cialdini, R. B. (1993). Influence: The psychology of persuasion. New York: Morrow.

12. Eagly, A.H. (1987). Sex differences in social behavior: the Social-role interpretation. Hillsdale, New Jersey: Lawrence Erlbaum.

13. Escribanoa, J. J., & Casado, A. B. (2016). Construction of gender differences in the discourse of entrepreneurship. Psychobiological, cultural and familiar aspects. SUMA DE NEGOCIOS, 18-24.

14. Eurostat Official Site. Entrepreneurship indicator programme. Retrieved from Eurostat: https://ec.europa.eu/eurostat/web/structural-business-statistics/entrepreneurship/indicators.

15. Field, E., Jayachandran, S., Pande R., Rigol N. (2016). "Friendship at Work: Can Peer Effects Catalyze Female Entrepreneurship?" American Economic Journal: Economic Policy, 8 (2), 125-53.

16. Garcнa-Lillo, F., Claver-Cortйs, E., Marco-Lajara, B., & Ъbeda-Garcнa, M. (2019). Identifying the `knowledge base' or `intellectual structure' of research on international business, 2000-2015: A citation/co-citation analysis of JIBS. International Business Review. doi: 10.1016/j.ibusrev.2019.02.001.

17. GEDI Official Site. Institute. Retrieved from The Global Entrepreneurship and Development Institute: https://thegedi.org/theinstitute/.