Do small and medium enterprises perceive finance more severely an obstacle than larger businesses?

Replication project based on the paper

Wang, Y., (2016). What are the biggest obstacles to growth of SMEs in developing countries? – An empirical evidence from an enterprise survey. Borsa Istanbul Review [online]. 16(3), 67–176. Available from: doi: 10.1016/j.bir.2016.06.001

## 
## Attaching package: 'dplyr'
## The following objects are masked from 'package:stats':
## 
##     filter, lag
## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union
## This is nonnest2 0.5-3.
## nonnest2 has not been tested with all combinations of model classes.
## Loading required package: Matrix
## Loading required package: arm
## Loading required package: MASS
## 
## Attaching package: 'MASS'
## The following object is masked from 'package:dplyr':
## 
##     select
## 
## arm (Version 1.10-1, built: 2018-4-12)
## Working directory is C:/Users/ggfun/Documents/R/win-library/3.6
## Install package "strengejacke" from GitHub (`devtools::install_github("strengejacke/strengejacke")`) to load all sj-packages at once!
## 
## Attaching package: 'pracma'
## The following object is masked from 'package:arm':
## 
##     logit
## The following objects are masked from 'package:Matrix':
## 
##     expm, lu, tril, triu
## Loading required package: carData
## Registered S3 methods overwritten by 'car':
##   method                          from
##   influence.merMod                lme4
##   cooks.distance.influence.merMod lme4
##   dfbeta.influence.merMod         lme4
##   dfbetas.influence.merMod        lme4
## 
## Attaching package: 'car'
## The following object is masked from 'package:pracma':
## 
##     logit
## The following object is masked from 'package:arm':
## 
##     logit
## The following object is masked from 'package:dplyr':
## 
##     recode
## 
## Attaching package: 'tidyr'
## The following objects are masked from 'package:Matrix':
## 
##     expand, pack, unpack
## Registered S3 method overwritten by 'pryr':
##   method      from
##   print.bytes Rcpp
## For best results, restart R session and update pander using devtools:: or remotes::install_github('rapporter/pander')

1. Introduction

Wang (2016), explores the main obstacles to business growth, as perceived by the business managers, that are facing small and medium enterprises globally. This paper is a replication that further investigates the largest perceived obstacle to business growth, finance. It is widely discussed in both literature and the media that small and medium enterprises are essential for encouraging economic development in emerging economies (Stawasz & Glodek 2010; Howard 1990).This analysis concerns significant obstacles that small-medium enterprises are facing as a collective, using data from the World Bank Group (2019). That said, recognising and mitigating these common issues is essential for economic development. This topic is discussed further in the replication.

2. Literature review

Wang (2016) situates his paper well amongst the literature; many authors have written cases for specific countries that explore challenges facing small businesses. For example, a study in Cameroon determined that financing constraints and corruption hindered firm growth (St-Pierre et al. 2018). Other studies support this claim that finance is the biggest obstacle (Tambunan 2008; Erdogan 2015).

Yet, there was a lack of literature on the subject concerning small enterprises on a global scale making his contribution necessary. The paper clearly discusses data, hypothesis, and variable coding reasoning, propositioning a well-structured study. However, the main detriment of the paper is the methodology; the one-level model is not suitable for the cross-sectional data structure. This was acknowledged by adding dummy variables for country and industry levels, aiming to control heterogeneity (Wang 2016, p.172). However, a multilevel model would have been more appropriate. Due to the large dataset size, there is likely individual errors and heterogeneity within results as suggested by Duncan and Jones (2000). Hence, the multilevel model has been adopted in replication.

Additionally, whilst most of the variable coding and literature supporting this is agreeable, literature contradicts the coding of the variable government ownership. For example, it is coded 1 when the government has any stake of ownership. However, the European Commission (2015) states that if 25% or more of its capital or voting rights are owned by public bodies then the business can not be classified as a small-medium enterprise. Consequently, it is not relevant to include businesses with any amount of government ownership if the focus of the study is small-medium enterprises. Furthermore, this argument does not provide inherent value because literature already supports that businesses with considerable government ownership receive generous public subsidies and resources (Tan 2002). Despite this, this may change within different country contexts.

Finally, given the survey was obtained from the manager’s perception, gender perspectives may have been interesting to include. Research supports that there are differences in perception between males and females (Smith & Oakley 1997). It can be determined in replication if there are specific differences in business attitudes and behaviours

2. Data and methodology

2.1 Data

The data used in Wang’s paper was obtained from the Enterprise Survey from the World Bank Group (2019). This includes data from over 130,000 firms in 135 countries of which obtained a sample size of 117105 observations.

2.2 Methodology

In paper replication, the methodology has been altered to a 3-level logistic model, so the intercept can vary at the country and year level. The years of study, 2006 to 2014, remain the same in this replication. Whilst the purpose of the original paper was to examine common obstacles that all small-medium businesses face, differences in these effects among countries and years are likely to exist. Furthermore, to account for heterogeneity between groups, this method is most suitable (Duncan & Jones 200). Remaining to the premise of the original paper, observing the collective obstacles, a random slopes model has not been included because I am not interested in individual differences. Additionally, a gender variable has been included and the government ownership variable has been recoded.

2.3 Variables

Dependants

  • finance – Binary variable with access to finance as the perceived biggest obstacle to business growth (1) or not (0)

Explanatory variables

  • sme - Binary variable: small-medium enterprises consisting of less than 100 employees (1) and not (0)

  • hgf – Binary variable indicating whether the business is a high growth firm (1) or not (0). High growth is indicated by whether the firm employee size has expanded by 20% or more in the last three years

  • age – The firm age

  • experience – The top managers years of experience working in the sector

  • employees – Number of employees working at the firm

  • gender – The top manager is a female (1) and not (0)

  • govowned – The businesses has 25% - 100% or 0% government ownership (0), 1% - 24% government ownership (1)

Null model

As shown in Table 1 from the ICC, approximately 9% of the variance is explained from the data structure varying by country and year.
Table 1. Base Multilevel logistic regression models of finance perception
  finance
Predictors Estimates Conf. Int (95%)
(Intercept) 0.16 *** 0.13 – 0.20
Random Effects
σ2 0.12
τ00 country 0.01
τ00 year 0.00
ICC 0.08
N country 135
N year 9
Observations 110529
Marginal R2 / Conditional R2 0.000 / 0.080
  • p<0.05   ** p<0.01   *** p<0.001

3. Results

3.1 Results for all businesses

As seen in figure 3 below, the indicator small-medium enterprise demonstrated a significant result; the log-likelihood increased 29% more for the managers of small-medium enterprises than managers from larger businesses. Moreover, managers of all high growth firms were 9% more likely to state finance as a major obstacle for all businesses.

3.2 Results for just SME businesses

The indicator for high growth firms increased the log-likelihood 9% (see Table 2). This means that owners of vastly expanding firms were more likely to perceive issues with financing. This is consistent with the finding of the indicator variable employees being significant albeit experiencing a less than 0.01 intercept change (see Table 2). Meaning that the smaller firms perceived financing as an obstacle more severely. This suggests that once firms mature out of the period of high growth and expand more gradually, the financing obstacle should lessen. It has been argued that job creation is predominantly from industries with below-average levels of productivity and likely to pay 20% less than larger firms (OECD 2019). Without these financial barriers, both employment and productivity growth could have the potential to increase with more room for investment into innovation and training.

Despite some variability being explained, results did not drastically change, suggesting that Yang’s findings were still consistent with the literature available and a multilevel model was not a necessity for the analysis. Both the manager’s experience and firm age were still insignificant to the model.

Table 2| 3 level regression model displaying indicators of finance perception
  finance
Predictors Odds Ratios Conf. Int (95%)
(Intercept) 0.13 *** 0.09 – 0.18
hgf 1.09 ** 1.03 – 1.16
sme 1.34 *** 1.24 – 1.44
age 1.00 1.00 – 1.00
govowned 0.86 0.67 – 1.11
experience 1.00 1.00 – 1.00
employees 1.00 *** 1.00 – 1.00
gender 0.93 * 0.87 – 0.99
Random Effects
σ2 3.29
τ00 country 0.66
τ00 year 0.16
ICC 0.20
N country 119
N year 8
Observations 71366
Marginal R2 / Conditional R2 0.007 / 0.204
  • p<0.05   ** p<0.01   *** p<0.001

3.3 Recoded variable differences

Gender

Gender effects proved somewhat significant to the model (see Table 2). Female managers were more likely to perceive finance as an obstacle. However, there were roughly only 14,000 female respondents in comparison to 85,000 males as most firms chosen for the survey are from developing in countries, in which there were fewer female managers from the time of data collection (World Bank Group 2019).Perhaps with a more even ratio of female and male managers, more conclusive results can be made.

gender
        0         1 
0.8575713 0.1424287

In contrast, manager differences may not affect perception at all. It is argued that survey responses based on perception are not the most accurate representation of true results because perception is subjective (Lavrakas 2008). Considering these findings are consistent with the literature, it is assumed that these perceptions represent real barriers.

Government ownership

Government ownership was no longer significant to the model due to the much smaller sample of small-medium enterprises with 1-24% government ownership. Although the original conclusions were agreeing with theory. Due to data privacy reasons, specific information on the individual companies and to what extent the government is involved in proceedings is unavailable. Consequently, decisive answers can not be made on whether small-medium enterprises with partial government ownership perceive financial obstacles more so than non-government owned enterprises.

4 Conclusions

Evidence amongst previous literature and this paper indicates that finance is perceived as the biggest obstacle to small-medium enterprises, more so than larger firms. Further, smaller high-growth firms are even more likely to state finance as a growth hindrance.

Small-medium enterprises accelerate economic growth, given these results, it is imperative that finance is more accessible. Personal managerial differences such as gender and experience did not influence the model, indicating that the survey collection method is appropriate for analysis on this topic. Whilst the multilevel model acknowledged the country and year context, the overall results did not significantly change from the original model. Consequently, replication results verify Wang’s paper.

#4.1 Limitations and further studies Whilst costly for such a large sample, further studies should consider incorporating panel data to understand if the population perception of obstacles differs over time. Both the age and number of employees significantly influenced the model. Accordingly, it would be interesting to understand if this obstacle lessens as the firm expands.

#5. References Duncan, C. & Jones, K., (2000). Using Multilevel Models to Model Heterogeneity: Potential and Pitfalls. Geographical Analysis [online]. 32(4), 279–305. [Viewed 23 May 2020]. Available from: [doi: 10.1111/j.1538-4632.2000.tb00429.x]

Erdogan, A., (2015). WHICH SMES PERCEIVE ACCESS TO FINANCE AS AN OBSTACLE TO THEIR OPERATIONS? EVIDENCE FROM TURKEY. Journal of Economic and Social Development [online]. 2(2), 13–19. [Viewed 23 May 2020]. Available from: https://search-proquest-com.sheffield.idm.oclc.org/docview/1725178732?accountid=13828

European Commission ; Directorate-General for Internal Market, Industry, Entrepreneurship Smes., (2015). User guide to the SME definition[online]. Luxembourg: Publications Office. [Viewed 23 May 2020]. [Available from: doi: 10.2873/782201]

Howard, R. (1990). Can small business help countries compete? Harvard Business Review [online]. Nov /Dec. [Viewed 23 May 2020]. Available from: https://hbr.org/1990/11/can-small-business-help-countries-compete

OECD., (2019). OECD SME and Entrepreneurship Outlook 2019 [online]. Paris: OECD Publishing. [Viewed 23 May 2020]. Available from: https://doi.org/10.1787/34907e9c-en

Lavrakas, P.J (2008). Encyclopedia of survey research methods, vol. 0, Sage Publications, Inc., Thousand Oaks, CA, [Viewed 24 May 2020]. Available from: [doi: 10.4135/9781412963947]

Smith, P. & Oakley, L., (1997). Gender-Related Differences in Ethical and Social Values of Business Students: Implications for Management. Journal of Business Ethics [online]. 16(1), 37–45. [Viewed 23 May 2020]. Available from: [doi: 10.1023/A:1017995530951]

Stawasz, E. & Glodek, P., (2010). SMEs innovation and job creation potential in the shadow economy context. Comparative Economic Research [online]. 13(4), 99–115. [Viewed 23 May 2020]. Available from: [doi: 10.2478/v10103-009-0048-x]

St-Pierre, J. et al., (2015). SME Development Challenges in Cameroon: An Entrepreneurial Ecosystem Perspective. Transnational Corporations Review [online].7(4). 441–462. [Viewed 23 May 2020]. Available from: [doi: 10.5148/tncr.2015.7405]

Tan, J., (2002). Impact of Ownership Type on Environment–Strategy Linkage and Performance: Evidence from a Transitional Economy. Journal of Management Studies [online]. 39(3), 333–354. [Viewed 23 May 2020]. Available from: [doi: 10.1111/1467-6486.00295]

Tambunan, T., (2008). SME development, economic growth, and government intervention in a developing country: The Indonesian story. Journal of International Entrepreneurship [online]. 6(4), 147–167. [Viewed 23 May 2020]. Available from: [doi: 10.1007/s10843-008-0025-7]

The World Bank Group., (2019). Enterprise Surveys [Data set]. [Accessed 24 May 2020]. Available from: https://datacatalog.worldbank.org/dataset/enterprise-surveys

Wang, Y., (2016). What are the biggest obstacles to growth of SMEs in developing countries? – An empirical evidence from an enterprise survey.Borsa Istanbul Review [online]. 16(3), 67–176. [Viewed 23 May 2020]. Available from: [doi: 10.1016/j.bir.2016.06.001]

Appendix

All R Code used

library(readstata13)
library(foreign)
library(dplyr)
library(nonnest2)
library(lme4)
library(powerlmm)
library(merTools)
library(sjPlot)
library(vpc)
library(pracma)
library(ggplot2)
library(car)
library(ggplot2)
library(ggpubr)
library(tidyr)
library(summarytools)
library(knitr)

opts_chunk$set(echo=TRUE,
                 cache=TRUE,
               comment=NA,
               message=FALSE,
               warning=FALSE)
indicators <- read.dta("C:/Users/ggfun/Documents/R/win-library/3.6/ES-indicator.dta")

mydata <- read.dta("C:/Users/ggfun/Documents/R/win-library/3.6/Newdata.dta")

#2.2 Data preparation: downloading data and subsetting

busdat1 = subset(indicators, select = c(  size, obst1 , obst14 , obst12 , obst8 , obst11 ,  wk10, perf2, car4, car1, wk8, car8 , region , country_official , year , car4 , wk12 , gend4, obst2, obst3, obst4 , obst5, obst6, obst7, obst9, obst10, obst13, obst15))


(busdat1 <- busdat1[busdat1$year %in% c("2006", "2007", "2008" , "2009" , "2010" , "2011" , "2012", "2013", "2014" , "wk12"), ])

#2.3 Renaming variable
names(busdat1)[names(busdat1) == "wk12"] <- "employees"
names(busdat1)[names(busdat1) == "country_official"] <- "country"
names(busdat1)[names(busdat1) == "wk8"] <- "experience"
names(busdat1)[names(busdat1) == "car1"] <- "age"
names(busdat1)[names(busdat1) == "perf2"] <- "employgrowth"
names(busdat1)[names(busdat1) == "wk10"] <- "uneducated"
names(busdat1)[names(busdat1) == "car4"] <- "govowned"
names(busdat1)[names(busdat1) == "obst1"] <- "finance"
names(busdat1)[names(busdat1) == "obst14"] <- "tax_rates"
names(busdat1)[names(busdat1) == "obst12"] <- "competitors"
names(busdat1)[names(busdat1) == "obst8"] <- "electricity"
names(busdat1)[names(busdat1) == "obst11"] <- "political"
names(busdat1)[names(busdat1) == "gend4"] <- "gender"
names(busdat1)[names(busdat1) == "obst2"] <- "land_access"
names(busdat1)[names(busdat1) == "obst3"] <- "licensing_permits"
names(busdat1)[names(busdat1) == "obst4"] <- "corruption"
names(busdat1)[names(busdat1) == "obst5"] <- "courts"
names(busdat1)[names(busdat1) == "obst6"] <- "crime"
names(busdat1)[names(busdat1) == "obst7"] <- "trade_regulation"
names(busdat1)[names(busdat1) == "obst9"] <- "uneducated_workforce"
names(busdat1)[names(busdat1) == "obst10"] <- "Labour_regulationl"
names(busdat1)[names(busdat1) == "obst13"] <- "tax_administration"
names(busdat1)[names(busdat1) == "obst15"] <- "transport"

busdat1$sme <- busdat1$size


#Recoding values
busdat1$finance[busdat1$finance== 100] <- 1
busdat1$tax_rates[busdat1$tax_rates== 100] <- 1
busdat1$competitors[busdat1$competitors== 100] <- 1
busdat1$electricity[busdat1$electricity== 100] <- 1
busdat1$political[busdat1$political == 100] <- 1
busdat1$experience[busdat1$experience == 100] <- 1
busdat1$gender[busdat1$gender == 100] <- 1

#Small/medium = 1 and big firm = 0
busdat1$sme <- as.numeric(busdat1$sme) 
busdat1$sme[busdat1$sme== 3] <- 0
busdat1$sme[busdat1$sme== 2] <- 1 



# 1 if high growth firm and 0 if not
#Less than 20% growth after 3 years means not high growth
busdat1$hgf <- ifelse(busdat1$employgrowth <20,0,1)

#25% to 100% ownership means not classified as sme 
busdat1$govowned <- recode(busdat1$govowned, "25:100=0; 1:24 = 1")

busdat2 <- busdat1
busdat2$sme <- busdat2$sme == 1 

busdat2$sme[busdat2$sme == 0] <- NA


#Data coding for graph
busdat3 <- busdat1

busdat3$finance[busdat3$finance== 1] <- "finance"
busdat3$tax_rates[busdat3$tax_rates== 1] <- "tax_rates"
busdat3$competitors[busdat3$competitors== 1] <- "competitors"
busdat3$electricity[busdat3$electricity== 1] <- "electricity_shortages"
busdat3$political[busdat3$political== 1] <- "political_issues"
busdat3$land_access[busdat3$land_access== 100] <- "land_access"
busdat3$licensing_permits[busdat3$licensing_permits== 100] <- "licensing_permits"
busdat3$corruption[busdat3$corruption== 100] <- "corruption"
busdat3$courts[busdat3$courts== 100] <- "courts"
busdat3$crime[busdat3$crime== 100] <- "crime"
busdat3$trade_regulation[busdat3$trade_regulation== 100] <- "trade_regulation"
busdat3$uneducated_workforce[busdat3$uneducated_workforce== 100] <- "uneducated_workforce"
busdat3$Labour_regulationl[busdat3$Labour_regulationl== 100] <- "labour_regulation"
busdat3$tax_administration[busdat3$tax_administration== 100] <- "tax_administration"
busdat3$transport[busdat3$transport== 100] <- "transport"

busdat3$sme <- as.numeric(busdat1$sme) 
busdat3$sme[busdat1$sme== 3] <- 0
busdat3$sme[busdat1$sme== 2] <- 1 

busdat3$sme[busdat3$sme == 1] <- "Small/medium_enterprise"
busdat3$sme[busdat3$sme == 0] <- "Not_sme"


busdat3$biggestobstacle <- paste(busdat3$finance,busdat3$tax_rates, busdat3$competitors, busdat3$electricity, busdat3$political, busdat3$land_access, busdat3$licensing_permits, busdat3$corruption, busdat3$courts, busdat3$crime, busdat3$trade_regulation, busdat3$uneducated_workforce, busdat3$Labour_regulationl, busdat3$uneducated_workforce, busdat3$Labour_regulationl, busdat3$tax_administration, busdat3$transport)

busdat3$biggestobstacle[busdat3$biggestobstacle== "finance 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0"
] <- "finance"

busdat3$biggestobstacle[busdat3$biggestobstacle== "0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 tax_administration 0"] <- "tax_administration"

busdat3$biggestobstacle[busdat3$biggestobstacle== "0 tax_rates 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0"] <- "tax_rates"

busdat3$biggestobstacle[busdat3$biggestobstacle== "0 0 competitors 0 0 0 0 0 0 0 0 0 0 0 0 0 0"] <- "competitors"

busdat3$biggestobstacle[busdat3$biggestobstacle== "0 0 0 electricity_shortages 0 0 0 0 0 0 0 0 0 0 0 0 0"] <- "electricity_shortages"

busdat3$biggestobstacle[busdat3$biggestobstacle== "0 0 0 0 political_issues 0 0 0 0 0 0 0 0 0 0 0 0"] <- "political_issues"

busdat3$biggestobstacle[busdat3$biggestobstacle== "0 0 0 0 0 land_access 0 0 0 0 0 0 0 0 0 0 0"] <- "land_access"

busdat3$biggestobstacle[busdat3$biggestobstacle== "0 0 0 0 0 0 licensing_permits 0 0 0 0 0 0 0 0 0 0"] <- "licensing_permits"

busdat3$biggestobstacle[busdat3$biggestobstacle== "0 0 0 0 0 0 0 corruption 0 0 0 0 0 0 0 0 0"] <- "corruption"

busdat3$biggestobstacle[busdat3$biggestobstacle== "0 0 0 0 0 0 0 0 courts 0 0 0 0 0 0 0 0"] <- "courts"

busdat3$biggestobstacle[busdat3$biggestobstacle== "0 0 0 0 0 0 0 0 0 crime 0 0 0 0 0 0 0"] <- "crime"

busdat3$biggestobstacle[busdat3$biggestobstacle== "0 0 0 0 0 0 0 0 0 0 trade_regulation 0 0 0 0 0 0"] <- "trade_regulation"

busdat3$biggestobstacle[busdat3$biggestobstacle== "0 0 0 0 0 0 0 0 0 0 0 uneducated_workforce 0 uneducated_workforce 0 0 0"] <- "uneducated_workforce"

busdat3$biggestobstacle[busdat3$biggestobstacle== "0 0 0 0 0 0 0 0 0 0 0 0 labour_regulation 0 labour_regulation 0 0"] <- "labour_regulation"

busdat3$biggestobstacle[busdat3$biggestobstacle== "0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 transport"] <- "transport"

busdat3$biggestobstacle[busdat3$biggestobstacle== "0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0"] <- "N/A"

busdat3$biggestobstacle[busdat3$biggestobstacle== "NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA"] <- "N/A"


  
#Data for percentage of firm rating obstacles
df3 <- busdat3 %>%
  filter(sme %in% c("Small/medium_enterprise", "Not_sme")) %>%
  group_by(biggestobstacle, sme) %>%
  count(biggestobstacle, sme) %>%
  group_by(biggestobstacle) %>%
    mutate(percent = n / sum(n) * 100) %>%
    ungroup()


ggplot(df3, aes(x = biggestobstacle, y = percent, fill = sme)) +
     geom_bar(stat = "identity", position = "dodge") +theme(axis.text.x = element_text(angle = 90) ) +ggtitle("Figure 1. Percentage of firms perceived obstacles")

     

  
#Data for firm counts of obstacles

ggplot(df3, aes(x = biggestobstacle, y = n, fill = sme)) +
     geom_bar(stat = "identity", position = "dodge") +theme(axis.text.x = element_text(angle = 90) ) +ggtitle("Figure 2. Number of firms votes for perceived obstacles")
     

#Descriptive Statistics 


#Percentage of sme to non sme firms 
#More sme than non
crostab_2=xtabs(~ sme, data=busdat1)
prop.table(crostab_2)

#Percentage of firms that descrive finance as obstacke
#More vote it to not be an obstacle
crostab_3=xtabs(~ finance, data=busdat1)
prop.table(crostab_3)

crostab_4=xtabs(~ finance + sme, data=busdat1)
prop.table(crostab_4)


#Null model multilevel regression
nullmodel1 <- glmer(finance ~ (1 | country ) + (1 | year), data= busdat2)

tab_model(nullmodel1,
            string.ci = "Conf. Int (95%)",
            p.style = "a", title = "Table 1. Base Multilevel logistic regression models of finance perception"
)


MFINlm <- glm(finance ~ hgf + sme + age + employees +  govowned + experience , gender, data = busdat1 ,  family = "binomial")



MFIN <- glmer(finance ~ hgf + sme + age +  govowned + experience + employees + gender + (1 | country) + (1|year),  data = busdat1, family = "binomial", nAGQ=0)

plot_model(MFIN, transform = NULL, show.values = TRUE,  value.offset = .4)+ggtitle("Figure 3. Factors influencing the perception of financial constraints on all businesses")


MFIN2 <- glmer(finance ~ hgf + age +  govowned + experience + employees + gender + (1 | country) + (1|year),  data = busdat2, family = "binomial", nAGQ=0)


tab_model(MFIN,
    string.ci = "Conf. Int (95%)",
     p.style = "a", title = "Table 2| 3 level regression model displaying indicators of finance perception"
)


#Percentage of male to female managers 
#0 = Male, #1 = Female
crostab_1=xtabs(~ gender, data=busdat1)
prop.table(crostab_1)


#6. Data visualization