Preliminary Analysis

Correlation Matrix for Preliminary Set of Variables

Basic correlation matrix with some preliminary variables.

• IRR2010: Rural index (larger = rural)
• proprietors_2017: # of sole proprietors in TX
• pct_proprietors_employment_2017: % of sole proprietors in total employment
• pct_broadband_FCC: FCC broadband availability in % (2017)
• pct_broadband_MS: MS broadband availability in % (2019)
• venturedensitydec19: Venture density by GoDaddy
• highlyactive_vddec19: Highly active venture density by GoDaddy
• prosperityindex2016: Prosperity index
• frac_BB.dec: % of M-Lab testers with over 25mbps of DL speed & 3mbps of UL

#### Correlation between variabes ####

## Basic Correlation Table ##

tx_bb_entrepreneur_merged %>% 
  select(IRR2010, proprietors_2017, pct_proprietors_employment_2017, pct_broadband_FCC, pct_broadband_MS,
         venturedensitydec19, highlyactive_vddec19, prosperityindex2016, frac_BB.dec) %>% 
  cor(method = "pearson", use = "complete.obs") %>% knitr::kable(format = "markdown", digits = 3)

#install.packages("PerformanceAnalytics")

tx_bb_entrepreneur_merged %>% 
  select(IRR2010, proprietors_2017, pct_proprietors_employment_2017, pct_broadband_FCC, pct_broadband_MS,
         venturedensitydec19, highlyactive_vddec19, prosperityindex2016, frac_BB.dec) %>%
  PerformanceAnalytics::chart.Correlation(histogram = T)

Preliminary Exploration of Relationships further with Regressions

Few models to examine relationships between entrepreneurship and broadband. In general, I take the entrepreneurship measures as the dependent variable and others as the independent. Following the ASU white paper analysis, venture density is explored as a measure of entrepreneurial activities online. Sole proprietors share in total employment is taken into IV as it represents general small-mid sized business activities, but not all of them are active online as reflected in the venture density factor.

#### Exploring the relationships further with regressions ####

## DV: Venture Density as Entrepreneurship Index
## IV: Rural index, Proprietors share in employment, Broadband (FCC), Broadband (MS), Broadband (M-Lab), Broadband Subscription
#install.packages("stargazer")
library(stargazer)

prem_model <- lm(venturedensitydec19 ~ IRR2010 + pct_proprietors_employment_2017 + pct_broadband_FCC + pct_broadband_MS + frac_BB.dec + pctbbfrac_ASU, data = tx_bb_entrepreneur_merged)

## DV: Highly active venture density

prem_model2 <- lm(highlyactive_vddec19 ~ IRR2010 + pct_proprietors_employment_2017 + pct_broadband_FCC + pct_broadband_MS + frac_BB.dec + pctbbfrac_ASU, data = tx_bb_entrepreneur_merged)

## DV: Proprietor's share in employment as an indicator of entrepreneurship

prem_model3 <- lm(pct_proprietors_employment_2017 ~ IRR2010 + pct_broadband_FCC + pct_broadband_MS + frac_BB.dec + pctbbfrac_ASU, data = tx_bb_entrepreneur_merged)

Comaparing Different Broadband Measures

Preliminary explorations reveal interesting discrepancies between different measures of broadband. Here I will explore how these measures paint different pictures of broadband of Texas.

#### Comparing Broadband Measures in the Dataset ####

## Broadband measures
## pct_broadband_FCC: FCC's broadband availability (reported by the service providers)
## pct_broadband_MS: MS's broadband availability (reported by the MS software service users)
## frac_BB.dec: Broadband speed users' share based on M-Lab data (reported by the M-Lab test respondents)
## pctbbfrac_ASU: ASU team's measure derived from the ACS survey estimates (based on reported BB subscription info from respondents)

## Summarize some key statistics of all these broadband variables

# Some basic statistics
tx_bb_entrepreneur_merged %>% 
  select(c("pct_broadband_FCC", "pct_broadband_MS", "frac_BB.dec", "pctbbfrac_ASU")) %>% 
  psych::describe() %>% knitr::kable(digits = 3)

# Who are the counties with minimum BB?

# According to FCC BB data
tx_bb_entrepreneur_merged[which(tx_bb_entrepreneur_merged$pct_broadband_FCC == min(tx_bb_entrepreneur_merged$pct_broadband_FCC, na.rm = T)),]$county

## [1] "Llano County" "Upton County"

# According to MS BB data
tx_bb_entrepreneur_merged[which(tx_bb_entrepreneur_merged$pct_broadband_MS == min(tx_bb_entrepreneur_merged$pct_broadband_MS, na.rm = T)),]$county

## [1] "Borden County" "Kenedy County"

# According to M-Lab BB data
tx_bb_entrepreneur_merged[which(tx_bb_entrepreneur_merged$frac_BB.dec == min(tx_bb_entrepreneur_merged$frac_BB.dec, na.rm = T)),]$county

##  [1] "Briscoe County"       "Castro County"        "Cochran County"      
##  [4] "Crane County"         "Culberson County"     "Delta County"        
##  [7] "Donley County"        "Hartley County"       "Jim Hogg County"     
## [10] "Kinney County"        "McMullen County"      "Mitchell County"     
## [13] "Newton County"        "Refugio County"       "Roberts County"      
## [16] "San Augustine County" "Sherman County"       "Sterling County"

# According to ACS BB data
tx_bb_entrepreneur_merged[which(tx_bb_entrepreneur_merged$pctbbfrac_ASU == min(tx_bb_entrepreneur_merged$pctbbfrac_ASU, na.rm = T)),]$county

## [1] "Terrell County"

# Who are the counties with maximum BB?

# According to FCC BB data
tx_bb_entrepreneur_merged[which(tx_bb_entrepreneur_merged$pct_broadband_FCC == max(tx_bb_entrepreneur_merged$pct_broadband_FCC, na.rm = T)),]$county

##  [1] "Aransas County"      "Bastrop County"      "Baylor County"      
##  [4] "Bee County"          "Bexar County"        "Bosque County"      
##  [7] "Brown County"        "Caldwell County"     "Cameron County"     
## [10] "Dallas County"       "Denton County"       "DeWitt County"      
## [13] "Ellis County"        "Erath County"        "Gonzales County"    
## [16] "Grayson County"      "Guadalupe County"    "Hill County"        
## [19] "Hood County"         "Johnson County"      "Karnes County"      
## [22] "Kleberg County"      "Knox County"         "Lampasas County"    
## [25] "Live Oak County"     "McLennan County"     "McMullen County"    
## [28] "Nueces County"       "Palo Pinto County"   "Parker County"      
## [31] "San Patricio County" "Somervell County"    "Tarrant County"     
## [34] "Travis County"       "Willacy County"      "Wilson County"      
## [37] "Wise County"

# According to MS BB data
tx_bb_entrepreneur_merged[which(tx_bb_entrepreneur_merged$pct_broadband_MS == max(tx_bb_entrepreneur_merged$pct_broadband_MS, na.rm = T)),]$county

## [1] "Loving County"

# According to M-Lab BB data
tx_bb_entrepreneur_merged[which(tx_bb_entrepreneur_merged$frac_BB.dec == max(tx_bb_entrepreneur_merged$frac_BB.dec, na.rm = T)),]$county

## [1] "Motley County"    "Stonewall County"

# According to ACS BB data
tx_bb_entrepreneur_merged[which(tx_bb_entrepreneur_merged$pctbbfrac_ASU == max(tx_bb_entrepreneur_merged$pctbbfrac_ASU, na.rm = T)),]$county

## [1] "Fort Bend County"

## Take a look at the frequency distribution of each BB measures

grid.arrange(
    ggplot(tx_bb_entrepreneur_merged, aes(x = pct_broadband_FCC)) + geom_histogram() + theme_minimal() + xlab("FCC Broadband"),
    ggplot(tx_bb_entrepreneur_merged, aes(x = pct_broadband_MS)) + geom_histogram() + theme_minimal() + xlab("MS Broadband"),
    ggplot(tx_bb_entrepreneur_merged, aes(x = pctbbfrac_ASU)) + geom_histogram() + theme_minimal() + xlab("ACS Broadband Subscription"),
    ggplot(tx_bb_entrepreneur_merged, aes(x = frac_BB.dec)) + geom_histogram() + theme_minimal() + xlab("M-Lab Broadband"),
    nrow = 2, ncol = 2, top = "Broadband Measure Distribution"
  )

## Plotting relationships between BB measures

grid.arrange(
  ggplot(tx_bb_entrepreneur_merged, aes(x = pct_broadband_FCC, y = pct_broadband_MS)) + geom_point() + geom_smooth(method = "lm") + theme_minimal() + ylab("Microsoft Broadband") + xlab("FCC Broadband"),
  ggplot(tx_bb_entrepreneur_merged, aes(x = pct_broadband_FCC, y = pctbbfrac_ASU)) + geom_point() + geom_smooth(method = "lm") + ylab("ACS BB Subscription") + xlab("FCC Broadband") + theme_minimal(),
  ggplot(tx_bb_entrepreneur_merged, aes(x = pct_broadband_FCC, y = frac_BB.dec)) + geom_point() + geom_smooth(method = "lm") + theme_minimal() + ylab("M-Lab Broadband") + xlab("FCC Broadband"),
  ggplot(tx_bb_entrepreneur_merged, aes(x = pct_broadband_MS, y = frac_BB.dec)) + geom_point() + geom_smooth(method = "lm") + theme_minimal() + ylab("M-Lab Broadband") + xlab("Microsoft Broadband"),
  ggplot(tx_bb_entrepreneur_merged, aes(x = pct_broadband_MS, y = pctbbfrac_ASU)) + geom_point() + geom_smooth(method = "lm") + ylab("ACS BB Subscription") + xlab("Microsoft Broadband") + theme_minimal(),
  ggplot(tx_bb_entrepreneur_merged, aes(x = pctbbfrac_ASU, y = frac_BB.dec)) + geom_point() + geom_smooth(method = "lm") + theme_minimal() + ylab("M-Lab Broadband") + xlab("ACS BB Subscription"),
  nrow = 2, ncol = 3, top = "Correlations b/w Broadband Measures"
)