| |

Spatial Patterns and Demographic Influences on Ham Radio Operators in US Counties

Introduction

This is an exploration of where amateur radio operators, or “ham radio operators,” are located, and who they are, across the United States. Using data from the Federal Communications Commission (FCC), I’ve mapped out where these operators live and tried to understand what influences their distribution across various counties. The tools for this task included the PostGIS Tiger geocoder for mapping the data, and ArcGIS Pro and R for analyzing the data.

One interesting finding is that ham radio operators tend to cluster in certain areas, possibly due to cultural or economic factors that make these regions more conducive to the hobby. This clustering was detected through ArcGIS Pro, which allowed for the visualization of the geographic data. This project offers insights into not only where ham radio operators are most common but also what demographic factors might influence these patterns.

The results of this work have been incorporated into an ArcGIS Online Dashboard.

Methodology

I aimed to explore the geographic distribution and influencing factors of amateur radio operators across U.S. counties. The data for this analysis was sourced from the FCC’s Universal Licensing System (ULS), which maintains records of all licensed ham radio operators in the United States. Data were also obtained from the US Census Bureau’s American Community Survey.

Data Collection

The first step involved extracting operator records from the FCC ULS database. Each record was then geocoded using the PostGIS Tiger geocoder. This process converted address data into geographic coordinates, ensuring that each operator was accurately mapped to their respective county.

Data Analysis Tools:
The analysis was conducted using two main software tools:

  1. ArcGIS Pro: This geographic information system (GIS) was utilized to handle spatial data manipulation and visualization, as well as geostatistics.
  2. R: Most data manipulation and some statistical analysis was performed in R, which offered. Packages such as sf for handling spatial data, dplyr for data manipulation, and spdep for spatial dependence were instrumental.

Statistical Methods

The project employed several statistical techniques to assess the distribution of ham radio operators:

  • Spatial Empirical Bayes Smoothing: This technique helps stabilize rates in counties with small populations. By considering data from neighboring counties, it provides more reliable estimates, reducing the effect of random fluctuations in sparsely populated areas.
  • Negative Binomial Regression: This method addresses overdispersion in data, where the variance is much larger than the mean. It provides a better fit for count data, like the number of ham radio operators.
    • Lagged Variables: Spatially lagged variables were calculated to incorporate the influence of neighboring counties into the regression model.

Outcome Measures

The main outcomes measured were the rate of ham radio operators per county, adjusted for socio-demographic factors such as median household income, total population, and educational attainment. These factors were included as covariates in the regression models to isolate the effect of geographical location on operator rates. In addition to these demographic variables, I obtained counts by religious affiliation from the US religion census to aid in the analysis.

Geocoding Results

In this project, the geocoding process was a crucial step to accurately map the locations of ham radio operators. Utilizing the PostGIS Tiger geocoder, Isuccessfully geocoded 1,589,052 out of a total of 1,612,849 FCC Universal Licensing System (ULS) records, achieving a geocoding success rate of 98.53%. Despite this high success rate, there were 5,486 records (0.35%) that could not be geocoded at the street address level using the Tiger geocoder. These records were excluded from the analysis.

The average geocoding rating for records processed through the Tiger geocoder was 16.35 on a scale from 0 to 100, where lower ratings indicate higher geocoding accuracy. This rating reflects a high level of precision in the geocoding efforts, ensuring that the majority of operators were mapped accurately to their respective locations. The combination of the Tiger geocoder’s high accuracy and the supplementary use of Nominatum for less precise locations allowed me to create a comprehensive and reliable dataset, available here.

Statistical Analysis

To further understand the distribution of ham radio operators across U.S. counties, I employed a negative binomial regression model. This statistical method is particularly suited for count data that exhibits overdispersion, where the variance is significantly greater than the mean.

Model Overview

The negative binomial model was used to assess the influence of various socio-demographic and geographic factors on the number of active ham radio operators in each county. The predictors included:

  • Total Population: The total number of residents in the county.
  • Spatial Lag: The influence of ham radio operator counts in neighboring counties.
  • Median Household Income: The median income of households in the county.
  • Percent White Population (white_perc): The percentage of the population that is white.
  • Percent Black Population (black_perc): The percentage of the population that is Black.
  • Houses per 100 Population (housing_perc): The number of houses per 100 residents.
  • Percent with Bachelor’s or Above (bach_perc): The percentage of residents with at least a bachelor’s degree.
  • Median Age: The median age of residents in the county.
  • Religious Adherence Rates: Including LDS adherents, Mainline Protestant adherents, Evangelical adherents, and Catholic adherents per 1,000 residents.

Key Findings

The results of the negative binomial regression model revealed several significant predictors of the number of active ham radio operators per county. The residual deviance (2045.7) of the model is substantially lower than the null deviance (5576.3), indicating that the model explains a significant portion of the variance in the data. The dispersion parameter (theta) of the model is 1.48, suggesting a moderate amount of over-dispersion which indicates that the negative binomial model is an appropriate choice.

Here are the key findings:

  1. Total Population: As expected, counties with larger populations tend to have more active ham radio operators. For every additional person, there’s a small but significant increase in the expected number of operators. This relationship is highly significant, meaning larger populations strongly predict more operators.
  2. Spatial Lag: There is a significant positive effect of the number of operators in neighboring counties, indicating spatial clustering. This suggests that ham radio operators tend to be near other operators, possibly due to shared interests or community support.
  3. Median Household Income: Higher median household income is associated with a slight decrease in the number of active operators. This might be because areas with higher incomes have different leisure activities or hobbies.
  4. Percent White Population (white_perc): Higher percentages of white residents are associated with fewer active ham radio operators. This finding might reflect varying interests or accessibility of the hobby across different racial groups.
  5. Percent Black Population (black_perc): Although included in the model, this variable was not found to be a significant predictor, indicating that the percentage of Black residents does not significantly impact the number of ham radio operators.
  6. Houses per 1,000 Population (housing_perc): Higher housing density is associated with fewer active operators. This could be due to limited space or interference in densely populated areas making ham radio operations less practical.
  7. Percent with Bachelor’s or Above (bach_perc): Higher educational attainment is positively associated with the number of active ham radio operators. This suggests that individuals with higher education levels might be more inclined toward technical hobbies like ham radio.
  8. Median Age: Older median age slightly increases the number of active operators. This aligns with the notion that ham radio is a popular hobby among older adults.
  9. Religious Adherence Rates: Higher rates of LDS, Mainline Protestant, Evangelical, and Catholic adherents are associated with fewer active operators. This might reflect cultural or community differences influencing hobby choices.

Visualizing the Results

To make these findings more accessible, I created a plot of the marginal effects of each predictor on the expected count of active ham radio operators. The plot includes 95% confidence intervals for each estimate, allowing us to see the uncertainty around these effects. The significance levels of the predictors are indicated by stars, with *** denoting p < 0.001, ** denoting p < 0.01, and * denoting p < 0.05.

This visualization helps convey the relative impact of each factor on the number of active ham radio operators, making the statistical results more interpretable for a general audience. The x-axis shows the change in the number of ham operators per county with a 1-unit increase in the variable shown on the y-axis. So, with a 1% increase in population with at least a Bachelor’s degree, there is around a 0.11 count increase in ham operators in that county, holding all other variables constant.

The vertical bars show the 95% confidence intervals, meaning that we can be 95% sure that the true effect of the variable on the ham radio count can be found within the range of those bars.

The negative binomial regression model provided valuable insights into the demographic and socio-economic factors influencing the distribution of ham radio operators across U.S. counties. By accounting for overdispersion in the data and incorporating spatial dependencies, the model offers a robust analysis of the patterns and predictors of ham radio operator distribution.

Overall, these results suggest that population size, educational attainment, and regional clustering are key drivers of where ham radio operators are found, with income and housing density also playing significant roles. Understanding these factors can help in promoting the hobby and supporting ham radio communities across different regions.

Geostatistical Analyses

Now that the regression modeling is complete, we turn to geostatistics, to determine where ham radio operators are clustering. A preliminary view of the rates of all active ham radio operators per 10,000 residents per county shows that there appear to be some clustering going on. The Northwestern United States is an area with high rates, as well as parts of the South and Texas.

To determine if there is significant clustering going on, we perform a hotspot analysis in ArcGIS Pro. The results of the analysis indicate that there are indeed hot and cold spots of active ham radio operators in the US. As expected, the Northwestern US is a hotspot, along with several other areas scattered across the continental US.

But which license class is clustering where? We can perform the same analysis on the percentage of each license class out of the total number of active operators to determine this. We will go in order of highest license to lowest, beginning with Extra class.

Extra Class Clusters

The results for Extra class operators are somewhat surprising. The hotspot of operators in the Northwest is a cold spot for Extra class hams. The cold spot indicates a lower percentage of Extra class licensees than expected if the distribution were random. Conversely, the Northeastern US, South Texas, much of Alabama, and Southern Florida are significant hotspots for Extra class licensees.

Advanced Class Clusters

There is a significant hotspot of Advanced class licensees around the Great Plains area, including South Dakota, Nebraska, Iowa, and parts of Minnesota, stretching East into Illinois. Interestingly, this area is also a cold spot for the overall rates of operators, meaning that while there are fewer hams per 10,000 population than expected, a higher proportion of these operators hold Advanced class licenses.

General Class Clusters

The hotspots of General class licensees are more dispersed than those of the higher license holders, with notable clusters in the North-Central US, starting in Nebraska and extending northward. Conversely, Utah, much of California, and West Virginia are cold spots for General class licensees.

Technician Class Clusters

As expected, there are numerous large hotspots of Technician class licensees. The cluster of all active licensees in the Northwest US has a high proportion of Technician class license holders. Nearly all of California is a hotspot for Technician licenses. Surprisingly, most of Florida is a cold spot for Technician class licensees, whereas it is a hotspot for Extra and Advanced level licensees.

Novice Class Clusters

The pattern of Novice class licensees is intriguing. While I suspected it might mirror that of the Advanced class licensees, given both are grandfathered license classes, the results show otherwise. Advanced class licensees have one large cluster over the Great Plains, whereas Novice class licensees are broken into many smaller clusters, with the largest centered over New York.

Conclusion

This look into the ham radio operator distribution across the United States revealpatterns and insights. By using data from the Federal Communications Commission (FCC) and tools like ArcGIS Pro and R, we mapped out where ham radio operators live and analyzed factors influencing their distribution.

I was happy that 98.53% of FCC Universal Licensing System records were geocoded. The negative binomial regression model highlighted key predictors of operator counts, including population size, educational attainment, and regional clustering, while factors like income and housing density also played significant roles.

Geostatistical analyses identified significant hot and cold spots of operator clustering. The Northwestern US emerged as a major hotspot of all active hams, with varying patterns for different license classes. For instance, Extra class operators were more concentrated in the Northeastern US and parts of the South, while Technician class operators formed large hotspots in California and the Northwest.

These findings highlight the importance of understanding local demographics and cultural factors in supporting and promoting the ham radio hobby.

Similar Posts

Leave a Reply