Farmers in the Southeast US

UEP 235 Advanced Geospatial Modeling

Melissa Gordon May 2020

Image by  Bishnu Sarangi  from  Pixabay  

What are some of the conditions of black farmland access in the southeastern US?

Black farmland ownership in the US peaked in 1910 at 15 million acres, and today has declined to less than 3 million acres owned (Rainge, 5). Despite land ownership decline, black producers are still farming: of the 2,042,220 farms in the US today, 35,470 have Black producers (1,973,006 have white producers) (USDA, 2017). Many farmers nationwide rent the land they farm, including many Socially Disadvantaged Farmers and Ranchers, a designation which includes farmers of color, beginning farmers, female farmers, and veterans. Renting agricultural land can be expensive and lead to instability as land is not secure.

"Tenants have a less secure position, in that they have to negotiate a lease on a regular basis, comply with the owners’ demands, typically have little say about the future of the land, and do not build wealth long-term from the land." (Horst and Marion, 2018)

This research set out to investigate land access for Black Farmers in the southeast, “Southern Belt” or “Black Belt,” of the United States. The name Black Belt is the official agricultural region designation for the rich, black soil in the middle of Alabama. This research compares Black and White farmers’ farmland conditions in the southeast US.

First, let's explore the farmland in the southeastern US.

The darkest green areas are where there is prime farmland.

Here you can explore what crops are grown in the region. Open the legend or click on a county to see what its main crop is.

Where are black farmers in the southeast?

Data Exploration & Visualization

All static maps are in the USA Lambert Conformal Conic Projection, and data come from United States Department of Agriculture's (USDA) National Agricultural Statistics Service (NASS), from the 2017 Census of Agriculture. Map data, such as the county and state borders, are from ESRI.

This is the % of producers who are Black.

And the percent of producers who are white. Notice the different scales: % of producers who are black ends at 50%; % who are white starts at 43%. There is a bit of an inverse relationship--areas with higher % black are lower % white.

Here are the number of acres with a black producer in each county.

And here are the number of acres with a white producer. Again, notice the scales. # of acres with a Black producer tops out at 62,000, whereas # of acres with a white producer is over 450,000 in some counties.

After looking at the racial demographic distribution of farmers in the region, we can look at some economic aspects.

What are cropland rents in these areas?

Rent ($/acre) Projection: USA Lambert Conformal Conic, Source: USDA NASS 2017 Census of Agriculture and ESRI.

What are net incomes in this region?

Net Incomes per Operation

"The discriminatory implementation of farm policy over previous decades has meant lasting negative economic implications for black farmers, particularly those living in the rural South. The effects are still seen today, as evidenced by the fact that the average farm income for all full-time and part-time farms in 2017 was $10,276 for white-operated farms, while the average income for all black-operated farms was just $795." (Castro and Willingham, 2019)

Here we can examine the relationship between acres with black producers and land rent rates. Use the legend in the bottom left to guide you. Red is where both acres with black producers and rent rates are high. Purple is where there are relatively high rates of acres with a black operator and low rent rates.

Here we see % of producers who are black, with red dots in an overlay, indicating small to large net incomes. We can see a mix of high incomes coinciding with high rates of black producers, but some of the areas with high rates of black farmers do not have high incomes, like in the middle of MS and AL.

We can also compare net incomes to % of farmers who are white. Remember that % of producers who are white is over 60% in almost all regions.

This map shows the relative prominence of various tenure arrangements: red shows where most acres in a county have a full owner; yellow is part owner; blue is tenant. The transparency levels show how strong the predominance is. We can see that not many places have predominantly tenant arrangements.

Analysis

Tabulate area, Local Moran's I, Ordinary Least Squares Regression

I started with a prime farmland raster from the Natural Resource Conservation Service.

Then I tabulated area to get the percentages of prime farmland in each county. We can see lots of prime farmland in Louisiana and along the Mississippi River. There are low rates of prime farmland in along the Appalachian mountains and in many coastal areas.

I used statistical tools in ArcMap to examine the situation and look for significant high-occurrence and low-occurrence clusters of each variable using Local Moran’s I. This tool compares a county to it's neighbors, to see if it is significantly different. I could then explore other summary statistics in those clustered areas.

The red clusters show where there is statistically significant clustering of Black producers.

This is where there is statistically significant clustering of white producers. In some instances, the high clusters are inverse from the high clusters of black producers.

These are the clusters of prime or high-quality farmland, as a percentage of total area of the county.

Though not what I expected, the significantly clusters of high full ownership are not in the places with the most prime farmland, like along the MS AK border, and high full ownership areas include areas with high rates of black producers, like in southern MS.

These are the clusters of high and low amounts of net income.

Finally, clusters of high and low rent rates. High rent clusters coincide in some areas with clusters of prime farmland, like along the MS-AK border.

I explored what was going on in those clusters by selecting some significant clusters areas by attribute. I then explored various average/mean characteristics of those clusters. For example, this exploration indicated that in high-high clusters of % black producers, the average % of Prime Farmland is 64%, while for white producers it is 33%. Comparing clusters of high rent rates to low rent rates indicated that in on higher rent land, more operations have gains, fewer have losses, and net incomes are higher, than on lower rent land.

This did not prove to be a fruitful activity, as there was great variation within each cluster, as indicated by very large standard deviations. While some of the data showed comparisons that I may not have expected, I cannot really place weight on these findings. Further research could perhaps compare clusters within the same state, for a more equivalent comparison.

Selected means for clusters by race and rent rates.

Finally, I ran an Ordinary Least Squares regression to see if there is any statistically significant correlation between the variables.

In this regression, I looked into the impact on net income per operation of median operation size in acres, rent rate in $/acre, % of acres with a tenant, % of operations with a black producer, and % of prime farmland in the county. The R-Squared shows that 54% of the variation in the outcome variable, net income, is explained by the model. All the variables are statistically significant except % of operations with a black producer, and they all have positive coefficients, though the intercept is negative, reflecting that some net incomes are negative. This is to say that as median operation size, rent rate, % tenants, and % prime farmland increase, net income also increases.

This regression looks at similar variables impact on net gain ($) per operation, of operations with gains. Instead of % of operations with a black producer, I used % of acres with a black producer. The R-Squared value is .40, and again, the variables are statistically significant, except % of acres with a black producer.

As a final regression example, here is a regression where the outcome is net income per operation, and the inputs are % prime farmland of all the land in the county, median acres per operation, rent, and % of acres with a white producer. The R-squared value is .51. This model indicates that as % prime farmland increases, net income decreases, but that relationship is not even close to statistically significant. The other variables, include % of acres with a white producer are statistically significant, indicating that as they increase, so does net income.

Peanuts Photo by  shattha pilabut  from  Pexels 

The relationships I was expecting to see were not evident from my analysis. There were not significant correlations between race, rent rates, and access to prime farmland. Relationships that one would expect seemed present, such as that larger farms have higher incomes. Rent is often higher for better quality farm land, so it also makes sense that incomes would be higher here. However many areas with high gains on some farms also had large losses on others.

Peaches Photo by  Sugar Bee  on  Unsplash 

There were numerous limitations to the data, so these findings are not conclusive. Data limitations include missing data for some counties. For example, rent rates were not available for some counties, so I imputed by filling in an average. Also, I only explored crop land, but many farmers, including farmers of color, raise livestock, so examining pasture rents or disaggregating livestock from crop cultivation could enhance future research. Another possible variable to include in the future is gross income which indicates more about the size of the operation than net income.

Despite inconclusive findings, I remain certain that there is a lot of work to be done in this area. Organizations like the  Black Belt Justice Center  and the  Federation of Southern Cooperatives / Land Assistance Fund exist explicitly to further the interests of farmers of color in the Black Belt region of the US. Hopefully further research can continue to support work like theirs.

Image from Black Belt Justice Center website

Image from the Federation of Southern Cooperatives Website

References

Thank you to Sumeeta and Carolyn!

Rent ($/acre) Projection: USA Lambert Conformal Conic, Source: USDA NASS 2017 Census of Agriculture and ESRI.

Net Incomes per Operation

Selected means for clusters by race and rent rates.

Peanuts Photo by  shattha pilabut  from  Pexels 

Peaches Photo by  Sugar Bee  on  Unsplash 

Image from Black Belt Justice Center website

Image from the Federation of Southern Cooperatives Website