1 / 38

Consolidation in U.S. Agriculture

Consolidation in U.S. Agriculture. James M. MacDonald October, 2019. Contact: jmmacdonald240@gmail.com. The Goals of this Project:. Extend some earlier reports with the 2017 census of agriculture. ERS, August 2013. OECD, July 2016. ERS, March 2018.

bronwen
Download Presentation

Consolidation in U.S. Agriculture

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Consolidation in U.S. Agriculture James M. MacDonald October, 2019 Contact: jmmacdonald240@gmail.com

  2. The Goals of this Project: Extend some earlier reports with the 2017 census of agriculture. ERS, August 2013 OECD, July 2016 ERS, March 2018 2. Put some more statistical evidence behind an assertion in those reports: that consolidation, particularly in crops, has been “persistent, ubiquitous, and large”. 3. Explain why that assertion matters.

  3. This Paper Focuses on Farm and Enterprise Size • Measured by acreage and animals • Whole farm, and commodity enterprises • Production has shifted to larger operations over time • How do we measure that, and how big is the change? • Of course, there’s more to consolidation • Commodity mix, on-farm vs off-farm tasks, ownership and management structure, vertical linkages • And, you can use $ measures for size

  4. ERS Had Significant Earlier Work on Farm Consolidation • Late 1970’s: Carter Administration interest • Spilled over into early 1980’s reports • Lots of capable people, but very limited data • Public census aggregates • Lots of information on technology & organization • Much better data now, and a more complicated story

  5. A Pre-1980 Story, Still Influential... • Kislev and Peterson • “Prices, Technology, and Farm Size”(JPE, 1982) • Focus on growth in average acreage per farm (mean size) • 1930-1970 • A very stripped down model • Relative prices (↑ wl/rk) drive growth in farm size • Rising non-farm wages drive higher opportunity cost of farm labor • ↑ wl/rk drives substitution of capital and land for farm labor, leads to more land per farmer. • Argued that factor price trends, coupled with estimates of elasticity of substitution, accounted for all growth in farm size.

  6. Kislev and Peterson were committed to this model... • Factor price, coupled with estimates of elasticity of substitution, accounted for all growth in farm size. • Argued for no scale economies • Technological change was labor-saving, but all induced • That is, relative factor prices are all that mattered • They noted that growth in mean acreage stopped in 1970-76, as did growth in the factor price ratio • Post-1980: opportunity cost of labor no longer rising, and mean farm size no longer rising, but the focus on mean size focus misses lots of consolidation.

  7. The Simple Mean Trend Shows No Consolidation from 1974-2017

  8. A Starting Point for this Project • Cropland and livestock consolidation has been large, and often dramatic, since the 1980’s • As acreage and animals shift to much larger farms • But simple measures don’t show it • That’s interesting in itself, but also creates a reporting problem for USDA. Our basic measures conflict with farmers’ experience. • This project emphasizes data and measurement • Some reimagining of what USDA should routinely report

  9. The Data • Census of Agriculture • 1982 (1987)-2017 (7-8 censuses) • Farm-level records • So, confidential; Needs access and NASS disclosure clearance • Focus is on all farms with a specified commodity, or with cropland/pasture and rangeland/farmland. • Farms are not linked—any panel features are at an aggregated level • I also use ARMS in the larger project • Agricultural Resource Management Survey • Large annual USDA farm survey, run by ERS and NASS

  10. Start with Cropland Consolidation Farms with at least 2,000 acres of cropland had 15% of U.S. cropland in 1987, and 41% in 2017. Wait: what does “with” mean? Mid size farms (100-999 acres) had 57% of cropland acres in 1987, and 33% in 2017. So, about 100 million acres of U.S. cropland shifted from midsize to large farms over 1987-2017 Note: this is not in census public reports

  11. Cropland is 44% of all Farmland, While Permanent Pasture and Rangeland is 45%. And there’s no consolidation in pasture/range Woodland is 8%. The other 3% is roads, ponds, homes, buildings, livestock facilities, farmstead, and wasteland

  12. The Shift in All Farmland is Far More Muted Than That of Cropland

  13. I Can Generate a Lot of Area Charts, but These Soon Become Overwhelming What I Need to Work With Are Measures of Average Farm or Enterprise Size, But These Measures Can Be Misleading

  14. The Farm Size Distribution is Really Skewed, and This Creates Analytical Problems. Not unlike other industries... Very few farms, and very little land, are near the mean. This has something to do with how we define farms. Mean values are not very informative

  15. I Use the Midpoint Farm Size in this Analysis • Midpoint: a median (not the median), where half the land is on smaller operations and half on larger. • It’s the median of the distribution of acres by farm size • In this example, 1,445 acres • Informative measure for skewed distributions • Use in IO & Labor goes back to Florence’s Logic of Industrial Organization (1933) • Scherer’s Industrial Market Structure and Economic Performance (1970, 1980) references many uses • But they relied on published census aggregates, so the measures were imprecise. Used as additional r.h.s variable in profit-concentration studies. • Today, we can use farm-level census records • Allows precision, and temporal comparisons

  16. A Visual Comparison of 3 Measures of Average Farm Size Note: the “hectare-weighted” median is the midpoint. It’s the median of the distribution of hectares (or acres, or animals) by farm size.

  17. Mean Farm Size (Cropland) Shows Little Consolidation, While the Midpoint Shows a Lot The midpoint, 589 acres in 1982, rises steadily, hitting 1,445 acres by 2017. Of course, this accords with the cropland area chart. Meanwhile, the mean shows no trend between 1987 and 2007, and rises by 10% over 2007-17. Source: Census of agriculture farm-level records.

  18. Why Does the Mean Show So Little Growth? Note: Cropland acres operated is cropland owned and cropland rented in, minus cropland that is rented out. Source: National Agricultural Statistics Service, census of agriculture. In the aggregate, an 11 percent decline in cropland and a 20 percent decline in farms, leading to a modest increase in average acreage. Where’s the growth in farm numbers? Less than 50 acres, and 2,000 or more. A dramatic decline in middle (50-999 acres). Land is shifting to larger farms, while we get better at counting small farms.

  19. So, the Post-1980 Consolidation Story is Complicated • Cropland, not Pasture and Range • A hollowing out of the middle, so simple averages miss • Livestock (we’ll see), except cow-calf, • And some is revolutionary

  20. Let’s Take a Closer Look at Crops Here, midpoints for harvested acreage, for 5 major field crops, 1987-2017

  21. Field Crop Consolidation is: • Ubiquitous: all 5 crops • Persistent: midpoint increases between every census for each crop (except cotton, 2007-2012) • Large: 1987-2017 change ranges from 166% (wheat) to 243% (corn). • And, this pattern continues in 2012-17

  22. I Expanded the Analysis to 55 Crops • 15 field crops, 20 vegetable and melon crops, 20 fruit, tree nut and berry crops • Harvested acreage, census years from 1987-2017 • 385 crop/year observations • Midpoints, and total harvested acreage, by crop • In the Three Decades report, we reported summary estimates to argue that consolidation was ubiquitous across crops, persistent over time, and large. • Here, I quantify and test those assertions

  23. Start with Long-term Changes: A Scatterplot of 2017 vs. 1987 Midpoints We calculated midpoint acreages (harvested acres) for 55 crops: Here, we plot 2017 vs. 1987 midpoints, to visualize ubiquity and magnitude of consolidation. So each dot plots 2017 vs. 1987, for one of 55 crops. (The outlier is lettuce)

  24. A Summary: Add a 45 degree line and a regression line With no long-term change, points would be scattered around the 45 degree line. Two points fall below the line, indicating decline in 1987-2017. 53 points are above, and the average increase is 148.7%.

  25. Deleting Lettuce Give a Clearer View Two crops (lemons and plums) are clearly below the 45 degree line. Lettuce tilted the mean estimate a bit; now 126.9% growth.

  26. Now Let’s Look at States Midpoints based on cropland, and from 1982 Five states show declines. Largest % increases in Corn Belt, Delta, Plains

  27. A Closer Look at Crop Estimates • I want to model persistence of consolidation over time, and ubiquity across commodities. • So, I’m going to focus on growth rates of midpoints as my measure of consolidation. • I don’t really care about levels—the corn midpoint (685 acres in 2017) versus the avocado midpoint (54 acres). • So, focus on persistence over time in growth rates, and ubiquity across crops in growth rates

  28. A Closer Look at Crop Estimates • Dependent variable: log difference between crop i midpoint in year t and mean crop i midpoint (that is, fixed crop effects) • Ln Mit-Ln Mi, where is Mit is the midpoint for crop i, year t, and Mi is the mean midpoint for crop i. • Reminder: we observe crop midpoints every 5 years from 1987 through 2017 • Relate to year/time effects, crop category dummy variables, and growth in harvested acreage for the crop.

  29. I Ran Fixed Effects Regressions to Test for Persistence and Ubiquity Dependent variable: log difference between crop midpoint in year t and mean crop midpoint (that is, fixed crop effects) Year effects, vs. time trend (t) to test persistence Short-term shocks to total crop planted acreage (acreage growth) matter. A F test for (3) vs. (2): F5,377 =2.08. The critical F value at 95% significance is 2.21; cannot reject constant growth rate of midpoints over 1987-2017. I find this, and the relatively high R2, to be remarkable. A time trend and total acreage growth account for a high share of the variation in 5-year midpoint growth rates across 55 crops.

  30. Here, we test for differences across crop groups Dependent variable: log difference between crop midpoint in year t and mean crop midpoint (that is, fixed crop effects) D2 is a dummy for vegetable & melon crops D3 is a dummy for fruit, tree nut, berry crops F test, for (4) vs. (3): F2,380=3.97 Critical F at 95% significance=3.00 F test, for (5) vs (4): F2,378=3.56 Critical F at 95% significance=3.00 However, estimated annual midpoint growth rates are quite close.

  31. Let’s Look at Livestock Consolidation, 1987-2017 Source: USDA/ERS calculations, from census of agriculture records Measurement: inventories versus removals... What do we see: variance across commodities * Radical change in hogs, egg layers, milk cows * Relatively little change in cow-calf * Ongoing change in broilers, fed cattle

  32. ERS Has Lots of Work on Livestock Consolidation Focus on scale, specialization, vertical relations, multi-farm operations. Primary focus on broilers, dairy, & hogs

  33. What Happens When We Add Livestock to the Analysis of Midpoint Growth? Now we have 7 years times 62 commodities, or 434 observations. D4 is a dummy for a livestock sector. “Sector growth” is now based on inventory, removals, or acreage, expressed as log difference from commodity mean. R2 falls—livestock adds variance. We still cannot reject constant temporal growth. F test for (3) vs (2): F5,434=2.0. Midpoint growth rates don’t vary widely across commodity types, based on estimates for t.

  34. The Patterns Are Informative of Causes • Do commodity programs drive farm consolidation? • Well, commodity programs focus on field crops. • Yet consolidation has occurred in livestock and in specialty crops, at similar rates • Commodity support changes over time • Consolidation is remarkably persistent over time

  35. This is a Dialogue on Roberts and Key • Commodity Payments, Farm Business Survival, and Farm Size Growth • An ERS project—a 2008 report, several journal articles • Measured midpoints, for cropland and farmland, across counties and zipcodes. • Tracked midpoint growth, 1987-2002 • Tied to differences in per-acre government payments • Found highest growth where payments highest • Attributed all of consolidation to commodity programs

  36. Some clear patterns to 1987-2002 consolidation • Sorted farmland and cropland into 6 classes: • No payments in county/zipcode • Quintiles (5) sorted by payments per cropland acres • For farmland, midpoint growth much higher in highest quintile, • Negative in lower & no-payment classes • For cropland, midpoint growth negative in no-payment class, low in 2 lower quintiles • Higher in highest three quintiles, but little difference among those three. • Attribution of commodity payments come from difference in midpoint growth between highest and lowest/no pay quintiles

  37. Pairing Kislev & Peterson w/ Roberts & Key • K&L: • Capital-labor substitution and labor-saving tech change allow farms to manage more acres and more animals • Or, “Farmers farm more acres because they can” • R&K • Land consolidation does not occur in places with low or no government payments per acre • Mostly concentrated in top 3 quintiles of cropland • That’s about 240 million acres • A lot of ground, with more than payments in common • Flat, contiguous, high-quality land: suitable for labor-saving tech change?

  38. Conclusions and Next Steps • Consolidation in cropland and livestock continues • The exception is the pasture and range/cow-calf complex • I think labor-saving technological change has played an important role • But I don’t think it’s all induced • And I think that effects are conditional on land attributes • The big future question: how does precision agriculture play into this? • I aim to keep a now-dispersed team in touch with each other and ERS data

More Related