# BRIC agricultural yields: Growth factor effects

*In my previous article “BRIC agricultural yields: Trends and theory“, I presented a number of graphs showing both BRIC country yield trends for maize, soybean, and sugarcane crops and posed a number of questions about economies of scale and the effects of various agricultural input factors on crop yield growth rates. In this article, I present a summary of results from statistical analyses of 1990-2014 agricultural data from Brazil, India, and China to answer some of these questions (Russia is omitted due to lack of data). *

The statistical evidence from the analyses can be summarized as follows:

**Efficiencies of scale** — Brazil has experienced economies of scale (in sugarcane), and China has experienced inefficiencies of scale (in maize and soybeans).

**Fertilizer effects** — Brazil and China have experienced reliable, positive yield growth effects from fertilizers, while India has experienced negative yield growth effects from fertilizers.

**Pesticide effects** — Yield growth effects from pesticides are largely dependent on country- and crop-related factors; i.e., results are inconsistent across BRIC country and crop type.

**Fertilizer-pesticide synergies** — Of the BRIC countries, only Brazil has experienced reliable, positive yield growth effects from fertilizer-pesticides synergies.

**Overall yield growth effects** — Economies of (land use) scale, fertilizer inputs, and pesticide inputs explain a large percentage of BRIC crop yield growth: 63% to 66% of maize yield growth; 64% to 71% of soybean yield growth; and, 65% to 71% of sugarcane yield growth.

More details on the results are presented below, so please read on!

#### 1. Statistical model and data overview

Statistical model. Because most readers are not interested in statistical models per se, I will omit technical details of the yield growth factor models used in my analyses and just present a qualitative summary of the statistical results for each country and each crop below. For the record, however, I will mention that I used a seemingly-unrelated regression model (“SUR model”) where there is a regression equation for each *country* (Brazil, India, China) and *crop* (maize, soybean, sugarcane) but–because of cross-correlations in the model error terms–parameter estimation is more efficient when the equations are estimated as a system.

Each regression equation included in the SUR model includes the following variables:

*— country-crop level yield growth rate
*

*— country-crop level land use growth rate*

*— country level fertilizer use growth rate*

*— country level pesticide use growth rate*

The model is formulated in terms of growth rates for a variety of technical reasons, mainly for the purpose of improved parameter (marginal effect) estimation and inference reliability.

Data used in estimating the SUR model was obtained from FAOSTAT and consisted of 1990-2015 annual data from Brazil, India, and China on …

— Agricultural crop yields for each country and crop

— Land area used for each country and crop

— Fertilizers used for each country (measured in aggregate tonnes)

— Pesticides used for each country (measured in aggregate tonnes)

The data is used to construct the growth rate variables discussed above under “Statistical model”.

Marginal effects of growth factors. In the summaries below, “positive” and “negative” refer to the estimated directional effect of the factor on yield growth, and “reliable” refers to whether the estimated effect of the factor is statistically significant at conventional levels (*p* = .01, .05, .10).

#### 1. Maize yield growth factor effects

Estimation and inference statistics from the *maize yield growth equations* of the SUR model can be summarized qualitatively as follows:

The maize yield growth model estimation and inference results are *reliably consistent with* the following hypotheses:

(1.1) Land use economies of scale, fertilizer use, pesticide use, and fertilizer-pesticide synergies account for between 63% and 66% of BIC maize yield growth.

(1.2) China has experienced maize land use dis-economies of scale, suggesting that China’s new land area planted in maize is of lower overall agricultural quality.

(1.3) Brazil has experienced reliable positive marginal effects from fertilizer use on maize crops, while India and China appear to have not.

(1.4) Both Brazil and China have experienced reliable positive marginal effects from pesticide use on maize crops, while India has not.

(1.5) Brazil has experienced reliable, positive synergistic effects from fertilizer-pesticide use on maize crops.

(1.6) India has experienced reliable, positive effects on maize yield growth from factors other than land use scale economies, fertilizers, or pesticides.

From an economic / agricultural development perspective, the important open question about these results is, Why are there significant disparities in the marginal effects of land use, fertilizer use, and pesticide use on maize yield growth across Brazil, India, and China?

#### 2. Soybean yield growth factor effects

Estimation and inference statistics from the *soybean yield growth equations* of the SUR model can be summarized qualitatively as follows:

The soybean yield growth model estimation and inference results are *reliably consistent with* the following hypotheses:

(2.1) Land use economies of scale, fertilizer use, pesticide use, and fertilizer-pesticide synergies account for between 64% and 71% of BIC soybean yield growth.

(2.2) Similar to the result on maize yield growth, China has experienced soybean land use dis-economies of scale, suggesting that China’s new land area planted in soybeans is of lower overall agricultural quality.

(2.3) Brazil and China have experienced positive marginal effects from fertilizer use on soybean crops, while India has had negative marginal effects from fertilizer use on soybean yield growth.

(2.4) Brazil has experienced positive marginal effects from pesticide use on soybean crops while China and India have had negative marginal effects on soybean yields from pesticide use.

(2.5) China and India have experienced positive effects on soybean yield growth from factors other than land use scale economies, fertilizers, or pesticides.

It’s unfortunately not clear why China and India have experienced negative marginal effects on soybean yield growth from the modeled factors. To determine why this might be the case, substantially more data would be needed on other factors influencing yield growth; perhaps including detailed farm level data.

#### 3. Sugarcane yield growth factor effects

Estimation and inference statistics from the *sugarcane yield growth equations* of the SUR model can be summarized qualitatively as follows:

The sugarcane yield growth model estimation and inference results are *reliably consistent with* the following hypotheses:

(3.1) Land use economies of scale, fertilizer use, pesticide use, and fertilizer-pesticide synergies account for between 65% and 72% of BIC sugarcane yield growth.

(3.2) Brazil has experienced sugarcane land use economies of scale, perhaps due to Brazil’s land and climate being well-adapted to sugarcane growth in combination with large areas of land available for cultivation.

(3.3) China has experienced negative marginal effects, while India has had positive marginal effects, from fertilizer use on sugarcane crops.

(3.4) Brazil and India have experienced positive marginal effects, while China has had negative marginal effects, from pesticide use on sugarcane crops.

(3.4) India has experienced positive marginal effects on sugarcane crops from fertilizer-pesticide synergies.

(3.5) China has experienced positive effects on sugarcane yield growth from factors other than land use scale economies, fertilizers, or pesticides.

Again, it’s unfortunately not clear why the inconsistent-across-country and negative marginal effect results exist. It is perhaps the case that agronomy experts would have good, clear theory and hypotheses explaining such results. But it might also be the case that more data–particularly at the farm level–would provide clear insights into the negative and inconsistent results.

#### 4. Conclusions

What might we learn from the statistical analyses presented above? Noting that–as is the case with all statistical results–the results are only *consistent with* the above hypotheses described in (1.1) through (3.5), we do at least learn what empirical regularities exist in three major agricultural producers and three major commercial crops over the last 45 years. Perhaps most importantly, from an economic development perspective the results suggest there remain significant differences in agricultural practices across Brazil, India, and China. If nothing else, this in turn suggests the need to further refine the agricultural yield models underlying this article and find better data sources that will allow more sophisticated, clearer analyses of the causes of crop yield in Brazil, India, and China. (Perhaps more on these things in future articles … .)

**Note for the interested reader**: Anyone who would like to see the actual statistical analyses and results underlying this article, please post a comment below or send an e-mail to me. I will be pleased to send you a copy of the results in the form of a presentation my partner, Rafael Palazzi, and I developed for the 10th Brazil AgroChemShow in São Paulo, Brazil on 17 August 2017.

MMc

São Paulo

**Caveats.** Please note: (i) views presented above are my own and do not reflect those of others; (ii) like anyone, I’m not infallible and am responsible for any errors; (iii) I greatly appreciate being informed of any significant errors in facts, logic, or inferences and am happy to give credit to anyone doing so; (iv) the above article is subject to revision and correction; and, (v) the article cannot be construed as investment or financial advice and is intended merely for educational purposes. MMc