Agricultural Productivity Growth and Environmental Externalities

“Can Soil Conservation Practices Increase

Upland Farm Productivity?”Agnes C. Rola, Asa Jose U.

Sajise, Dieldre S. Harder, Joe Marvin Alpuerto

(Paper presented during the Asian Society of Ag Economists

Meeting, Makati,Philippines, 28-30 Aug. 2008)

Introduction

Present results from a larger study Southeast Asian Regional Center for

Graduate Study and Research in Agriculture (SEARCA) Productivity Growth for Philippine Agriculture Project

Project component on Agricultural Productivity and Environmental Externality

Upland-Lowland Agriculture (Aggie to Aggie externality)

In-Situ Effects of Soil Erosion on Upland Farm Productivity

Technology Adoption & Soil Erosion

Soil Erosion and Irrigation Systems

Introduction: Detail of Larger Study

Ex-Situ Effects of Soil Erosion (thru Irrigation Systems) on Lowland Farm Productivity

Review of Literature on Upland Corn Cultivation and Soil Erosion

Empirical Case Study: Panel Data Analysis

Review of Literature on Irrigation and Lowland Farm Productivity

“Empirical” Case Study: Productivity Change Analysis

Focus of Presentation

Rationale for Scope of the Study

Ecosystem: Upland Agricultural Systems Commodity: Corn Natural Resource Base: Soil Environmental Factor: Soil Loss/ Erosion Why?

Soil erosion has been identified as a leading problem in upland resource management

Upland areas will be the potential bread basket because of dwindling lowland areas

Huge amount of R and D investments on soil erosion control (agroforestry, SALT, etc.)

General Methodology

No direct measure of soil erosion Therefore, look at the indirect

relationship Adoption practices -> Soil Loss -> Decline/

Increase in Productivity Review literature on corn production and

soil erosion Case Study: use data and estimate a

panel stochastic frontier on a primal production determination function

Empirical Methodology

Posit the following yield/ production determination function: Corn Yieldit =f(Ωit, θit, dit)

where Ωit is a vector of household/ plot characteristics

for household i at time t Θt is a vector of economic/ institutional /

physical variables for household i at time t dit is an indicator variable d=1 if adopts a soil

conservation practice, assumes value of 0 otherwise for household i at time t

Econometric Specification: Stochastic Frontier Corn Yieldit =f(Ωit, θit, dit) + εit + μit

where εit is the usual error term for household i at

time t μit is the truncated error term that

represents technical inefficiency for household i at time t

Furthermore we specify a time varying technical efficiency term of the form

( ( ))

2( , )

it Tit

iid

i

e

where

N

Econometric Issue

Adopting a conservation measure or not is endogenous A function of some set of variables

A better or alternative way to look at it: Measurement error soil erosion is a latent variable (in the

absence of an indicator) or may not be measured properly

using only adoption of soil conservation technology as a proxy

Roughly: Prob(d=1) = soil erosion + Є

Solution: Two Stage Estimation or Treatment Effects 1st Stage: Discrete Choice

Probit on discrete decision to adopt Get Mill’s ratio for non-adoption and

adoption 2nd Stage: Stochastic Frontier

Revised specification: E[Corn Yield|Ωit,θit,1]= f(Ωit,θit,1) + E[Corn Yield|Ωit,θit,0]=f(Ωit,θit,0) +

)(

)(

z

zua

)(1

)(

z

zu

na

Results of Review of Literature: Some Statistics Corn production (kg./ha) in the

Philippines is lagging behind it’s SEA neighbors (FAO Statistics)

Country 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005

Ave

growth

rate

1995-2005

Cambodia 1,219 1,374 1,243 1,217 1,595 2,735 2,761 2,080 3,747 3,320 3,516 15.46

Indonesia 2,258 2,486 2,614 2,653 2,663 2,765 2,845 3,088 3,241 3,344 3,428 4.30

Malaysia 1,870 1,800 1,846 1,852 2,111 2,407 3,046 3,044 3,000 3,000 3,000 5.22

Myanmar 1,703 1,710 1,908 1,642 1,706 1,724 2,124 2,251 2,477 2,630 2,842 5.65

Philippines 1,521 1,592 1,589 1,624 1,735 1,797 1,820 1,803 1,915 2,142 2,045 3.10

Thailand 3,289 3,448 3,198 3,345 3,552 3,676 3,735 3,729 3,855 3,869 3,760 1.42

Viet Nam 2,114 2,498 2,490 2,481 2,534 2,747 2,963 3,076 3,436 3,462 3,569 5.52

Results of Review of Literature: Some Statistics In terms of agro-

zone Uplands has lower

minimum and maximum production

In general, wide range of production (i.e. more erratic)

Range of maize yield by agro-ecozone (t/ha) Type of maize

material Rainfed

lowlands Upland plains

Rolling-to-

hilly

Local/traditional

Most common - 1.0 – 2.0 0.1 – 2.5

Minimum attained - 0.5 – 1.5 0.1 – 3.0

Maximum attained - 1.0 – 2.0 0.2 – 3.8

Improved OPVs

Most common - 2.0 – 4.0 0.9 – 2.4

Minimum attained - 1.0 – 3.5 1.0 – 2.1

Maximum attained - 2.0 – 4.5 2.0 – 4.5

Hybrids

Most common 4.6– 6.0 3.0 – 5.7 1.6 – 5.0

Minimum attained 3.0– 5.0 1.5 – 5.0 2.0 – 4.3

Maximum attained 5.5– 9.0 4.0 – 7.0 4.0 – 7.0

Summary of Literature Review Yield gap is due to a variety of reasons:

biological constraints, soil and water constraints, socio-economic constraints, ‘erratic and unpredictable’ weather conditions and storms, use of sub-optimal fertilizer input due to capital constraints, soil acidity, and declining soil fertility

Literature acknowledges that corn in the uplands is very soil erosive

However, obviously dampened by adoption of technology (e.g. crop rotation, hedge rows, etc.)

Results of Empirical Study SANREM panel data

from 1994 to 2006 of from 195 households in 1994 to 80 households in 2006; and about 109 plots in 2006 Data series includes corn

input, output, soil conservation practices, demographic characteristics, some data on rainfall patterns, village level data on water quality

Site is in Lantapan Bukidnon

Casual look at trends in production show increasing production for adopters

05000

10000

1995 2000 2005 1995 2000 2005

0 1

Pro

duct

ion p

er

ha.

Year0 - Non-Adopters; 1 - Adopters

Production (per ha.), Adopters vs. Non-Adopters (1994-2006)

Data Problem

One problem is corn plot attrition (around 80%) Various reasons: out migration; shift to other crops

Can test for effects of attrition based on observables Use baseline data Regress yield with inputs, other variables and an

attrition indicator Test shows that attrition indicator is not

significant, implying attrition is not a real problem

Determinants of Technology Adoption Probability of

adoption higher in farms located in higher slopes

Favorable price expectations leads to higher likelihood of adoption

Households relying on family labor are more likely to adopt soil conservation technologies

Variable Coefficient Std. Err. Slope 0.15* 0.06 Watershed -0.05 0.24 Education of Household Head -0.03 0.08 Tenure -0.05 0.07 Age of Household Head -0.05 0.06 Squared Age of Household Head 0.00 0.00 Fertilizer Use -0.17 0.18 Favorable Price Expectation 0.38** 0.23 Above Mean Production -0.62* 0.17 Used Family Labor Only 0.61* 0.20 Year Dummies:

1996 -1.28* 0.24 1998 -1.27* 0.30 1999 -1.11* 0.27 2000 -0.36 0.28 2002 -0.98* 0.31 2006 -0.81* 0.34

Constant** 2.44** 1.43 * - significant at 5%

** - significant at 10%

Determinants of Technology Adoption (continued) Farms with higher

productivity are more likely to be put under a soil conservation regime

Adoption is also more likely during the base year than later years.

Variable Coefficient Std. Err. Slope 0.15* 0.06 Watershed -0.05 0.24 Education of Household Head -0.03 0.08 Tenure -0.05 0.07 Age of Household Head -0.05 0.06 Squared Age of Household Head 0.00 0.00 Fertilizer Use -0.17 0.18 Favorable Price Expectation 0.38** 0.23 Above Mean Production -0.62* 0.17 Used Family Labor Only 0.61* 0.20 Year Dummies:

1996 -1.28* 0.24 1998 -1.27* 0.30 1999 -1.11* 0.27 2000 -0.36 0.28 2002 -0.98* 0.31 2006 -0.81* 0.34

Constant** 2.44** 1.43 * - significant at 5%

** - significant at 10%

In-Situ Impact of Technology Adoption (Soil Erosion) on Upland Farm Productivity All production inputs

have the right signs and directly related to production

Adoption of soil conservation increases production implying that controlling erosion increases productivity among upland farms

Variables Coefficients Std. Err. Production Inputs: Ln (Seed Inputs)

0.12* 0.05

Ln (Urea Applied)

0.09* 0.03

Ln (Manure Applied)

0.05** 0.03

Ln (Complete Fertilizer Applied)

-0.03 0.02

Ln (Man-days of HH Labor)

0.13* 0.03

Ln (Man-days Hired Labor)

0.06* 0.03

Biophysical Variables

Drought Dummy

-1.27* 0.13

Watershed Dummy

-0.38* 0.19

Slope Dummy 0.02 0.05 Adoption Dummy

0.82* 0.29

Interaction Term Between Watershed and Adoption

0.45* 0.20

Interaction Term Between Slope and Adoption

0.02 0.06

Mills Ratio -0.79* 0.11 Constant 6.71* 0.24

In-Situ Impact of Technology Adoption (Soil Erosion) on Upland Farm Productivity Drought dummy for

(1997=8;others=0) drastically reduced production

Farms located in higher watershed also had lower corn yields

However, the interaction term between watershed location and adoption imply that farmers in the upper watershed who use soil conservation techniques will tend to have higher yields Soil conservation can

mitigate locational disadvantages

Variables Coefficients Std. Err. Production Inputs: Ln (Seed Inputs)

0.12* 0.05

Ln (Urea Applied)

0.09* 0.03

Ln (Manure Applied)

0.05** 0.03

Ln (Complete Fertilizer Applied)

-0.03 0.02

Ln (Man-days of HH Labor)

0.13* 0.03

Ln (Man-days Hired Labor)

0.06* 0.03

Biophysical Variables

Drought Dummy

-1.27* 0.13

Watershed Dummy

-0.38* 0.19

Slope Dummy 0.02 0.05 Adoption Dummy

0.82* 0.29

Interaction Term Between Watershed and Adoption

0.45* 0.20

Interaction Term Between Slope and Adoption

0.02 0.06

Mills Ratio -0.79* 0.11 Constant 6.71* 0.24

Other interesting insights

Calculated marginal effects for each variable Marginal effects show that the impact of adoption

on productivity is higher compared to the productivity effects of conventional input usage (singly or combined)

The technical inefficiency term is increasing across time Some farmers are indigenous people (IP), thus, practices

time old traditional conservation and cultivation Why then is there declining efficiency?

Labor opportunities in the lowland More variability in weather

Summary and Policy Implications (so far) Target extension (and R and D) efforts

on the upper watersheds There are productivity gains to R and D

efforts and soil conservation adoption. Gains commensurate (or even higher) than efforts to increase productivity through encouraging input use Another bite at the Green Revolution

Summary Preview in the Context of the Larger Study Our initial review of literature for the

latter half of the study Hard to trace soil erosion from upland

farms to irrigation systems Many other sources of soil erosion (mining,

roads, etc.) Not all soil erosion goes directly towards

the irrigation systems. Some ends up in the banks of rivers and other reservoirs

Summary Preview in the Context of the Larger Study Point #1: Literature is clear however that

large development (and more disruptive land use changes like mining) are more likely to be more significant sources of soil erosion

Point # 2: Soil erosion from upland farms may therefore have very miniscule impact on irrigation systems and therefore on lowland agriculture

Point #1 + Point #2 implies that aggie to aggie externality may not be that great an issue

Summary Preview in the Context of the Larger Study Can we therefore justify intervention on

the basis of efficiency? Results of the larger study say: Maybe Not But it maybe justified on equity grounds,

i.e. to increase production (and thus income) in poverty stricken upland areas

We have been doing the right things for the wrong reasons because of lack of empirics.

What we failed to get

Actual irrigation and farm data in the lowland areas

Focused on aggie to aggie or farm to farm externalities

What about Farm to Fish externalities? Large part of eroded soil goes to coastal

areas and river banks Coastal non-point source pollution is a big

issue But this might be a more wieldy study

Maraming Salamat Po