You are on page 1of 16

th

10 National Convention on Statistics (NCS)


EDSA Shangri-La Hotel
October 1-2, 2007

Technical Efficiency of Philippine Rice-Producing Regions:


An Econometric Approach

by

Niño T. Pate and Agustina Tan-Cruz

For additional information, please contact:

Author’s name : Niño T. Pate


Designation : Institutional Research Analyst and Assessment Consultant
Affiliation : Administration Division
Address : Palau Community College, Republic of Palau
Tel. no. :
E-mail :

Co-author’s name : Agustina Tan-Cruz


Designatio : Faculty Member, School of Applied Economics
Affiliation : University of Southeastern Philippines
Address : Bo. Obrero, Davao City
Tel. no. : (06382) 225-4696 loc 231
E-mail : a_tancruz@yahoo.com
Technical Efficiency of Philippine Rice -Producing
Regions: An Econometric Approach
by
Niño T. Pate1 and Agustina Tan-Cruz2

ABSTRACT

This study measures the technical efficiency of rice production in fifteen (15)
regions of the Philippines for the period 1991-2002. It measures the adequacy of different
estimation methods to find out a more accurate model to represent irrigated and rainfed
rice production in the Philippines and to provide a thematic map of the spatial distribution
of technical efficiency among rice-producing regions using Geographic Information
System (GIS). The frontier function involves inputs such as area of production, fertilizer
applied in kilogram of nutrients (N:P2O 5:K 2O), cost of labor, seeds/planting materials in
pesos, crop protection products, other miscellaneous inputs, year of observation, dummy
for seve re drought due to El Niño phenomenon and tropical cyclone passage.

Results showed that the half-normal distribution with time-varying technical


efficiency is an adequate representation of irrigated rice-producing regions in the
Philippines. The preferred model for the rainfed rice production in the Philippines is the
stochastic frontier production function with time-invariant technical efficiency, having
half-normal distribution.

I. Introduction

Rice is the most important food crop with more than 90% of the global production
occurring in tropical and semi-tropical Asia. The production of rice has increased faster than
population over the last three decades, despite being produced on small and marginal farms in
countries with extreme population pressure on limited land resources. Most of the growth in
production originated from technological progress in the irrigated and favorable rainfed
ecosystem (Hossain, 2004).

Recent trend analyses indicate that the growth of the rice sector has become
completely dependent on yield improvements (David and Balisacan, 1995 and Gonzales,
1998). Yield improvement can come in either of two ways: a) shifting the yield frontier, i.e.,
breeding varieties that have significantly higher yield potential than our current varieties, and b)
developing and promoting yield-enhancing technologies such as the use of high quality seeds
and efficient fertilizers. This raises the aggregate questions of how technically efficient the rice
production in the Philippines and how much production improvement can be made through
increased technical efficiency.

1
Institutional Research Analyst and Assessment Consultant, Administration Division, Palau Community College, Republic of Palau
Majority of the studies in the Philippines used the classical linear regression analysis to
estimate the production function. Most of the conducted studies were concerned only in
maximizing output but overlooking other variables that may influence the productivity of a
particular firm. Efficiency is a very important factor of productivity growth especially in
developing agricultural economies where resources are meager, and opportunities for
developing and adopting better technologies have lately started dwindling (Ali and Chaudhry,
1990).

Efficiency analyses using stochastic frontier approach on rice production in the


Philippines are rare. The application of stochastic frontier approach in the study may not appear
to be a radical innovation, but the linkage between methodology and theory allows more solid
hypothesis testing.

1.1 Efficiency Studies of Rice Production in the Philippines

Some of earlier studies in rice farming applying frontier methods done in the Philippines
include those of Kalirajan and Flinn (1983), Lingard, Castillo and Jayasuriya (1983), Kalirajan
(1984), Fare, Grabowski and Grosskopf (1985), and Dawson and Lingard (1989). Table 1 is a
summary of related studies, some of which were referred to by this paper.

Table 1. Technical Efficiency Studies of Rice Production in the Philippines,


1983-2004.
Year of
Author(s) Location Method
Publication

Kalirajan and Flinn 1983 Bicol Stochastic


Covariance
Lingard, Castillo and Jayasuriya 1983 Central Luzon
Analysis
Fare, Grabowski and Grosskopf 1985 Philippines Deterministic

Kalirajan* 1989 Selected Regions Stochastic

Dawson and Lingard 1989 Central Luzon Stochastic

Dawson, Lingard and Woodford 1991 Central Luzon Stochastic-Panel

Rola and Quintana-Alejandrino 1993 Selected Regions Stochastic

2
Faculty Member, School of Applied Economics, University of Southeastern Philippines, Davao City
Importante and Tan-Cruz* 2000 Selected Regions Stochastic

Larson and Plessman 2002 Bicol Stochastic

Gragasin, Maruyama and Kikuchi 2002 Mindoro & Cavite Stochastic


Umetsu, Lekprichakul &
2003 All Regions Malmquist Index
Chakravorty
Villano and Fleming* 2004 Central Luzon Stochastic

Adapted from Villano (2004); * author’s own literature search.

It has been observed that most of the studies on technical efficiency and produc tivity in
Philippine rice production are on a farm-level basis. Because of this limited literature, the
researcher conducted a study to estimate the technical efficiency of irrigated and rainfed
lowland rice production using stochastic frontier approach us ing regional-level panel data.

II. Objectives

The general objective of the study is to measure the technical efficiency of rice
producing-regions in the Philippines for the period 1991-2002.
Specifically it aims:
1. To come up with an adequate model to represent each of the irrigated and rainfed
lowland rice production of the entire Philippines;
2. To estimate the technical efficiency of irrigated and rainfed lowland rice production
of different regions;
3. To provide a thematic map of the spatial distribution of technical efficiency among
rice-producing regions using Geographic Information System (GIS).

III. Methodology

A production function is defined in terms of the maximum output that can be attained
from a given combination of inputs. For quite a time , the usual methodology used in the
optimization process had been the ordinary least squares (OLS) estimation.
In 1977, an alternative approach, called Stochastic Frontier Production estimation was
proposed, and this paper made use of said methodology.
3.1 Stochastic Frontier Production Function and
Technical Efficiency Measures

Aigner, Lovell and Schmidt (1977) and Meeusen and van den Broeck (1977)
independently proposed the stochastic frontier production function. Battese and Coelli (1992)
proposed a stochastic frontier production function for (un)balanced panel data, which has firm
effects assumed to be distributed as truncated normal random variables, and are also
permitted to vary systematically with time. For the i-th firm in the t-th time period, the model may
be expressed as:
Yit = X it ? + (V it - Uit) i = 1, 2, …, N, t = 1, 2, …, T
where
Yit is (the logarithm of) the production of the i-th firm in the t-th time period
Xit is a k×1 vector of input quantities (or their transformations)
? is a vector of unknown parameters
? v 2), independent of
Vit are random variables, which are assumed to be iid N(0,?
Uit which are non-negative random variables assumed to account for technical
inefficiency in production, and are assumed to be iid 1/2N(0,?? u2) .

Battese and Tessema (1993) specified the more general form of stochastic frontier
production function for (un)balanced panel data as:
Y it = Xit ? + (Vit - Uit) i = 1, 2, …, N, t = 1, 2, …, T
Uit = ? i + Ui = {exp[-? (t-T)] Ui
where ? ?
shows the time varying behavior of the non-negative firm effects. If ? ?
is zero, then the
inefficiency effects does not vary over time. The hypothesis that ? ?
is zero can be tested using
likelihood ratio test. The technical efficiency of a given firm is defined as the ratio of its mean
production given its realized firm effect, to the corresponding mean production if the firm effect
was zero. Thus, the technical efficiency of the i-th firm, denoted by TEit, is defined as

In particular, this study assumed half-normal and truncated normal distributions for Uit.
Further, the assumption of constant return to scale was relaxed.
3.2 Model Specification

Using the Cobb-Douglas production function, the stochastic frontier production is


expressed in logarithmic form as:
ln (QRPit) = ? ?+ ? ??ln(LRPit) + ? ??ln(QFAit) + ? ??ln(COLit) +
? ? ln(SPM
? it) + ? ? ln(CPP
? it) + ? ? ln(OMI
? it) + ? ? Year
? i+

? ? (Dummy1) + ? ? (Dummy2) + V it – Uit,


where the subscript i represents region and t represents the year of observation, and
QRPit = production in kilograms
LRP it = land area in hectare
QFA it = fertilizer applied, in kilograms of nutrients (N:P2O5:K2O)
COLit = the cost of labor in pesos
SPM it = cost of seeds/planting materials (in pesos, cash and non-cash)
CPP it = cost of crop protection products in pesos
OMI it = cost of other miscellaneous inputs in rice production in pesos
Yeari = year of observation for the i-th region
Dummy1 = severe drought dummy variable
Dummy2 = tropical cyclone dummy variable (D2=1 if the observation is from
region with more than 1 tropical cyclone “typhoons” per year
over its geographical zones; 0 otherwise).

The input costs are deflated using the regional Consumer Price Index with 1994 as
the base year. Data on quantities, prices, input cost of rice production and area planted are
from the publications of the Bureau of Agricultural Statistics (BAS).

Table 2 shows the six basic model specifications formulated in the estimation
process. Model 1 is the stochastic frontier production function in which the farm effects, Uit,
have the time-varying structure ( ? i?s an unknown parameter and the Ui’s are non-negative
truncations of the N(? ? u2) distribution). If ? ?is statistically significant the SFA presentation is
appropriate for the model but if it is closer to 0 h
t en the usual response function is the
appropriate presentation for the model. Model 2 is the special case of model 1 in which the Ui
? v2). Model 3 is the time-invariant model with
have half-normal distribution N(0,?
truncated-normal distribution considered by Battese, Coelli and Colby (1989). The model
implicitly assumes that variation in all regional effects (technical efficiency) is monotone
throughout time periods and that one rate of change applies to all regions. Furthermore, if the
estimated ? is significantly different from zero and positive, then the non-negative firms’
technical inefficiency in production is not only time-variant, but also decreasing, i.e., technical
efficiency is improving. Model 4 is the time-invariant model in which the farm effects, Ui have
half-normal distribution. Model 5 is the traditional average response function in which farms
are assumed to be fully technically efficient (i.e., the regional effects, Uit are absent from the
model). Finally, Model 6 is the corrected ordinary least squares (COLS) estimation method
suggested by Greene (1980).

Table 2. Major Characteristics of Models Used in the Analyses.

Efficiency Functional One-sided


Model Assumption
Measure Form Error Term

Model 1 ? ?and ? ?unrestricted TV CD TN

Model 2 ??
=0 TV CD HN

Model 3 ??
=0 TI CD TN

Model 4 ? =??
=0 TI CD HN

Model 5 ? =? =??
=0 Ordinary Least Square (OLS)

Model 6 ? ?Adjusted Corrected Ordinary Least Square (COLS)

Legend: TV – Time-Varying; TI – Time-Invariant; CD – Cobb-Douglas;


TN – Truncated-normal Distribution; HN – Half-normal Distribution

IV. Results and Discussion

The estimates of the parameters and results of the likelihood ratio tests are presented
in Tables 3-6.
It is evident that the average production function is not an adequate representation of the
data. Further, the hypotheses that the models are time-invariant are also rejected (i.e., Ho: ? =
??
= 0 and Ho: ? ?
= 0, rejected). However, the half-normal distribution with time-varying technical
efficiency is an adequate representation for the data given the specification of the stochastic
frontier model (i.e., Ho: ? = 0, is not rejected).
Table 3. Estimates of Stochastic Frontier Production Function ( Irrigated )
Maximum Likelihood Estimation
OLS COLS
Variable Parametr Model Model Model Model
Model 5 Model 6
1 2 3 4
?? 6.48* 6.47* 6.39* 6.39* 6.36* 6.32*
Constant
(0.22) (0.24) (0.26) (0.27) (0.21) (0.20)
ln (LRP) ?? 0.80* 0.78* 0.76* 0.68* 0.77* 0.73*
- ha (0.05) (0.06) (0.07) (0.08) (0.05) (0.05)
ln (QFA) ?? 0.19* 0.20* 0.23* 0.26* 0.21* 0.22*
- NPK,kg (0.04) (0.04) (0.05) (0.05) (0.33) (0.03)
ln (COL) ?? 0.05 0.05 0.03 0.03 0.03 0.02
- Pesos (0.04) (0.04) (0.04) (0.05) (0.04) (0.04)
ln (SPM) ?? 0.06* 0.06* 0.05* 0.05* 0.06* 0.05*
- Pesos (0.01) (0.01) (0.02) (0.02) (0.01) (0.01)
ln (CCP) ?? 0.01 0.01 0.02 0.02 0.01 0.01
- Pesos (0.01) (0.01) (0.01) (0.02) (0.01) (0.01)
ln (OMI) ?? 0.06* 0.06* 0.05* 0.05* 0.04* 0.04*
- Pesos (0.01) (0.02) (0.02) (0.02) (0.01) (0.01)
Year ?? 0.05* 0.06* 0.03 0.04 0.06* 0.07*
Drought ?? -0.06* -0.08* -0.07* -0.07* -0.07* -0.07*
Dummy (0.01) (0.01) (0.01) (0.01) (0.01) (0.01)
Tropical ?? -0.13* -0.13* -0.12* -0.12* -0.11* -0.11*
Cyclone (0.02) (0.02) (0.02) (0.02) (0.02) (0.02)
Dummy
? s2 0.11* 0.13* 0.01 0.12* 0.01 0
(0.01) (0.01) (0.01) (0.00)
? 0.27* 0.86* 0.56* 0.29* 0 0
(0.01) (0.02) (0.27) (0.21)
? -0.01 0 -0.16 0 0 0
(0.02) (0.26)

? 0.36* 0.42* 0 0 0 0
(0.04) (0.05)
217.6 210.1 210.0
Log likelihood Function 217.92 207.79 208.71
2 7 5
Note: 1. Model 1 is with ? a
?nd ? unrestricted, Model 2 is with ? = 0,
Model 3 is with ? ? ?Model 4 is with ??
?? ??
Model 5 is OLS estimation with ? ? ? ? ? ? ?and
Model 6 is COLS estimation ? ?adjusted).
2. Figures in parentheses are standard errors of
estimates

Table 4. Likelihood-Ratio Tests for Irrigated Rice Production in the Philippines.

Model Assumption LR (λ) χ 2 (0.95, df) Decision

Model 1 µ?and η?unrestricted Given Model


Model 2 µ=0 0.61 3.84 Do not reject
Model 3 η?= 0 14.90 3.84 Reject
Model 4 µ=η =0 15.11 5.99 Reject
Model 5 (OLS) γ = µ = η =0 19.65 7.82 Reject
Model 6 (COLS) (β ?Adjusted) 17.82 9.49 Reject

The preferred model for irrigated rice production is Model 2. Rainfed rice production in
the Philippines exhibits technical inefficiency and do not improve the level of
efficiency over the period. The hypothesis that half-normal distribution with time varying
technical efficiency (i.e., Ho: ? = 0) was not rejected.

The preferred model for the rainfed rice production is the model with time-invariant
technical efficiency having half-normal distribution (i.e., H o: ? = ? ?
= 0), or Model 4.

Table 6. Estimates of Stochastic Frontier Production Function for Rainfed Rice


Maximum Likelihood Estimation
OLS COLS
Variable Parametr Model Model Model Model
Model 5 Model 6
1 2 3 4
?? 5.71* 5.68* 5.42* 5.30* 4.75* 4.72*
Constant
(0.28) (0.35) (0.30) (0.34) (0.24) (0.23)
ln (LRP) ?? 0.64* 0.64* 0.64* 0.64* 0.71* 0.71*
- ha (0.08) (0.09) (0.08) (0.09) (0.07) (0.07)
ln (QFA) ?? 0.12* 0.12* 0.16* 0.14* 0.17* 0.13*
- NPK,kg (0.04) (0.04) (0.04) (0.04) (0.04) (0.04)
ln (COL) ?? 0.07 0.07 0.08 0.08 0.12* 0.10
- Pesos (0.08) (0.08) (0.07) (0.08) (0.05) (0.05)
ln (SPM) ?? 0.15* 0.16* 0.20* 0.15* 0.19* 0.16*
- Pesos (0.11) (0.19) 3.00 (0.19) (0.05) (0.05)
ln (CCP) ?? 0.09* 0.08* 0.09* 0.11* 0.07* 0.09*
- Pesos (0.02) (0.02) (0.02) (0.02) (0.01) (0.01)
ln (OMI) ?? 0.13* 0.14* 0.21* 0.12* 0.20* 0.22*
- Pesos (0.07) (0.08) (0.06) (0.08) (0.03) (0.03)
?? 0.03 0.04 0.09* 0.08* 0.04 0.03
Year
(0.07) (0.08) (0.06) (0.08) (0.03) (0.03)
Drought ?? -0.05* -0.06* -0.03 -0.15* -0.04 -0.04
Dummy (0.02) (0.02) (0.02) (0.02) (0.02) (0.02)
Tropical ?? -0.12* -0.08 -0.18* -0.18* -0.18* -0.16*
Cyclone (0.06) (0.10) (0.08) (0.10) (0.03) (0.02)
Dummy
? s2 0.32* 0.11 0.37 0.21* 0.02 0
(0.17) (0.06) (0.22) (0.06)
? 0.97* 0.90* 0.97* 0.90* 0 0
(0.02) (0.06) (0.02) (0.06)
? -0.12 0 -0.21 0 0 0
(0.82) (0.43)
? 0.00 0.00 0 0 0 0
(0.01) (0.01)
Log likelihood Function 132.16 131.16 132.23 132.25 118.19 118.03

Elasticities of output for land area and cost of seeds and planting materials are the two
most important inputs determining the yield of rainfed rice production. The signs are positive
which conforms to a priori assumption that the estimated production function is an increasing
function.
4.1 Technical Efficiency Estimates

The estimated mean technical efficiency of irrigated rice-producing regions in the


Philippines tends to fluctuate over time. As can be seen in Figure 1, an abrupt decline of
technical

Figure 1. Predicted Mean Technical Efficiency of Philippine Rice Production


efficiency for both r ainfed and irrigated rice production are observed in 1995, 1996 and in 1998.
One of the major causes for this drop is the severe drought experienced by some regions in the
Philippines in 1995 and by all regions in 1998. It was in 1998 also that the super-typhoons
Emang and Gading ravaged the Philippines. The recovery in 1997 could be partly explained by
the Gintong Ani program launched by the Department of Agriculture.

V. Conclusions

Following are the important empirical findings of the study:


(1) The traditional average production function is not an adequate representation of
irrigated rice-producing regions in the Philippines. The technical efficiency is time-variant.
(2) The preferred model for the production of rainfed rice in the Philippines is the
model with time-invariant technical efficiency, having a half-normal distribution.
.(3) The predicted mean efficiencies are estimated to be 0.924 and 0.898 for irrigated
and rainfed rice production respectively. About 75.56% of the technical efficienc ies for irrigated
regions clustered around 0.901 to 1.00. For the rainfed regions, the figures are 51.11%
clustering around 0.901 to 0.950, respectively.
(4) For irrigated rice production, the Caraga, Cagayan Valley and Northern Mindanao
regions are considered to be within the more efficient bracket, while Bicol region, Ilocos region
and Western Visayas are within the less efficient group. For the rainfed case, the Ilocos region,
Central Luzon and Southern Mindanao are considered to be the more efficient while Western
Visayas, Western Mindanao and CAR are among the less efficient.
References

Aigner, D.J., Lovell, C.A.K. and Schmidt, P., 1997. “Formulation and Estimation of Stochastic Frontier
Production Function models,” Journal of Econometrics, 6: 21-37.

Battese, G.E. and Coelli, T.J., 1992. “Frontier Production Function, Technical Efficiency and Panel Data:
with Application to Paddy Farmers in India,” Journal of Productivity Analysis,

Coelli, T.J., 1994. A Guide to FRONTIER Version 4.1: “A Computer Program for Stochastic Frontier
Production and Cost Function Estimation,” Department of Econometrics, University of New
England, Armidale, Australia.

Hossain, M., 2004. “Long-term Prospects for the Global Rice Economy,” Paper presented to FAO Rice
Conference 04/CRS.1.
Pitt, M.M. and Lee, L.F., 1981. “Measurement and Sources of Technical Inefficiency in the Indonesian
Weaving Industry,” Journal of Developmental Economics, 9:43-64.

Villano, R.A. and Fleming, E., 2004. “Analysis of Technical Efficiency in a Rainfed Lowland Rice
Environment in Central Luzon Philippines using a Stochastic Frontier Production Function with
Heteroskedastic Error,” Working Paper Series in Agricultural and Resource Economics.
University of New England, Armidale, NSW, Australia.
Appendix A

Appendix A. Distribution of Technical Efficiency for Irrigated Rice Production in the


Philippines, 1995-1998, by Region
Appendix B

Appendix B. Distribution of Technical Efficiency for Rainfed Rice Production in the


Philippines, 1995-1998, by Region

You might also like