Abstract
Among several
factors causing food insecurity problems, inappropriate policy formulation and
absence of technological progress were repeatedly emphasized. Thus,
Autoregressive
Conditional Heteroscedasticity (ARCH) was used to measure the volatility of
food grain prices, and a Vector of Error Correction (VEC) models was applied
estimate its impact on the utilization of modern input and supply of major food
grains on time-series data. The results
indicated that both short run and long run price volatility of major cereals
have significantly increased in the post reform period. Furthermore, this price
volatility coupled with changes in weather conditions and relative prices ratio
was found to be considerably influencing farmers’ incentive to use modern
inputs and the amount of grain food supply.
Consequently, farmers cannot sustainably use of improved technologies to mitigate food insecurity problem. Neither the farm households have storage facilities nor has the government buffer stock facilities to mitigate the instability problem. Therefore, for a sustainable growth in food production and agriculture, improving storage facilities and genuine government intervention to stabilize the grain system, at least in the short-run till the domestic market is sufficiently developed, is greatly recommended. Enhancing irrigation technology and development of market infrastructure and information would be suggested for a long run development of the food sector.
Keywords: Food insecurity, market liberalization, modern
input use, instability and technological changes
1
Introduction
Severe food insecurity problems have been observed under almost
all government regimes in
The problem
of food insecurity may be associated not only with production but also with
marketing and distribution issues. Marketing policies affect the variability of
food prices and hence supply, which ultimately affects the vulnerability of
low-income households to food insecurity. In
The accumulated evidence does indeed indicate that market
liberalization affects the rate of economic growth posititively, and hence the
extent to which households may be lifted out of poverty (Schuh, 2002). However,
there is no guarantee that the free market system can work effectively in an
economy where the market itself is not (well) developed. Schultz (1978) argues
that the paradigm of free market presupposes full information, perfect
transport with minimum transaction costs, and a large number of both producers
and consumers. And he underlines that these assumptions are far from the
reality of many developing countries. Moreover, in a poorly developed market,
farmers lack full information on future prices and hence perceive prices as
unpredictable. Thus, it is strongly argued that fully liberalizing prices could
increase the risk of price instability, curbing the incentive to use improved
technologies and ultimately worsening food insecurity problems. So, the gains
from agricultural liberalization may be overstated. In such a case,
particularly, for basic food products that have an important role in economic
stability and for the welfare of the population, the free market paradigm may
be misleading.
Despite the considerable influence of price volatility on
the performance of the agricultural sector in particular, little research has
been done in this area in
In
During the study period,
chemical fertilizers (DAP and Urea) and improved seeds were the only modern
inputs available to small farmers in
For this analysis, data sets on annual quantities of
fertilizer and improved seeds consumed by small farmers were obtained from the
National Seed Enterprise and Fertilizer Agency. The data set of monthly prices
of the cereals observed in the central market (
2
Trends in
Prices, Modern Input Use and Food production
2.1
Trends in Prices of Modern Inputs and Cereal Products
During the socialist period, prices of fertilizer and improved
seeds were subsidized and controlled, keeping them at a low level. Thus, there
was only a slight upward movement. After the reform of 1992, the general trend
in the prices of these inputs was always upward, with minimum fluctuations. Two
main causes may be identified. The first cause was the increasing depreciation
of the local currency, and the second was poor market performance (development)
for such inputs. Market for fertilizer was monopolized by a few regional
government-affiliated importers and distributors, and participation by private
firms in the market was negligible. The market for improved seed was even less
well developed, and thus it was unavailable altogether in the open market. In
the period from 1993 to 2002, the price of DAP rose from 176 to 265 birr per
100kg (a 50% increase), while the price of urea rose from 156 to 195 (a 25%
increase). Similarly, prices of improved seeds increased dramatically in the
same period. Improved maize seed (hybrid) increased from 178 to 235 birr per
100kg, improved wheat seed from 140 to 260 birr, and improved tef seed from 153
to 353 birr. These amounted to an increase of 32%, 86% and 131% respectively
for improved seeds of maize hybrid, wheat and tef.
Conversely, prices of major agricultural products swing
seasonally and were highly volatile over all the years after the reform.
Seasonal ups and downs in producer prices were documented in Mulat (1999) and
Wolday (2002). Prices fall dramatically immediately after harvest, from January
to March. This is because nearly all farmers undertake the largest part (79%)
of their annual sales at this time due to lack of appropriate storage
facilities and the need to repay loans (particularly for the purchase of
fertilizer and improved seed) and to meet other financial obligations (e.g.,
taxes). Hence prices generally drop sharply, which leads to a decline in the
net return to the farm. On the other hand, food crop prices rise later (June to
September) in the year when many farmers run out of stock and hence supply is
relatively low, and some poor farmers even start to buy from the market.
Three of the most important cereals, namely maize, wheat
and tef, all with average quality, were considered for illustration. Figure 1
depicts the yearly nominal price fluctuation for these crops in the capital,
Following the substantial increase in total cereal
production due to increased use of fertilizer and improved seed coupled with
favorable rainfall immediately after the reform till 2001, in the subsequent
years (2000 to 2002) prices frustratingly dropped sharply. Prices fell from
their respective maxima in 1999 to minima of 45, 95 and 120 birr per 100kg for
maize, wheat and tef, respectively. This amounted to a fall of 68%, 55% and 55%
for maize, wheat and tef respectively. This is the largest recorded price slump
in the history of
Table 1: Variability in Crop Prices (measured
in Standard of Deviation)
|
Observation
period |
Tef
price |
Wheat
price |
Maize
price |
Weighted
average crop price |
|
before 1991 |
22.67 |
20.82 |
12.81 |
18.14 |
|
1991-1993 |
42.83 |
31.95 |
32.65 |
35.55 |
|
1994-2000 |
29.33 |
45.24 |
31.28 |
34.75 |
|
2001-2003 |
37.74 |
49.10 |
57.61 |
35.52 |
|
1992-2003 |
36.95 |
44.69 |
46.33 |
40.32 |
Source: Own computation from CSA database
The standard
deviations of prices of the selected crops are presented in Table 1. The
differences in the standard of deviation in the different periods constitute a
significant structural break in the level of volatility of crop prices. On
average, the standard deviation of individual and aggregate prices increased
strongly, suggesting an increase in the volatility of the food grain prices
after the reform. A similar situation could be observed with regard to their
aggregate volume of production (Table 2).
2.2
Trends in
Modern Input Use
Figure 2 portrays the general
trends as regards purchased improved seed and chemical fertilizer consumption.
It is clearly discernible that consumption of these modern inputs followed that
of output prices (compare with Figure 1). Compared to the pre-reform period
(before 1992), fertilizer consumption dramatically increased, with acceptable
ups and downs. For instance, in 2001, about 2.6 times as much fertilizer as in
1992 was consumed by small-scale farmers, representing a 161% increase. Over
the period between 1994 and 2002, fertilizer consumption grew at an average
compound rate of only 5.61%. According to Mulat (1999), this achievement was
mainly owing to increased donor support in the form of funds for fertilizer
imports, coupled with favorable climatic conditions in the stated period. Nevertheless,
it was still far below the level recommended by the World Bank)
Improved seed is almost always used in tandem with fertilizer. Similar to the case of fertilizer, use of improved seeds grew rapidly amounting to a compound growth rate of about 47% per annum after the reform (1992-2001). In 2001, for instance, about 2.6 times the amount of improved seeds was used compared to 1992 (i.e. an increase of over 160%). Hitherto, ESE (2001) revealed that overall utilization of improved seeds had only been 2% of the annual average national improved seed requirement, which is estimated at about 4 million tons for the different crops.
Figure 1: Trends in Prices of
Major Food Grains
Source: Own compilation based on CSA database
The lowest prices were recorded in 2001 which is almost the same with the prices in 1994 (about 7 years back). The data is to be updated for 2002/2003.
Disappointingly, the use of both chemical fertilizer and
improved seeds has fallen radically since 2001. Owing mainly to plummeting
grain prices, farmers have reverted to the meager level they used 10 years ago;
and because of the lag effect, input consumption did not rise along with the
crop output prices in the subsequent years. Consumption of purchased improved
seed fell much faster than fertilizer use. Such trends could have significant
repercussions for future food production and food security.
Figure 2: Trends in
Total Fertilizer and Improved Seed Consumption by Small Farmers
Source: Own compilation based on ESE database
Note: The figure shows that
the extent of fertilizer and improved seed use follows in the footsteps of
fluctuations in food grain prices. The amount of improved seed used by the
small farmers in the 2002 was exactly equal to that amount used before 10
years, in 1992/93.
2.3
Performances
and Trends in Cereal Food Production per capita
Figure
3 portrays the annual changes and long-run trends in major food grain
production per capita in
Comparing trends
between the pre- and post-reform period, the trends in the per capita food
production index for the two periods are essentially the mirror image of each
other. In the socialist period, between 1975 and 1991, food crop production was
growing at a mere 1.2% while the population was growing at 2.64%, resulting in
a downward trend in per capita production of food grain at a rate of -1.44%.
With the exception of the early 1980s, the food production index declined
continuously throughout the pre-reform period until the end of the regime.
Figure 1: Trends in Major Food Crop
Production per capita (kg per person per year)
LR Post-reform Pre-reform
Source: Own computation.
After the policy
reform, the overall performance of food crop production was appreciably
positive between 1992 and 2001. Food production per capita grew by about 34%
(i.e., from an average of about 146 kg in 1993 to about 196 kg in 2001). In
that period, the compound growth rate in food production was impressive, reaching
4.52% per annum (significant at 5%). The main source of the improvement in food
production in the period under observation was the improvement in cereals,
mainly due to technological progress and increasing use of modern inputs (in
fact only for some food crops), coupled with relatively good weather condition.
This observation gives the impression that, under favorable climatic conditions
and institutional support, there is considerable potential for increasing land
productivity in such a way as to enhance food self-sufficiency.
However, Output suddenly
collapsed in 2002/03 because of the customary incidence of drought mainly
because of a decrease in the use of chemical fertilizer and improved seed
caused by drought and a drastic fall in output prices. This implies that the government
policy failed to bring about sustainable food production. In general, for the
food security system to operate, the absolute magnitude of variability in
output is critical. Production was not stable in
In general, food
production per capita remained stagnant, and was characterized by severe
fluctuations during the whole observation period, both before and after the
reform. Consequently, an increasing share of the population, particularly in
marginal and drier areas, was tormented by hunger every year. Thus,
agricultural production was very vulnerable to many kinds of risks and
uncertainties. Moreover, the dynamics of product prices have greater
implications for food security in a liberalized market economy.
3
Econometric
Analysis of Output Price Volatility and its Impacts on Modern Input Use and
Food Grain Supply
3.1
Estimating
Volatility (Instability) in Grain Prices for Selected Cereals
3.1.1
The
Model
Most statistical tools are designed to model the conditional mean
of a random variable. Nevertheless, variance is often used to measure
volatility, which is a key element in pricing theories. This is because price
variability (volatility) is among the most important sources of uncertainty.
Most studies assume that price variances follow persistent short-term
stochastic models, such as the Autoregressive Conditional Heteroscedasticity
(ARCH) model developed by Engel (1982) and/or the Generalized ARCH (GARCH)
model developed by Bollerslev (1986). The models measure the degree of dynamics
in conditional mean and variances. Studies using these models have found that
the volatility process is highly persistent (Baillie et al., 1996; Bollerslev
and Mikkelsen, 1996). The standard specification of the GARCH model is as
follows:
; Mean equation (1)
; Conditional variance
equation (2)
where equation (1) represents
the mean equation of the price of a commodity (pt) as a function of
exogenous variables (Xt) and an innovation term (εt)
with a white noise (stationary) series. Equation (2) refers to the conditional
variance; δt2 is the one-period-ahead forecast
variance based on past information; and ε2t-1 (also
called the ARCH term) denotes news information about volatility from the
previous period, measured as the lag of the squared residual from the mean
equation (1); δ2t-1 (the GARCH term) is the last
period’s forecast variance. It is worth mentioning that model (2) is simply the
result of the first model (1). The coefficients γ, ω, β and
α are unknown parameters to be estimated. Nelson and Coa (1992)
demonstrate that restricting ω > 0, α ≥ 0, and β
≥ 0 is a sufficient but not a necessary condition for the non-negativity
of δ2. Bollerslev and Ghysels (1996) suggest that the α
coefficient can be interpreted as a measure of the short-run (immediate) impact
of “news arrival” on volatility, while β controls the long-term evolution
of the volatility process. The sum of α and β measures the degree of
price volatility.
3.1.2
Analysis
and Results
Descriptive statistics of some selected economic variables
(1980 to 2003) are given in Table A1 in the Annex, and the ADF and PP test
results for selected variables (crop prices) are presented in Table A2 in the Annex.
Both unit root test tools fail to reject the null hypothesis of a unit root in
all price series at level, and hence they are all non-stationary. According to
the PP test, we reject the unit root in the first differences, suggesting that
the data series contains one unit root and is of integration order one I(1).
Nevertheless, with the ADF test statistics, the variables were found to be of a
I(2) process, which is in line with the findings of much of the literature
which suggest prices are mostly I(2) (Wooldridge, 2003). Weighted prices and
volume of production of cereals are also I(1) according to the unit root test
results of both the ADF and the PP. Based on the PP test results, and according
to the suggestion by Engle and Granger (1987), the variables under consideration
are, by definition, cointegrated since they have the same order of integration.
That is, the series are drifting together. So, their linear combination (a
regression equation) at level can produce a long-run relationship between the
variables.
The models presented in equations (1) and (2) above were
simultaneously estimated using maximum likelihood. The dependent variable is
product monthly price of the selected crops. Initially, many variables
hypothesized to explain changes in cereal prices were included in the
estimation. These included: current and lagged output of major cereals
(supply), money supply to GDP ratio (income), real exchange rate, a weather
dummy (to capture productivity change and hence any shift in the supply
function), and total volume of food imports (supply of substitutes). Moreover,
prices of substitutes for each individual crop were included as determinants,
while the price index of pulses was included to explain the overall price index
of the major cereals.
Table 1: GARCH (1, 1) Maximum
Likelihood Estimation for Volatility of Major Food Grain Prices (1992-2003)
|
Coefficients |
Tef price |
Wheat price |
Maize price |
Cereals price |
|
|
Mean (ω) |
0.012 |
0.015* |
0.023** |
0.004 |
|
|
ARCH (α) |
0.037 |
0.308** |
0.436** |
0.414* |
|
|
GARCH (β) |
0.544** |
0.646** |
0.532** |
0.443* |
|
|
(α +
β) |
0.581 |
0.954 |
0.968 |
0.857 |
|
|
Log likelihood
|
-15.21 |
-8.180 |
15.57 |
10.89 |
|
|
Durbin-Watson
stat. |
1.67 |
1.743 |
1.87 |
2.160 |
|
|
Akaike info.
criterion |
-0.610 |
0.743 |
-0.460 |
0.258 |
|
|
Schwartz
criterion |
-0.156 |
0.034 |
-0.159 |
0.021 |
|
|
F-statistics |
7.56** |
8.69** |
21.14** |
13.45** |
|
Source: Own computation based on monthly data from CSA
Note: * (**) indicates a 5% (1 %) significance level
It was found, however, that changes in prices of the selected crop grains were significantly explained only by their respective lag price, total cereal output and the weather dummy. All the other variables were statistically not significant, and the direction of the relation was economically not meaningful given the data and the model specified, and was therefore excluded from the model. (Further research could be suggested here.) As our main objective was to measure the extent of price volatility, only the results of model (2) of the GARCH (1, 1) analysis are reported in Table 3. In general, the results of the models are quite satisfactory. In each case, the test for normality of residuals was not rejected by the Jarque-Bera statistic, and all the other diagnostic tests moderately satisfied the classical regression axioms (not indicated here). The Eviews software performs both unit root tests, and the unit root procedure was applied for all data series used in the estimation of models 1 through 4, and description of the variables used in the analysis are summarized in Table A1 in the annex
From visual examination of the data series (section 2.2), it was observed that, after market liberalization, prices of the major food crops became increasingly unstable over time. The results of the econometric analysis also reinforce the volatility in the prices of the selected series. Both the short-run (α) and the long-run (β) volatility measures are statistically highly significant, particularly for wheat and maize. Because we used monthly data, the long-run volatility index roughly represents seasonal and annu