Production and Productivity Growth in Chinese Agriculture:
New National and Regional Measures
Shenggen Fan*
Xiaobo Zhang*
International Food Policy Research Institute
2033 K Street, NW
Washington, D.C. 20006 U.S.A.
(Forthcoming in EDCC and revised on June 19, 2001)
Production and Productivity Growth in Chinese Agriculture:
New National and Regional Measures
I. Introduction
Agricultural output in China has been reported growing rapidly in the past several decades, particularly since the rural reforms that began in 1979. The State Statistical Bureau (SSB 1998) reported that from 1952 to 1997, output grew at 4.4% per annum.[1] Up until 1978, the annual rate was 2.8% per annum. It then jumped to 6.5% per annum during the period from 1979 to 1997. This long-term growth rate was one of the highest worldwide during the same period.
There have been numerous literatures in explaining China’s dramatic agricultural growth. Most studies (Weins 1982, Lardy 1983, Perkins 1984, Tang 1984, Lin 1988 and 1992, Fan 1990 and 1991, World Bank 1991, Wen 1993, and Zhang and Carter 1997) on Chinese agriculture have used gross value of agricultural output (GVAO) as an indicator to measure growth in agricultural output.[2] GVAO is reported by the State Statistical Bureau (SSB) --- the official government agency specialized in collecting, measuring, and reporting statistics in China.[3] However, GVAO is usually measured in constant prices (or comparable prices as described by the Chinese statistical system) to represent total output in a particular year.[4] But constant prices may not be the appropriate weights in aggregating total output because the growth rates calculated from these constant prices may be seriously biased, especially when relative prices have changed (Diewert 1976, Lau 1979, and Jorgenson 1995).[5] Relative output prices in Chinese agriculture have changed dramatically over time and across regions since the economic reform in the late 1970s.
In more recent years, many scholars have also questioned the accuracies of livestock and fishery output in Chinese official statistics. For example, Zhong (1997), Lu (1998), and Fuller et al. (2000) estimated that major livestock output reported by SSB may have been overstated by as large as over 40%, and fishery output by 70% in 1996.
Biased estimates of production and productivity growth in Chinese agriculture may have serious consequences, as most previous studies have used these indicators to judge the performance of the agricultural sector and to measure the impact of the rural reforms in Chinese agriculture. Moreover, since the officially published output values measured in constant prices are also widely used in the regional inequality literature, a regional inequality index calculated based on official data may be misleading as well.
This paper aims to properly measure national and regional growth in output, input, and total factor productivity in Chinese agriculture, and to reassess the impact of the recent policy reforms on production and productivity growth and regional inequality. The paper is organized as follows. The next section will review some aggregation issues in production theory. We show how estimates of aggregate output and input, and therefore measured total factor productivity, can be biased as a result of using constant prices as weights. The third section will be devoted to the measurement of growth in output, input and productivity, and regional inequality in Chinese agriculture. We conclude the paper in section 4.
II. Conceptual Framework in Output and Input Aggregation
The agricultural sector in China produces a great number of products including major staple grains like rice, wheat, and corn, major livestock products such as meat, eggs, and diary products, and horticultural and fishery products. Aggregation is often needed in order to compare the performance of the whole agricultural sector over time and across regions. Similarly, aggregation over inputs used in the agricultural sector is also necessary for number of reasons, such as providing information for measuring technical changes and efficiency improvement, and for measuring total factor productivity.
Agricultural output in the Chinese statistical reporting system is measured as gross value of agricultural output (GVAO) by summing production values of all products produced in the sector. Production value of a certain product is calculated by multiplying quantity by price. GVAO is often measured in current prices, and for the purpose of comparison across years, it is also measured and reported in constant prices[6].
Many economists have pointed out that using constant prices to aggregate output may result in biased estimate of production growth. Despite these concerns, many countries and international organizations still report growth in output aggregated using constant prices. This potential bias is illustrated in Figure 1 where Q0 represents a production possibility curve, which indicates the different combination of products Y1 and Y2, using the same amount of inputs[7]. Profit-maximizing producers choose different combination of Y1 and Y2 based on relative prices of the two products. Producers would choose point a in the production possibility curve when relative prices are P1, and b when relative prices are P2. If total output is aggregated using a liner aggregation of the two products weighted by their relative prices P1, aggregate output at a (equals to output at b') would be greater than that at b. But if P2 were used in the aggregation, output at b (equals to output at a') would be greater than that at a. Different output measures are obtained using different price weights, although producers only move along the same production possibility curve.
Figure 2 shows the potential bias that arises from input aggregation where I0 represents an isoquant in which the same amount of output is produced using different input combination, X1 and X2. Cost-minimizing producers choose input combination based on relative input prices, W1 and W2. If producers face relative prices W1, then the optimal combination of inputs would be at point c. If relative prices change to W2, the optimal combination of inputs would be at d. This shift is the producers' response to input price changes (the substitution effect) along the same isoquant. But using different relative prices as weights yields different input aggregates. For example if relative prices W1 are used as weights, aggregated input at d is greater than that at c (equals to output at d'). Conversely, if the relative price W2 is used, aggregate input at c is greater than that at d (equals to output at c'). The resulting productivity index using these biased estimates of aggregate output and input is also biased even when there has been no change in quantities of either inputs or outputs.
In order to minimize the potential bias caused by relative price changes, several approaches have been developed in the literature. The most commonly used method is the Divisa index. As Richter (1966) has shown, the Divisia index is desirable because of its invariance property: if nothing real has changed (e.g., the only quantity changes involve movements around an unchanged isoquant) then the index itself is unchanged. In practice, the Törnqvist-Theil (TT) index is usually used to approximate the Divisia index.[8] The formula for a TT index of aggregate output is:
1
(1)
where lnQItis the log of the aggregate output index at time t, S i, t and Si,t-1 are output i's share in total production value at time t and t-1, respectively. Yi ,tand Yi, t-1 are quantities of output i at time t and t-1, respectively. The advantage of such an index is that rolling weights accommodate any substantial changes in relative prices over time. Diewert (1976) and Lau (1979) have proved that the TT index is exact for the more general class of translog aggregator functions. The TT index of aggregated input growth can be expressed in a similar way.
Based on the growth of aggregated output and input, total factor productivity (TFP) is defined as the difference of these two. Specifically, the TFP index can be written as follows.
(2)
Where lnTFPt is the log of total factor productivity index, Wi,tand W i,t-1 are the cost shares of input i in total cost at time t and t-1, respectively, and Xi,t and X i,t-1 are the quantities of input i at time t and t-1, respectively.
III. Measures of Growth in Output, Input and Productivity, and Changes in Regional Inequality
3.1.Output Measures and Adjustments
As defined by SSB, GVAOrefers to total volume of output of farming, forestry, animal husbandry, and fishery in value terms, reflecting the scale of and the achievements made in agricultural production during a given period of time. The scope of official statistics on farming, forestry, animal husbandry, and fishery is as follows:
Farming is defined as cultivation of farm crops and other agricultural activities. Farm crops include grains, cotton, oil-bearing crops, sugar crops, bast fiber plants, tobacco, vegetables, medicinal herbs, melon and gourd crops, and cultivation and management of tea plantations, mulberry fields, and orchards. Other agricultural activities are defined as harvesting wild vegetation fruits, fiber, gum, resin, oil-bearing plants, grass, wild medicinal herbs, fungus plants, and rural household nonfarm activities.
Forestry refers to planting trees of various kinds (excluding tea plantations, mulberry fields and orchards), collection of forestry products, and cutting and felling of bamboo and trees by villages and other cooperative organizations under villages. Animal husbandry refers to raising and grazing of all animals except fishing and cultivating, and hunting and raising of wild animals. Fishery refers to cultivation and catching of fish and other aquatic animals and cultivation and collection of seaweed and other aquatic plants.
As indicated in the last two sessions, GVAO measured in constant prices cannot reflect the real growth of agricultural production. In this study, we use the TT index to measure production growth in Chinese agriculture for the past several decades. There are several hundred agricultural products covered by SSB, we try to cover as many as products as we can. In addition, adjustments have to be made to avoid double counting of livestock and fishery output. The following is a list of the commodities we covered and adjustments made to correct overstated output in this study.
Crops are divided into grain and cash crops. Grain crops include rice, wheat, corn, sorghum, millet, soybean, and tubers. Cash crops include cotton, hemp, jute, rapeseed oil, peanuts, sugarcanes, sugar beets, apples, bananas, citrus, grapes, pears, tea, tobacco, and vegetables. Livestock includes pork, beef, mutton, poultry, eggs, cow milk, and goat milk. Zhong (1997) found that meat output might have been overstated by 40% in 1996’s official statistics by comparing consumption data from rural and urban households surveys and official reported meat production. Lu (1998) estimated that official meat and egg output was overestimated by 52% in 1995. Fuller et al. (2000) further estimated meat output by different types and found that 1996 official statistics overstated output of pork by 65%, beef by 139%, poultry by 239%, mutton by 87%, and eggs by 64% and most of livestock products were underreported back in 1985.[9] First Agricultural Census data also show that the inventories of large animals have been overstated by 30%, swine by 36%, goats and sheep by 40%. However, the inventory data does not reflect overstated number of animals slaughtered, and weight per head when slaughtered. Therefore, in this study, we use Fuller et al.’s estimates to adjust the official statistics for pork, beef, mutton, poultry, and eggs.[10] The SSB has adjusted its meat production data since 1996 beginning in 1998’s China Statistical Yearbook. For 1997 livestock output data, we simply use the growth rate reported by SSB to extrapolate the output data, assuming there is no over reporting of growth in 1997. Fishery products are total weights of both sea and fresh water fish and other aquatic animals produced in a year measured in tons.[11] We use Lu’s estimate of fishery output to adjust official fishery output.[12]Forestry products are excluded in this study. The exclusion of forestry does not bias estimate too much, because it accounts for only 3% of total SSB's GVAO in 1997.
The official quantity data are taken from various issues of China’s Statistical Yearbook and China’s Agricultural Yearbook.[13] Prices are taken from China Commodity Price Yearbooks (various issues), China Trade and Price Statistical Materials, 1952-83, China Domestic Marketing Statistical Yearbooks (various issues), and USDA Agricultural Statistics of the People's Republic of China, 1949-90.[14]
Table 1 reports various measures of output growth for China as a whole. The SSB index is reported by various official statistics without adjusting meat and fishery output. The constant price index was calculated by the authors using the official price and output data (no adjustments were made on meat and fishery data) and 1980 constant prices in aggregation. The TT1 is the output index constructed using the Törnqvist-Theil index approach and official statistics without adjusting meat and fishery output. The TT2 index is the Törnqvist-Theil index constructed by the authors using adjusted livestock and fishery output data. The difference between SSB index and index measured in 1980 prices is solely due to the commodity coverage because the methodology used in aggregating these two indexes are the same. When comparing SSB’s GAVO and ours, we found the difference ranges from 20% in 1952, 1% in 1980, to 7% in 1997. Except for the 1950s, our coverage is reasonably good. The reason for large difference in the 1950s is due mainly to the fact that SSB’s GAVO includes values of manurial fertilizer and rural household’s small industry, and they are not in our commodity coverage and have been excluded from SSB’s GAVO since 1957.
The difference between the index measured in 1980 prices and TT1 index is due to the aggregation bias as we have illustrated in the previous section. Prior to the 1960s, the difference between these two indexes is small, being less than 2%. But the disparity grew larger over time. Particularly since the reforms, the difference has grown from less than 8% in 1978 to more than 31% in 1997, implying that part of the rapid growth in agricultural output measured in constant prices may come from the aggregation bias. There have been rapid price changes in agricultural products in Chinese agriculture since the reforms. Some of the substitution effects (or price effects) may have been captured as part of the production increase if constant prices are used in the aggregation.
The difference between TT1 and TT2 is due to the adjustments made to livestock and fishery output. The TT1 uses the official statistical meat and fishery data, while TT2 uses the meat and fishery data adjusted by the authors to avoid the over reporting of these outputs. The difference in 1981 is less than 1%, but gradually increased to 18% in 1997. Prior to 1988, agricultural production value was underreported while it was overstated thereafter due to the adjustment of livestock and fishery data.
Although still growing at a respective rate, the appropriately measured production growth is 4.6% per annum from 1979 to 1997, compared to 6.5% per annum reported by official SSB publications. In other words, the annual agricultural production growth reported by official statistics has been overstated as high as 1.9 % per annum from 1979 to 1997.
Table 2 reports measures of output growth at the provincial level. Due to data unavailability, we were able to construct provincial growth index only after 1979. But this may not be a serious problem because overestimation of production growth predominantly occurred after the reforms. The constant price index represents measures of growth using the 1980 constant prices and official SSB output data without adjusting meat and fishery output. The TT2 index is measured by the authors using the Törnqvist-Theil index approach and the adjusted SSB output data. One significant finding from the table is that overestimation has mainly occurred in less-developed areas. For example, growth in GAVO in Guizhou and Qinghai has been overestimated by more than 4% per annum, and that in Gansu has been overestimated by almost 3% per annum. This implies that the difference in development level between less- and more-developed areas is actually larger than that calculated from the official statistics.
3.2Measures of Inputs
In order to aggregate total input, both quantities and prices of individual inputs are required. Quantities are easy to obtain as SSB has been consistently publishing these input data since the 1950s. But prices of inputs are not readily available. Certain statistical techniques are required to extrapolate price data for some inputs.
3.2.1Quantities of Inputs
Labor Input. Labor input is measured in person-year equivalent of workers directly engaged in production of farming, animal husbandry, and fishery.[15] The labor of rural industry is excluded from agricultural labor input.
Land Input. Land input is measured as arable land. The official data on arable land are extremely inaccurate, and is reported to be understated by 30-40% from various sources. However, as long as the over reporting of land quantity is constant over time and the land input cost share is measured as residual as we will discuss later, the measure of input growth index will not be affected.