China’s Invisible Chicken Losses: Production Costs Effect of Highly Pathogenic Avian Influenza

Abstract

Highly pathogenic avian influenza (HPAI) affects chicken production not only during outbreaks but also afterward. Understanding its delayed effect is essential for facilitating timely production recovery. Employing a dynamic panel data model with annual Chinese provincial data obtained between 2004 and 2021, we quantified the impact of previous-year HPAI outbreaks on current-year chicken production through production costs. The results indicated that a 1% increase in provincial HPAI outbreaks raised production costs per 100 broilers by 0.372%, ultimately reducing annual chicken production by 0.038%. These findings remained robust after controlling for endogeneity and conducting extensive robustness checks. The impact was most pronounced in provinces characterized by high chicken production, a high proportion of scale broilers, and yellow-feathered broiler specialization, where both production costs and production losses were significantly greater. Additionally, previous-year HPAI outbreaks significantly increased production costs by increasing both epidemic prevention and broiler chick costs. Our findings offer robust empirical evidence and actionable insights for managing cost volatility risks along the chicken supply chain in post-epidemic contexts.

1. Introduction

China is the third largest chicken producer and second largest chicken consumer in the world [1]. In 2022, China’s chicken production reached 14.3 million tons, accounting for 11.57% of the global production, ranking third after the United States (20.65%) and Brazil (14.27%), which increased by 70.97% over 2000 [2,3]. With the increasing demand for chicken, China’s chicken consumption per capita increased from 7.41 kg in 2000 to 10.18 kg in 2023 [4]. Ensuring and increasing China’s self-sufficiency in chicken production not only meets the growing domestic demand but also contributes to stabilizing the global chicken supply and prices.

Highly pathogenic avian influenza (HPAI) is a zoonotic infectious disease caused by the H5 and H7 HA subtypes of avian influenza virus [5]. It is classified as a notifiable animal disease by the World Organization for Animal Health (WOAH) due to its high infectivity and mortality rates in poultry [6]. HPAI directly caused extensive mortality and culling of poultry, leading to substantial production losses [7,8]. Veterinary Bulletin of China’s Ministry of Agriculture and Rural Affairs indicated that there were 176 HPAI outbreaks in 24 Chinese provinces from 2004 to 2021. A total of 804,901 live chickens were infected during the epidemics, 577,105 died, and 120,622,970 were culled [9].

Although HPAI was ultimately brought under control, poultry production capacity failed to rebound promptly, resulting in invisible production losses [10]. A survey conducted in Ningxia Province of China revealed that, among poultry producers previously affected by HPAI, 44.64% downsized their flocks and 27.54% resumed production slowly [11]. Approximately 49% of surveyed Chinese poultry producers whose flocks were affected by HPAI have yet to restore their production capacity to pre-outbreak levels [12]. Hence, it is imperative to elucidate the reasons for the slow recovery of poultry production capacity once HPAI outbreaks have been brought under control, as this information could be used to formulate evidence-based policies that ensure a stable supply of poultry products.

The sluggish restoration of poultry production is ascribed to the persistent impacts of HPAI on poultry production [13,14]. Prior studies have examined how past HPAI outbreaks affect subsequent poultry production, focusing on market price responses and producers’ risk aversion. For example, You and Diao [15] employed a spatial equilibrium model and showed that HPAI-induced contractions in Nigerian poultry demand depressed domestic poultry prices, which, in turn, cut the subsequent batch production by up to 21%. Ifft et al. [13] estimated that medium-scale Vietnamese broiler producers reduced their flock size by an average of 2.96 birds in response to HPAI outbreaks in the preceding year.

To our knowledge, little attention has been paid to production costs. Although producers incur direct costs from poultry mortality during HPAI outbreaks [16,17], evidence on whether and how past HPAI outbreaks influence current production costs remains scarce. Previous research has shown that past HPAI can reduce current poultry production by raising both epidemic prevention costs and chick costs. In response to the perceived risk of future HPAI outbreaks, poultry producers typically raise expenditure on preventive biosecurity measures, but a subset appear to engage in precautionary over-investment, incurring epidemic prevention costs that exceed expected marginal benefits [18,19]. Antimicrobial Agents Intended for Use in Animals reported that China’s use of antibiotics in chickens is at a relatively high level compared to the rest of the world [20]. For example, 75% of surveyed broiler producers used prohibited antibiotics, and 4.8% of them continued to use antibiotics during the drug withdrawal period [21]. Data suggest that broiler producers are prone to raising production costs by over-investing in immunization, medication, and disinfection aimed at HPAI risk mitigation, thereby potentially impeding production capacity restoration. Moreover, commercial broiler production is contingent on a multi-generational breeding pyramid, requiring approximately 1.5 years for genetic material to propagate from primary breeders to market-ready broilers [22,23].

In China, HPAI has repeatedly affected great-grandparent and grandparent breeder farms in Shanxi, Anhui, Hunan, Liaoning, and Guangxi, causing substantial mortality and compulsory culling since 2004 [9]. This is expected to elevate broiler feed costs within one year, thereby exerting negative effects on chicken production. Past HPAI outbreaks are likely to significantly impact these two types of costs, so our study aimed to confirm whether HPAI outbreaks in the previous year affect chicken production in the current year, considering these costs.

Relative to the extant literature, this study makes two important contributions. One involves examining the impact of the previous year’s HPAI on the chicken production in the current year from the perspective of production costs to identify the impact of past HPAI on the poultry industry. The other involves testing whether the previous year’s HPAI impacts chicken production via the rising costs of epidemic prevention and chicks, thus informing cost management strategies. The remainder of this paper is structured as follows: Section 2 presents the data and empirical approaches used in the analysis. Section 3 describes our main empirical results. Section 4 discusses the results. Section 5 concludes this paper.

2. Materials and Methods

2.1. The Context of HPAI Outbreaks in China

The pattern of HPAI outbreaks in Chinese chickens typically begins by initially detecting the HA HPAI subtype, followed by nationwide compulsory poultry immunization to transform concentrated HPAI outbreaks into sporadic and local HPAI outbreaks. However, HPAI has not been completely eradicated. China’s first HPAI outbreak occurred in 2004. The period from 2004 to 2005 saw concentrated outbreaks of the H5 HPAI subtype, with 46 and 27 HPAI outbreaks in 2004 and 2005, respectively. Since 2006, China has implemented compulsory vaccination of all poultry against the H5 HPAI subtype, in addition to the policy of culling all poultry within 3 km of an epidemic point. Then, there were sporadic outbreaks of the H5 HPAI subtype. Although there were no HPAI outbreaks in chickens from 2009 to 2011, there was an average of three HPAI outbreaks per year from 2006 to 2016. In 2017, China experienced the first and concentrated outbreaks of the H7 HPAI subtype, totaling ten HPAI outbreaks in chickens. Since the autumn of 2017, China has compulsorily vaccinated all poultry with the H5 and H7 subtype vaccines. This intervention reduced the outbreaks of the H5 and H7 HPAI subtypes between 2018 and 2021, with an average of 3 outbreaks per year.

2.2. Theoretical Model and Hypotheses

2.2.1. Impact of the Previous Year’s HPAI on Production Through Costs

In this study, we applied production cost theory to explain how broiler producers adjusted production in response to HPAI outbreaks that occurred in the previous year. The profit function for broiler producers in province $i$ is specified as follows:

$${\pi }_{it}=R_{it}\left(Q_{it}\right)-C_{it}(Q_{it},I_{i,t-1})$$

(1)

where ${\pi }_{it}$ is the farming profit in province $i$ in year $t$ and $R_{it}\left(Q_{it}\right)$ is the income function in province $i$ in year $t$. Let $R_{it}\left(Q_{it}\right)=P_{it}{Q}_{it}$, where $P_{it}$ is the chicken price in province $i$ in year $t$, $P_{it}$ > 0. $C_{it}(Q_{it},I_{i,t-1})$ is the production costs function in province $i$ in year $t$, and $C_{it}\left(Q_{it},I_{i,t-1}\right)$ = $C_0\left(Q_{it}\right)+C_1\left(I_{i,t-1}\right)$, where $C_0\left(Q_{it}\right)$ is the production costs related to chicken production $Q_{it}$ in province $i$ in year $t$, and $C_0\left(Q_{it}\right)$ $=$ $\alpha {Q_{it}}^2+b{Q}_{it}+c$. $\alpha , b, \mathrm{a}\mathrm{n}\mathrm{d} c$ are constants. $C_1\left(I_{i,t-1}\right)$ is the production costs related to the number of HPAI outbreaks in chickens in province $i$ in year $t$, and $C_1\left(I_{i,t-1}\right)={k{I}_{i,t-1}}^d$, $I_{i,t-1}$ ≥ 0, $k$ > 0, and $d$ > 0. The partial derivative of $C_{it}$ with respect to $I_{i,t-1}$ is as follows:

$${\displaystyle \frac{\partial C_{it}}{\partial I_{i,t-1}}}={\displaystyle \frac{\partial C_1}{\partial I_{i,t-1}}}={kd{I}_{i,t-1}}^{d-1} >0$$

(2)

A first-order condition in which broiler producers seek to maximize profit ${\pi }_{it}$ is specified as follows:

$${\displaystyle \frac{d{\pi }_{it}}{d{Q}_{it}}}=P_{it}-2a{Q}_{it}-b=0$$

(3)

$$Q_{it}={\displaystyle \frac{P_{it}-b}{2a}}$$

(4)

The partial derivative of $Q_{it}$ with respect to $C_{it}$ is as follows:

$${\displaystyle \frac{\partial Q_{it}}{\partial C_{it}}}={\displaystyle \frac{\partial Q_{it}}{\partial [\alpha {Q_{it}}^2+b{Q}_{it}+c+C_1(I_{i,t-1}\left)\right]}}={\displaystyle \frac{\partial Q_{it}}{\partial (\alpha {Q_{it}}^2+b{Q}_{it}+c)}}+{\displaystyle \frac{\partial Q_{it}}{\partial C_1\left(I_{i,t-1}\right)}}$$

(5)

where ${\displaystyle \frac{\partial Q_{it}}{\partial (\alpha {Q_{it}}^2+b{Q}_{it}+c)}}=0$, and ${\displaystyle \frac{\partial Q_{it}}{\partial C_1\left(I_{i,t-1}\right)}}=-{\displaystyle \frac{1}{2d}}{\displaystyle \frac{\partial C_1}{\partial I_{i,t-1}}} <0$.

The partial derivative of $Q_{it}$ with respect to $I_{i,t-1}$ is as follows:

$${\displaystyle \frac{\partial Q_{it}}{\partial I_{i,t-1}}}={\displaystyle \frac{\partial Q_{it}}{\partial C_{it}}}{\displaystyle \frac{\partial C_{it}}{\partial I_{i,t-1}}}<0$$

(6)

Specifically, the increase in the number of HPAI outbreaks in province $i$ in year $t-1$ would lead to a decrease in chicken production in province $i$ in year $t$ through an increase in production costs in province $i$ in year $t$. Therefore, Hypothesis 1 is proposed as follows:

More HPAI outbreaks in a province in the previous year lead to greater increases in production costs for that province in the previous year and larger reductions in chicken production in the current year.

2.2.2. Impact of the Previous Year’s HPAI on Epidemic Prevention Costs

The risk of HPAI outbreaks function is specified as follows:

$${\theta }_{it}={\theta }_0-\alpha C_{it}^p+\beta H_{i,t-1}$$

(7)

where ${\theta }_{it}$ is the risk of HPAI outbreaks in province $i$ in year $t$, ${\theta }_0$ is the base risk, $C_{it}^p$ is the epidemic prevention cost in province $i$ in year $t$, and $H_{i,t-1}$ is the number of HPAI outbreaks in year $t-1$. $\alpha >0;$ $\beta >0.$

The cost associated with the risk of HPAI outbreaks in province $i$ in year $t$ is as follows:

$${TC}_{it}=C_{it}^p+{\theta }_{it}·L_{it}$$

(8)

where ${TC}_{it}$ is the costs of HPAI outbreaks in province $i$ in year $t$ and $L_{it}$ is the potential losses due to HPAI outbreaks in province $i$ in year $t;L_{it}>0$.

The partial derivative of ${TC}_{it}$ with respect to $C_{it}^p$ is as follows:

$${\displaystyle \frac{\partial {TC}_{it}}{\partial C_{it}^p}}=1-\alpha L_{it}=0$$

(9)

The first-order total differentiation is as follows:

$$\alpha {\mathrm{d}L}_{it}+L_{it}\mathrm{d}\alpha =0$$

(10)

The derivative of $C_{it}^p$ with respect to $H_{i,t-1}$ is as follows:

$${\displaystyle \frac{d{C}_{it}^p}{d{H}_{i,t-1}}}=-{\displaystyle \frac{\left[\frac{{\partial }^2{\theta }_{i,t}}{\partial C_{it}^p{H}_{i,t-1}}\right]}{\left[\frac{{\partial }^2{\theta }_{it}}{{\left(\partial C_{it}^p\right)}^2}\right]}}={\displaystyle \frac{\alpha \frac{\partial L_{it}}{\partial H_{i,t-1}}}{L_{it}\frac{\partial \alpha }{\partial C_{it}^P}}}$$

(11)

An HPAI outbreaks in year $t-1$ could increase potential losses in year $t,$ and the effectiveness of epidemic prevention could decrease due to the diminishing effect of epidemic prevention costs invested, so ${\displaystyle \frac{\partial L_{it}}{\partial H_{i,t-1}}}$> 0 and ${\displaystyle \frac{\partial \alpha }{\partial C_{it}^P}}>0$, plus $\alpha >0$, $L_{it}>0$. The following can be derived.

$${\displaystyle \frac{d{C}_{it}^p}{d{H}_{i,t-1}}}>0$$

(12)

Specifically, an increase in the number of HPAI outbreaks in province $i$ in year $t-1$ is positively correlated with the epidemic prevention costs in year $t$. Therefore, Hypothesis 2 is proposed as follows:

More HPAI outbreaks in a province in the previous year will lead to a greater increase in epidemic prevention costs for the province in the current year.

2.2.3. Impact of the Previous Year’s HPAI on Chick Costs

The chicks supply function for province $i$ in year $t-1$ is specified as follows:

$$Q_{it}=\alpha S_{it}=\alpha \left(S_{i,t-1}-\gamma H_{i,t-1}\right)$$

(13)

where $Q_{it}$ is the chick supply in province $i$ in year $t$; $\alpha$ is the broiler breeding rate; $S_{it}$ represents the number of breeding broilers in stock in province $i$ in year $t$, $\alpha >0$; $S_{i,t-1}$ represents the number of breeding broilers in stock in province $i$ in year $t-1$; $H_{t-1}$ is the number of HPAI outbreaks in year $t-1$; and $\gamma$ is culling rate, $\gamma >0$.

The equilibrium of chick supply and demand is as follows:

$$Q_{it}=D_{it}\left(P_{it}\right)=\alpha \left(S_{i,t-1}-\gamma H_{i,t-1}\right)$$

(14)

where $D_{it}$ is the chick demand in province $i$ in year $t$ and $P_{it}$ is the unit price of chicks in province $i$ in year $t$.

The partial derivative of both sides of Equation (14) with respect to $H_{t-1}$ is as follows:

$${\displaystyle \frac{\partial P_{it}}{\partial H_{i,t-1}}}={\displaystyle \frac{-\alpha \gamma }{\frac{\partial D_{it}}{\partial P_{it}}}}$$

(15)

where $D_{it}\left(P_{it}\right)$ meets the condition where demand decreases as price increases (i.e., ${\displaystyle \frac{\partial D_{it}}{\partial P_{it}}}<0$)

$${\displaystyle \frac{\partial P_{it}}{\partial H_{i,t-1}}}>0$$

(16)

Suppose the chick cost is specified as follows:

$$C_{it}=P_{it}\times Q_{it}$$

(17)

where $C_{it}$ is the chick cost. The partial derivative of both sides of Equation (17) with respect to $H_{t-1}$ is as follows:

$${\displaystyle \frac{\partial C_{it}}{\partial H_{i,t-1}}}={\displaystyle \frac{\partial P_{it}}{\partial H_{i,t-1}}}{Q}_t+{\displaystyle \frac{\partial Q_{it}}{\partial H_{i,t-1}}}{P}_{it}$$

(18)

where ${\displaystyle \frac{\partial P_{it}}{\partial H_{i,t-1}}}{Q}_{it}$ > 0. It is inferred that ${\displaystyle \frac{\partial Q_{it}}{\partial H_{i,t-1}}}{P}_{it}<0$ based on Equation (14) with respect to $H_{t-1}$ for partial differentiation and $P_{it}>0$.

Typically, the production recovery cycle for broiler breeder flocks is prolonged, resulting in a price-inelastic supply of chicks. When the price increase exceeds the decrease in production, ${\displaystyle \frac{\partial P_{it}}{\partial H_{i,t-1}}}{Q}_{it}>{\displaystyle \frac{\partial Q_{it}}{\partial H_{i,t-1}}}{P}_{it}$; therefore, ${\displaystyle \frac{\partial C_{it}}{\partial H_{i,t-1}}}>0$. The increase in the number of HPAI outbreaks in province $i$ in year $t-1$ is positively correlated with the chick costs in year $t$. Therefore, Hypothesis 3 is proposed as follows:

More HPAI outbreaks in a province in the previous year lead to a greater increase in chick costs for the province in the current year.

2.3. Estimation Strategy

We employed two regression models to examine the lagged effect of HPAI outbreaks on chicken production. The dynamic panel data model used to estimate the impact of the previous year’s HPAI outbreaks on production costs is specified as follows:

$${lnprocost}_{it}={\alpha }_0+{\alpha }_1{hpai}_{it-1}+\vartheta {Controls}_{it}+{\phi }_i+{\tau }_t+{ϵ}_{it}$$

(19)

where ${lnprocost}_{it}$ is the natural logarithm of production costs in province $i$ in year $t$, which would be replaced by the epidemic prevention costs and chick costs at the stage of the mechanism analysis. ${hpai}_{it-1}$ is the number of HPAI outbreaks in province $i$ in year $t-1$. ${Controls}_{it}$ are control variables, including the natural logarithm of production costs in province $i$ in year $t-1$, the number of HPAI outbreaks in chickens in province $i$ in year $t$, the natural logarithm of chicken production in province $i$ in year $t$, the natural logarithm of the square term of chicken production in province $i$ in year $t$, the natural logarithm of labor inputs in province $i$ in year $t$, the natural logarithm of feed inputs in province $i$ in year $t$, the natural logarithm of broiler total factor productivity in province $i$ in year $t$, and the proportion of scale broiler producers to the total number of broiler producers in province $i$ in year $t$. ${\phi }_i$ is the province fixed effect, ${\tau }_t$ is the year fixed effect, ${ϵ}_{it}$ is the random error term, ${\alpha }_0$ is the intercept term, ${\alpha }_1$ is the elasticity of the number of HPAI outbreaks in chickens on production costs, and $\vartheta$ is the coefficients of the control variables vector.

The dynamic panel data regression model for the impact of production costs on chicken production is specified as follows:

$${lnchickenpro}_{it}={\beta }_0+{\beta }_1{lnprocost}_{it}+\theta {Controls}_{it}+{\mu }_i+{\delta }_t+{\epsilon }_{it}$$

(20)

where ${lnchickenpro}_{it}$ is the natural logarithm of chicken production in province $i$ in year $t$. ${lnprocost}_{it}$ is the natural logarithm of production costs in province $i$ in year $t$. ${Controls}_{it}$ are the control variables, including the natural logarithm of chicken production in province $i$ in year $t-1$, the number of HPAI outbreaks in chickens in province $i$ in year $t$, the number of HPAI outbreaks in chickens in province $i$ in year $t-1$, the natural logarithm of the chicken price in province $i$ in year $t$, the natural logarithm of the pork price in province $i$ in year $t$, the natural logarithm of labor inputs in province $i$ in year $t$, the natural logarithm of feed inputs in province $i$ in year $t$, and the natural logarithm of the number of broiler producers in province $i$ in year $t$. ${\mu }_i$ is the province fixed effect, ${\delta }_t$ is the year fixed effect, ${\epsilon }_{it}$ is the random error term, ${\beta }_0$ is the intercept term, ${\alpha }_1$ is the elasticity of the number of HPAI outbreaks in chickens on production costs, $\vartheta$ is the coefficients of the control variables vector, ${\beta }_1$ is the elasticity of the production costs on chicken production, and $\theta$ is the coefficient of the control variables vector.

$${\displaystyle \frac{\partial {lnchickenpro}_{it}}{\partial {hpai}_{it-1}}}={\displaystyle \frac{\partial {chickenproduction}_{it}}{\partial {lnprocost}_{it}}}\times {\displaystyle \frac{\partial {lnprocost}_{it}}{\partial {hpai}_{it-1}}}={\alpha }_1\ast {\beta }_1$$

(21)

where ${\alpha }_1\ast {\beta }_1$ is the elasticity of the number of HPAI outbreaks on chicken production. ${\alpha }_1$ and ${\beta }_1$ are obtained using GMM estimation.

2.4. Data Collection

The dependent variable, annual chicken production per province, was calculated as 70% of the annual poultry meat production reported in the China Animal Husbandry and Veterinary Yearbook [8].

The independent variables included the previous year’s number of HPAI outbreaks in chickens per province and the annual production costs per province. Given that HPAI outbreaks in China began in 2004, and the Veterinary Bulletin issued by China’s Ministry of Agriculture and Rural Affairs ceased publishing outbreak data after 2021, we constructed an annual provincial panel covering 30 provinces (except Tibet) for the 2004–2021 period. The annual number of HPAI outbreaks in chickens for each province was obtained by summing the monthly HPAI outbreaks reported in the Veterinary Bulletin. Although official reports did not specify whether HPAI-infected birds were broilers or layers, culled layers were ultimately processed for meat; thus, all HPAI outbreaks in chickens were treated uniformly. The annual provincial number was then lagged by one year to yield the previous year’s outbreak. Additionally, annual provincial production costs were computed as the simple arithmetic mean of the cost per 100 broilers reported for small (300–1000 birds/year), medium (1000–10,000 birds/year), and large (>10,000 birds/year) farms, drawn from Compilation of Cost–benefit Data on National Agricultural Products, published by the China Development and Reform Commission.

After interpolating sporadic missing values via Auto-Regressive Moving Average (ARMA) [24], we imputed the unpublished production cost series for Jiangxi, Shaanxi, Qinghai, Chongqing, Sichuan, and Xinjiang (2004–2021) using cluster analysis, assuming that provinces with similar attribute vectors form homogeneous groups while differing markedly across groups [25]. Scale economy theory predicts that the average cost decreases as output expands [26], so production costs were expected to differ across farm sizes. Because the China Ministry of Agriculture and Rural Affairs defined farms with ≥50 000 broilers per year as scale broiler farms, we used the ratio of such farms to total broiler farms in each province, obtained from the China Animal Husbandry and Veterinary Yearbook, as our measure of scale. A higher proportion implied larger average farm size and, consequently, lower unit costs, giving both production costs and the scale proportion a clustered provincial pattern. We therefore applied the incremental sum-of-squares method to group the 30 provinces into five clusters based on their scale-broiler shares. For Jiangxi, Shanxi, Qinghai, Chongqing, Sichuan, and Xinjiang, we imputed missing annual production cost data as the simple arithmetic mean of the two nearest provinces within the same cluster. All provincial cost series were then deflated by the price index of agricultural means of production for animal products or feeding animals, sourced from the China Statistical Yearbook.

The control variables comprised annual provincial chicken and pork prices, taken from the China Animal Husbandry and Veterinary Yearbook. Each annual price was calculated as the simple arithmetic mean of the 12 monthly provincial prices, then deflated by the corresponding consumer price index of poultry meat and pork. Additional control variables, such as annual epidemic prevention costs, broiler chick costs, broiler feed inputs, and broiler labor inputs, were compiled with the same procedure used for production costs. The values were drawn from the Compilation of Cost–benefit Data on National Agricultural Products. Advances in broiler farming technology were measured by total factor productivity (TFP), which was estimated using the Levinsohn–Petrin method. The TFP calculation incorporated data on main product output, material and service costs, labor costs, feed costs, and land costs, sourced from the Compilation of Cost–benefit Data on National Agricultural Products.

3. Results

3.1. Descriptive Statistics

Table 1. Definition and summary statistics.

Variable Types	Variables	Description	Obs.	Mean	Std. Dev	Min	Max
Dependent variables	chickenpro	Annual chicken production in each province (10,000 tons)	540	41.871	43.406	0.210	255.430
	procost	Annual 100 broiler real production costs in each province (CNY)	540	6655.088	13,940.930	813.850	136,004.50
	broilerepi precost	Annual 100 broiler real epidemic prevention costs in each province (CNY)	540	119.800	250.968	14.650	2448.332
	broilerchickcost	Annual 100 broiler real chick costs in each province (CNY)	540	983.623	2060.470	120.290	20,101.440
Independent variables	chickenhpai	The number of annual HPAI outbreaks in chickens in each province	540	0.248	0.929	0	10
	lchickenhpai	The number of annual HPAI outbreaks in chickens in the previous year in each province	510	0.263	0.954	0	10
	tfp	Annual broiler total factor productivity in each province	540	88.362	17.592	51.925	166.310
	laborinput	Annual 100 broiler labor inputs in each province (days)	540	3.758	1.9130	0.590	14.230
	feedinput	Annual 100 broiler feed inputs in each province (kilograms)	540	354.252	76.285	173.700	596.250
	scalepercent	The proportion of annual scale broiler producers to the total number of broiler producers (%)	540	1.218	4.721	0	58.540
	chickenprice	Annual real chicken price in each province in the previous year (CNY per kilogram)	540	19.561	6.191	7.872	34.662
	porkprice	Annual real pork price in the previous year (CNY per kilogram)	540	27.178	10.180	11.548	60.413
	farmers	The number of annual broiler producers in each province	540	628,266.500	1,013,842	0	8,523,092

Sources: China Statistical Yearbook (2005–2022); Compilation of Cost–benefit Data on National Agricultural Products (2005–2022); China Animal Husbandry and Veterinary Yearbook (2005–2022); Veterinary Bulletin of China’s Ministry of Agriculture and Rural Affairs (2004–2021).

presents summary statistics for 540 province-year observations across China’s 30 provinces from 2004 to 2021. The average annual provincial chicken production reached 418,710 tons, ranging from 2100 to 25.54 million tons. Production costs averaged CNY 6655.088 per 100 broilers, varying from CNY 813.850 to CNY 136,004.50. Epidemic prevention costs averaged CNY 119.800, while chick costs averaged CNY 983.623 per 100 broilers. The average number of HPAI outbreaks per province in the previous year was 0.263. The mean total factor productivity for broilers across provinces was 88.362. Production inputs per 100 broilers averaged 3.758 labor days and 354.252 kg of feed. Provincial annual prices averaged CNY 19.561 per kilogram for chicken and CNY 27.178 per kilogram for pork.

3.2. Baseline Regression Analysis

In

Table 2. The results of the baseline regression.

Variables	(1)	(2)	(3)	(4)
	Lnprocost	Lnchickenpro	Lnprocost	Lnchickenpro
lnprocost	—	−0.111 *** (0.043)	—	−0.103 *** (0.038)
lnlchickenhpai	0.426 *** (0.161)	—	0.372 *** (0.143)	—
Controls	N	N	Y	Y
Province FE	Y	Y	Y	Y
Year FE	Y	Y	Y	Y
AR(1) test	−3.26 (0.001)	−3.16 (0.002)	−1.72 (0.025)	−2.80 (0.005)
AR(2) test	0.92 (0.202)	0.85 (0.398)	1.04 (0.300)	−0.25 (0.801)
Hansen test	0.68 (0.774)	7.98 (0.890)	0.58 (0.690)	2.57 (0.965)
Obs.	510	510	510	510

Notes: The numbers in parentheses are robust standard errors; *** represent significance at 1%.

, Columns (1) and (2) (without control variables) show that the number of HPAI outbreaks in the previous year had a positive effect on current-year broiler production costs, while current-year production costs negatively affected current-year chicken production. Columns (3) and (4) (with control variables) further confirm these relationships, with both the lagged HPAI outbreaks and current-year production costs being statistically significant at the 1% level, namely, HPAI outbreaks in the previous year significantly reduced chicken production by increasing production costs in the current year. Holding all other conditions constant, a 1% increase in the number of HPAI outbreaks in a province led to a 0.372% rise in production costs per 100 broilers the following year, which, in turn, resulted in a 0.038% (−0.103 * 0.372%) decline in chicken production. Thus, the empirical results support Hypothesis 1.

3.3. Analysis of the Robustness Test

Endogeneity issues may arise due to reverse causality. Specifically, HPAI outbreaks can reduce consumer demand for poultry, prompting producers to cut production to lower costs [27]. In this study, we addressed the two-way causality between chicken production and production costs using a two-stage least squares regression. As instrumental variables for production costs, we utilized the annual provincial broiler feed price and broiler chick price, obtained from the China Animal Husbandry and Veterinary Yearbook. As is shown in Columns (1) and (3) of

Table 3. The results of the two-stage least squares regression.

Variables	(1)	(2)	(3)	(4)
	Lnprocost	Lnchickenpro	Lnprocost	Lnchickenpro
	First-Stage	Second-Stage	First-Stage	Second-Stage
Lnbroilerfeedprice	0.864 *** (0.013)	—	—	—
Lnprocost	—	−0.050 ** (0.024)	—	−0.046 ** (0.023)
Lnchickprice	—	—	0.805 *** (0.020)	—
Controls	Y	Y	Y	Y
Province FE	Y	Y	Y	Y
Year FE	Y	Y	Y	Y
Obs.	510	510	510	510
underidentification test	6.133 ***		5.685 **
weak identification test	445.372		157.255

Notes: The numbers in parentheses are robust standard errors; *** and ** represent significance at 1% and 5%, respectively.

, the first-stage results confirm that both instrumental variables were statistically significant at the 1% level. The Kleibergen–Paap Wald F-statistics exceeded the critical threshold of 10, strongly rejecting the null hypothesis of underidentification and ruling out concerns of weak instruments, which suggested that the selected instrumental variables were empirically valid. Columns (2) and (4) of Table 3 demonstrate that production costs had a statistically significant negative impact on chicken production in the second-stage regression, consistent with the baseline regression result.

The dependent variable was changed from the natural log of total chicken production to the natural log of broiler outputs from the China Statistical Yearbook. Column (1) in

Table 4. The results of the robustness test.

Variables	(1)	(2)		(3)		(4)		(5)
	Lnbroileroutput	Lnprocost	Lnprocost	Lnprocost	Lnchickenpro	Lnprocost	Lnchickenpro	Lnprocost	Lnchickenpro
lnlchickenhpai	—	—	—	0.234 ** (0.108)	—	0.004 ** (0.002)	—	0.142 ** (0.069)	—
lnprocost	−0.234 ** (0.114)	—	—	—	−0.101 * (0.057)	—	−0.013 ** (0.006)	—	−0.081 ** (0.037)
lnlchickenhpaiornot	—	0.234 ** (0.108)	—	—	—	—	—	—	—
lnchickenhpaiseverity	—	—	0.562 ** (0.280)	—	—	—	—	—	—
lnlaborinput^2	—	—	—	−0.103 ** (0.051)	−0.071 * (0.044)	—	—	—	—
lnfeedinput^2	—	—	—	3.499 * (1.891)	0.088 ** (0.040)	—	—	—	—
Controls	Y	Y	Y	Y	Y	Y	Y	Y	Y
Province FE	Y	Y	Y	Y	Y	Y	Y	Y	Y
Year FE	Y	Y	Y	Y	Y	Y	Y	Y	Y
AR(1) test	−1.70 (0.090)	−2.00 (0.046)	−1.87 (0.061)	−1.71 (0.087)	−2.51 (0.012)	—	—	−2.33 (0.020)	−2.80 (0.005)
AR(2) test	1.60 (0.111)	1.17 (0.241)	−0.40 (0.686)	0.88 (0.377)	−0.23 (0.819)	—	—	0.59 (0.554)	0.36 (0.715)
Hansen test	5.37 (0.890)	1.71 (0.887)	1.89 (0.902)	0.14 (0.576)	1.51 (0.755)	—	—	2.15 (0.867)	2.08 (0.820)
Obs.	510	510	510	510	510	510	510	480	480

Notes: The numbers in parentheses are robust standard errors; ** and * represent significance at 5%, and 10%, respectively.

shows that production costs had a significantly negative impact on broiler outputs.

The independent variable was modified from the natural log of HPAI outbreak counts to two alternative specifications: a binary variable indicating outbreaks and an ordinal measure of outbreak severity, which was constructed with the number of chickens infected with HPAI, the number of chickens that died from HPAI, and the number of chickens culled due to HPAI from Veterinary Bulletin using the entropy weight method. Column (2) in Table 4 demonstrates that both the HPAI outbreaks in the previous year and the severity of HPAI outbreaks had statistically significant positive effects on chicken production.

To capture potential diminishing or increasing returns to scale, we augmented the baseline model with quadratic terms for labor and feed inputs, consistent with production function theory [26]. Column (3) in Table 4 indicates that the number of HPAI outbreaks in the previous year retained a statistically significant positive effect, while production costs exhibited a significant negative effect.

To account for potential simultaneity between HPAI outbreaks, production costs, and chicken production, we employed a three-stage least squares estimator. This approach addressed the limitations of single-equation methods by modeling the system of equations jointly, thereby correcting for correlated disturbance terms across equations. Column (4) in Table 4 indicates that lagged HPAI outbreaks had a significant positive effect on chicken production, while production costs showed a significant negative effect.

We excluded 2004–2005 data from re-estimation due to their disproportionately high share of HPAI outbreaks (54.48% of total cases). Column (5) in Table 4 demonstrates that the number of HPAI outbreaks in the previous year had a statistically significant positive effect on production, while production costs exhibited a significant negative effect.

3.4. Heterogeneity Analysis

Given variations in broiler industry development, we conducted a heterogeneity analysis to examine how the impact of previous-year HPAI outbreaks on chicken production differed by the total production volume, share of large-scale producers, and yellow-feathered broiler specialization.

We stratified the provinces into two groups based on whether their chicken production exceeded the sample median and then performed separate regression analyses for each subgroup. Column (1) in

Table 5. The result of the heterogeneity analysis.

Variables	(1)				(2)				(3)
	High-Yield		Low-Yield		High Proportion of Scale Broiler Producers		Low Proportion of Scale Broiler Producers		Yellow-Feathered Broilers		Non-Yellow-Feathered Broilers
	Lnprocost	Lnchickenpro	Lnprocost	Lnchickenpro	Lnprocost	Lnchickenpro	Lnprocost	Lnchickenpro	Lnprocost	Lnchickenpro	Lnprocost	Lnchickenpro
lnlchickenhpai	0.152 ** (0.076)	—	0.111 * (0.062)	—	0.115 *** (0.046)	—	0.145 ** (0.073)	—	0.177 *** (0.069)	—	0.268 ** (0.135)	—
Lnprocost	—	−0.378 ** (0.186)	—	−0.168 ** (0.084)	—	−0.332 ** (0.166)	—	−0.109 ** (0.185)	—	−0.369 *** (0.144)	—	−0.160 ** (0.080)
Controls	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y
Province FE	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y
Year FE	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y
AR(1) test	−1.79 (0.074)	−2.18 (0.029)	−1.89 (0.059)	−2.19 (0.028)	−1.78 (0.075)	−2.53 (0.011)	−1.89 (0.058)	−2.46 (0.014)	−1.88 (0.060)	−1.82 (0.069)	−0.40 (0.062)	−2.86 (0.004)
AR(2) test	−1.36 (0.175)	−1.62 (0.104)	1.01 (0.313)	0.13 (0.899)	−1.45 (0.148)	−0.72 (0.471)	−1.73 (0.784)	0.69 (0.489)	−0.41 (0.679)	0.09 (0.865)	−0.40 (0.062)	0.61 (0.545)
Hansen test	0.12 (0.768)	0.12 (0.876)	0.08 (0.740)	0.07 (0.867)	0.12 (0.768)	0.24 (0.649)	63.25 (0.432)	0.15 (0.865)	0.02 (0.689)	0.07 (0.520)	0.03 (0.768)	0.09 (0.997)
Obs.	258	258	252	252	260	260	250	250	106	106	404	404
Group difference test	0.013 **				0.054 **				0.064 **

Notes: The numbers in parentheses are robust standard errors; ***, **, and * represent significance at 1%, 5%, and 10%, respectively.

indicates that a 1% increase in previous-year HPAI outbreaks reduced chicken production by 0.019% (−0.168 × 0.111%) in low-yield provinces and by 0.057% (−0.378 × 0.152%) in high-yield provinces.

We divided the samples into two groups based on whether the proportion of scale broiler producers in each province was greater than the sample median. Then, we conducted regression analysis for each group. Column (2) in Table 5 demonstrates that a 1% increase in previous-year HPAI outbreaks reduced chicken production by 0.016% (−0.109 × 0.145%) in provinces with below-median scale production and by 0.038% (−0.332 × 0.115%) in provinces with above-median scale production.

We stratified the provinces into yellow-feathered broiler production regions and non-yellow-feathered regions, performing separate regression analyses for each group. Column (3) in Table 5 demonstrates that a 1% increase in prior-year HPAI outbreaks led to a 0.043% (−0.160 × 0.268%) production decline in non-yellow-feathered broiler provinces and a 0.065% (−0.369 × 0.177%) production decline in yellow-feathered broiler provinces.

3.5. Mechanism Analysis

We recalculated production costs per 100 broilers using epidemic prevention costs and chick costs from the Compilation of Cost–benefit Data on National Agricultural Products. Column (1) in

Table 6. The result of the mechanism analysis.

Variables	(1)	(2)
	Lnlrbroilerepiprecost	Lnlrbroilerchickcost
lnlchickenhpai	0.148 ** (0.057)	0.221 *** (0.088)
lnlrbroilerepiprecost	—	—
lnlrbroilerchickcost	—	—
Controls	Y	Y
Province FE	Y	Y
Year FE	Y	Y
AR(1) test	−2.00 (0.045)	−1.01 (0.073)
AR(2) test	−1.34 (0.289)	−0.69 (0.489)
Hansen test	0.10 (0.767)	0.20 (0.875)
Obs.	510	510

Notes: The numbers in parentheses are robust standard errors; *** and **, represent significance at 1% and 5%, respectively.

demonstrates that higher epidemic prevention costs per 100 broilers significantly increased with previous-year HPAI outbreaks. Specifically, a 1% increase in outbreaks raised prevention costs by 0.148%, supporting Hypothesis 2. Column (2) in Table 6 reveals that broiler chick costs per 100 broilers significantly increased with previous-year HPAI outbreaks. A 1% rise in outbreaks led to a 0.221% increase in chick costs, supporting Hypothesis 3. Furthermore, our analysis revealed that HPAI outbreaks in the previous year caused a more substantial increase in chick costs compared to epidemic prevention costs.

4. Discussion

While prior research examined HPAI’s economic consequences through risk aversion and market price transmission, this study identified a novel cost channel from lagged HPAI outbreaks to reduced chicken production. Our production cost perspective complemented existing frameworks and provided new insights into the epidemic’s persistent economic effects.

While existing research examined HPAI’s economic impacts at district and industry levels, this study revealed substantial provincial heterogeneity in costs and production responses to previous-year outbreaks. Our findings advanced the understanding of HPAI’s spatial economic dimensions, enabling more targeted provincial policy interventions. Practically, we identified epidemic prevention costs and broiler chick costs as critical drivers of post-outbreak production cost volatility, highlighting the need for enhanced management along the poultry supply chain.

Based on the results of our study, we suggest several promising directions for future research. First, while we utilized provincial-level annual data, farm-level microdata could provide more granular insights and help mitigate potential aggregation biases. Expanding the sample to include finer-scale operational data would significantly enhance the analysis of HPAI’s economic impacts. Second, China’s broiler breeding cycles range from 32 to 155 days, with 2–11 batches produced annually. While our annual data captured producers’ aggregate adjustments, they could not identify batch-specific epidemic effects. Future studies should employ monthly data to assess HPAI’s intra-year production impacts. Third, although we analyzed short-term (one-year) cost-mediated effects, long-term impacts may differ as producers adjust all factor inputs. Extending the analysis to multi-year lags would clarify HPAI’s dynamic economic consequences.

5. Conclusions

This study yielded three key findings. First, previous-year HPAI outbreaks significantly increased current-year production costs, ultimately reducing chicken production. Second, production costs increased and chicken production reductions were most pronounced in provinces with intensive chicken production, predominant large-scale operations, and specialization in yellow-feathered broilers. Third, previous-year HPAI outbreaks increased production costs through elevated epidemic prevention costs and higher broiler chick costs.

Our findings suggest three key policy implications. First, provinces characterized by intensive chicken production, high concentrations of scale operations, and yellow-feathered broiler specialization should receive prioritized support to strengthen disease surveillance systems and emergency response protocols, thereby enhancing regional production stability. Second, targeted subsidy programs should be established to support broiler producers in HPAI prevention, alleviating production cost pressures and stabilizing market supply. Third, a broiler chick price stabilization mechanism should be implemented to mitigate excessive price surges during HPAI outbreaks, thereby helping producers manage costs and facilitating faster production recovery.