Impact of livestream selling on suppliers’ operational performance: an empirical study on an online retailer

Abstract

Purpose

Livestream selling is becoming an increasingly popular practice adopted by online retailers to develop a consumer-centric supply chain (CCSC). It improves consumer experience by integrating chat, watch and purchase functions, while also altering consumer behaviors by increasing impulse purchases. Online retailers’ responses to this change potentially impact suppliers’ operational processes. This study aims to empirically examine how livestream selling affects suppliers’ operational performance in terms of lead time and how suppliers’ product variety and order fulfillment capabilities moderate such an impact.

Design/methodology/approach

Using data from a leading online retailer in China, the authors use a least squares model with fixed effects to test the relationships. Both the two-stage instrumental variable model and the two-stage Heckman model are used to address potential endogeneity in this study.

Findings

The findings show that retailers’ usage of livestream selling can increase suppliers’ lead time. Furthermore, the negative impact is enhanced when a supplier has a higher level of product variety or a weaker order fulfillment capability.

Originality/value

This study explores how livestream selling alters consumer behavior, adversely affecting upstream suppliers’ operational performance. It underscores the need for a CCSC approach across all tiers, not just those closest to consumers. To achieve this, the research suggests that suppliers must align their capabilities with retailers’ consumer-centric practices to develop a CCSC, particularly by improving order fulfillment capability and cautiously expanding their product variety in livestream selling. The research further highlights the importance for retailers to consider changes in lead time to enhance the application of traditional inventory theory in the context of livestream selling.

Full text

Translate

Turn on search term navigation

1. Introduction

Social and technological changes in recent years have accelerated the development of consumer-centric supply chain management (CCSCM) which prioritizes the consumer experience in all supply chain decisions (Baldi et al., 2024; Esper et al., 2020). With the advancement of telecommunication and information technology, such as high-efficiency video coding and 5G networks, livestream selling is becoming a novel practice adopted by online retailers to build a CCSC, which has gained significant momentum around the world. Many giant e-commerce retailers, such as Amazon, eBay, Alibaba and JD, have adopted livestream selling over the past few years (Yao et al., 2024). In China, at least 530 million consumers participated in livestream selling in 2023, and sales from livestream selling are expected to reach $843.93 billion by 2025 (Shalabi, 2023). In the United States, the sales from livestream selling amounted to $17 billion in 2022, and it is expected to reach $130.10 billion by 2026 (Chevalier, 2024; Shalabi, 2023). Additionally, livestream selling has begun to emerge more prominently over the past several years in Europe, drawing in a growing number of viewers (Coppola, 2024; Muret, 2022).

In livestream selling, online retailers hire streamers to sell products through real-time interactions with consumers, who can make purchases during live videos. This approach significantly enhances consumer shopping experiences by integrating chat, video and purchase functions into an independent e-commerce platform (Wongkitrungrueng and Assarut, 2020). The chat function enables consumers to interact with both streamers and co-viewers during the live video, allowing them to ask questions and provide immediate and synchronous feedback in the chat room (Sun et al., 2019). The video function allows sellers (or streamers agents) to present detailed product features and respond promptly to consumer inquiries in the chat room. With the purchase function, consumers can buy products and complete transactions directly in the live video without jumping to other sites or applications. These characteristics of livestream selling provide great convenience for consumers and also distinguish it as a channel distinct from traditional e-commerce sales channels, such as TV home shopping and telemarketing. Specifically, livestream selling is two-way communication and provides real-time product displays and immediate purchases. Unlike conventional telemarketing, which relies on phone calls and audio mediums, livestream selling enables consumers to view real-time product displays and make immediate purchases. Compared to traditional TV home shopping, livestream selling enables consumers to give feedback on their real-time needs about the products and make immediate purchases.

Online retailers are expected to be responsive to consumer demand and flexible in adapting to changing consumer preferences (Thomas et al., 2014). Consequently, suppliers’ operational processes are often impacted by retailer-imposed time pressure from order alterations and accelerated decision-making. At the same time, suppliers are required to improve their capabilities to provide more product variety and better order fulfillment to meet consumer expectations. In the context of livestream selling, retailers’ responses to changes in consumer behavior are anticipated to hurt suppliers’ operational performance. Consumers can benefit from livestream selling by gaining a better understanding of products through real-time interactions with streamers, allowing them to purchase suitable products (Cui et al., 2023). However, driven by the informative and persuasive behavior of streamers and sometimes constrained by limited time and product availability, consumers are more prone to impulse buying (Lee et al., 2023; Sun et al., 2021). As a result, consumers’ spontaneous and unanticipated purchase decisions often lead to big spikes in orders within a short time, resulting in suppliers being forced to complete orders under retailers’ strengthened time pressure. Such time pressure can cause suppliers to feel resentful and may result in burnout within operational processes (Thomas et al., 2014). A direct performance outcome is suppliers’ lead time, which refers to the total time taken for each step from order processing to delivery (Wan, 2022). Due to the complexity of operational processes, any unplanned changes can disrupt subsequent operations, leading to increased waiting times and extended processing times, which may ultimately prolong suppliers’ lead times. A recent online survey report indicated that long delivery times for livestream orders are a major concern for consumers (Beijing Consumer Association, 2023). Therefore, the success of livestream selling also requires suppliers to build appropriate capabilities to meet consumers’ changing demands and expectations.

The extant research on livestream selling primarily examined its impact on online retailers’ sales performance (Yu et al., 2023; Chen et al., 2023a) and consumer purchase behaviors (Lee et al., 2023; Feng et al., 2024). Few studies have investigated how livestream selling affects suppliers’ operational performance, particularly concerning lead time. Recently, some studies have explored the impact of livestream selling from the suppliers’ perspective using analytical models and revealed the conditions under which suppliers could benefit from livestream selling (e.g. Chen et al., 2023b; Yang et al., 2023; Zhang and Tang, 2023; Zhang et al., 2024a). There remains a notable gap in empirical research regarding the impact of livestream selling on suppliers’ lead time. Hence, our first research question is: What is the impact of livestream selling on suppliers’ lead time?

Consumers’ preferences for more product variety in livestream selling require suppliers to have a strong order capability to provide a wide product variety. In livestream selling, consumers can quickly access and comprehend information about products through visual and auditory demonstrations presented by streamers, which leads them to favor diverse product options. To meet this demand, retailers will urge their suppliers to improve their capabilities to provide a wide product variety. However, an increase in product variety may lead to more impulse purchases and greater order variability from retailers, which creates additional challenges for suppliers in fulfilling orders on time. Accordingly, we expect that suppliers’ product variety has the potential to influence the impact of livestream selling usages on suppliers’ operational performance. Accordingly, we propose our second question: How does suppliers’ product variety moderate the relationship between livestream selling and suppliers’ lead time?

Furthermore, online retailers’ accelerated procurement and consumer expectations on delivery in livestream selling require suppliers to have a strong order fulfillment capability. Online retailers may modify orders and expedite decision-making processes in response to fluctuating consumer demand in livestream selling while still expecting suppliers to meet their fulfillment obligations. Timely delivery of orders significantly influences customers’ satisfaction and repurchase willingness (Baldi et al., 2024). Suppliers with superior order fulfillment capabilities are anticipated to manage operational processes such as order processing, manufacturing, order packing and sorting, and package delivery more effectively. In this case, they can better tackle the challenges from retailers’ time pressure in livestream selling. Thus, our third question is: How do suppliers’ order fulfillment capabilities moderate the relationship between livestream selling and suppliers’ lead time?

To answer these questions, we collaborated with a leading online retailer in China. We focus on a supply chain consisting of an online retailer, its suppliers and consumers. The retailer sources products directly from its suppliers to its self-owned warehouses and sells them to consumers through livestream selling or traditional online channels. The retailer determines which products are sold through livestream selling based on its marketing strategy. Figure 1 shows the structure of this supply chain. We collected data from the retailer between January and September 2020, including product, order, sales, inventory and delivery information, to explore the research questions.

Our results show that online retailers’ usage of livestream selling can increase suppliers’ lead time. Such adverse effects would be exacerbated when suppliers have a higher level of product variety. However, suppliers with stronger order fulfillment capabilities will be less subjected to this detrimental effect. These findings offer several significant contributions. First, this study contributes to the literature by investigating how changes in consumer behaviors caused by livestream selling impact the operational performance of upstream suppliers, in response to recent calls for research in CCSCM (Esper et al., 2020). Our study reveals that changes in consumer behaviors will adversely affect suppliers’ operational performance when online retailers adopt livestream selling. This highlights that a shift in consumer orientation to approach CCSC should extend to all tiers of the supply chain, rather than focusing solely on those closest to the consumer. Moreover, this study deepens our understanding of the traditional inventory theory by suggesting retailers should take into account the impacts of unpredictable demand change caused by livestream selling on a key factor of the theory, i.e. supplier’s lead time when optimizing the inventory model from a supply chain perspective. Finally, this study contributes to the literature by underlining that suppliers must build commensurate capabilities with retailers’ consumer-centric practices in CCSC. More specifically, we reveal that suppliers should enhance their order fulfillment capabilities while carefully expanding their product variety in livestream selling. This finding provides important managerial implications for online retailers and suppliers to successfully implement livestream selling.

The remaining sections of the paper are organized as follows. Section 2 reviews the relevant literature and develops the research hypotheses. Section 3 presents the research method, the data, and the measurement of variables. In Section 4, we provide estimated models and a summary of the results, followed by endogeneity analysis and robustness checks in Section 5. The final section discusses the contributions and implications of this study.

2. Literature review and hypotheses development

2.1 Livestream selling

Driven by the emergence of new technologies, such as high-efficiency video coding and 5G networks, livestream selling has become increasingly popular to enhance online retailers’ sales and improve consumer experiences (Ki et al., 2024; Zhang et al., 2024b). Many studies have demonstrated the huge benefits of livestream selling. For example, in the context of online international markets, Yu et al. (2023) found that foreign sellers experience increased sales in host markets through livestream selling. Similarly, Chen et al. (2023a) reported that both gross merchandise value (GMV, a widely-used proxy for the total sales of an e-commerce platform) and fan growth rise when more products are sold via live streaming.

The literature also highlights the effects of livestream selling on consumer purchase behaviors, suggesting that consumers participating in livestream selling are more prone to making impulse purchases (Feng et al., 2024; Lee et al., 2023; Sun et al., 2021). This tendency is influenced by the behavior of streamers and the atmosphere of the live video. During a live video, streamers could engage in live chats with consumers and promptly address their questions. Moreover, they can provide abundant product information through detailed explanations and displays. These interactions serve both informative and persuasive roles (Sun et al., 2021), encouraging impulsive buying. Specifically, characteristics of streamers, such as their sales effort, sales ability, linguistic persuasive styles, attractiveness and experience, are considered to have dominant influences on consumer purchase behavior (Chen et al., 2023a; Meng et al., 2021; Pan et al., 2022; Zhou et al., 2022). Additionally, the live video format often creates a tense atmosphere through time pressure and product scarcity related to exclusive discounts, reduced list prices, complimentary gifts, etc., further encouraging consumers’ impulsive purchases (Kong et al., 2023; Lee et al., 2023).

Changes in consumer purchasing behavior significantly influence demand variability, which in turn affects retailers’ procurement and subsequently suppliers’ operational processes. For example, suppliers serving physical stores may lose orders as consumers shift from in-store to online shopping (Hamzaoui et al., 2024). In the context of livestream selling, increased impulsive purchases lead to greater demand variability, e.g. big spikes in orders within a short time. As a result, suppliers are challenged to fulfill orders under retailers’ increased time pressure, potentially generating worse operational performance, particularly in terms of lead times.

Recent studies in operations management have employed analytical models to investigate supply chains that incorporate livestream selling. These studies discussed the conditions under which online retailers and suppliers could or could not benefit from livestream selling. They suggested that suppliers should participate in livestream selling when collaborating with online retailers who have key opinion leaders (Yang et al., 2023) or when the costs associated with livestream selling, production and consumer inconvenience are relatively low (Chen et al., 2023b; Zhang and Tang, 2023; Zhang et al., 2024a).

In sum, current research on the impact of livestream selling on suppliers’ operational performance remains limited, with a notable scarcity of empirical evidence. Extending these studies, we empirically investigated how retailers’ use of livestream selling impacts suppliers’ lead time, as well as how product variety and order fulfillment capabilities influence this relationship. This research provides a more comprehensive understanding of the impact of livestream selling on the entire supply chain and offers valuable insights for retailers and suppliers to optimize their use of livestream selling.

2.2 Impact of livestream selling on supplier’s lead time

Livestream selling introduces greater fluctuations in consumer demand into the supply chain. Driven by the informative and persuasive effects, consumers participating in livestream selling are more susceptible to making spontaneous and unanticipated purchase decisions (Sun et al., 2021). These unforeseen orders, often generated rapidly during a live video, are usually larger than normal orders. Overall, livestream selling boosts impulse buying, significantly increasing demand variability. As a result, retailers will experience considerable fluctuations in consumer demand when they implement livestream selling. This heightened demand variability directly affects retailers’ procurement, resulting in more changes to orders, quicker decision-making and a rise in inaccurate ordering quantities (Hançerlioğulları et al., 2016; Sweeney et al., 2022). This order variability leads retailers to impose more time pressure on suppliers. Such retailer-imposed time pressure can cause suppliers to feel resentful and may result in burnout within operational processes (Thomas et al., 2014), thus requiring more time to fulfill orders.

Moreover, suppliers base their production and distribution decisions on long-term demand forecasts derived from retailer orders. When order variability is high, suppliers may take longer to adapt to changing demand. In addition, significant fluctuations in orders can lead to worse demand forecasting of suppliers, especially when retailer orders are more variable than actual consumer demand (Zhou et al., 2023). Inaccurate forecasts hinder suppliers’ ability to plan and manage effectively, resulting in inefficient use of resources, such as equipment, workforce and transportation. As a result, this inefficiency leads to increased waiting times and disruptions in the production process, prolonging the time it takes for suppliers to fulfill retailer orders. Similarly, ineffective resource allocation in distribution planning can also extend delivery time. Consequently, suppliers will need more time to fulfill retailer orders from livestream selling. Thus, we propose the following:

H1. The use of livestream selling increases suppliers’ lead time.

2.3 Moderating effect of product variety

The product variety is defined as the number of different product variants (DeHoratius and Raman, 2008). Suppliers with a higher level of product variety can provide retailers with more product choices for livestream selling. Typically, online retailers tend to offer as much product variety as possible to satisfy consumers’ demand (Punj, 2011; Sharma et al., 2010). Having more products in livestream selling increases the likelihood that consumers will find what they desire, catering to their variety-seeking traits (Punj, 2011; Sharma et al., 2010; Ton and Raman, 2010). This, in turn, increases the chances of impulse purchases, further contributing to demand variability. In addition, studies have shown that livestream selling has a positive spillover effect on consumers’ willingness to purchase related products, the number of consumers and product sales across other channels for the same supplier (Fan et al., 2022; Yang et al., 2023; Zhao et al., 2022). Hence, retailers are likely to face increased demand variability when suppliers provide a wider product variety, not only from products sold via live stream but also from other channels. When demand variabilities from different products are aggregated, retailers may encounter amplified demand variability. Additionally, maintaining a wider product variety complicates retailers’ order decisions, resulting in more inaccurate order decisions. All these kinds of demand variability increase retailers’ order variability and impose more time pressure on suppliers, thus affecting suppliers’ operational processes. Moreover, suppliers who can offer more product variety usually have more complicated operations and less accurate demand forecasts. Combining these two aspects, suppliers managing more product variety may have more difficulty planning their production, transportation and other operations. As a result, they have the potential to take a longer time to meet the orders from livestream selling. Thus, we propose the following:

H2. The use of livestream selling leads to a longer lead time when suppliers can provide a higher level of product variety.

2.4 Moderating effect of order fulfillment capability

Order fulfillment capability, typically evaluated by the order fill rates, reflects the supplier’s ability to fulfill incoming orders within the expected quantity (Closs et al., 2010). The typical order fulfillment process includes order processing, manufacturing, order packing and sorting, and package delivery. According to Vaidyanathan and Devaraj (2008), order fulfillment is a dynamic capability that encompasses specific organizational and strategic processes that firms can optimize by effectively coordinating and leveraging their available resources (Barua et al., 2004). Therefore, a supplier with a strong order fulfillment capability is always able to efficiently organize its business processes. Suppliers with high order fulfillment capabilities tend to have streamlined order processing, effective inventory management and robust logistics networks, allowing them to respond quickly to changing consumer demand (Tracey et al., 1999). In this case, suppliers can better manage the challenges posed by increased time pressure from retailers, which is driven by amplified consumer demand variability from livestream selling. Thus, we propose the following:

H3. The use of livestream selling leads to a shorter lead time when suppliers have a higher order fulfillment capability.

3. Research methodology

3.1 Data

We gathered data for each product category of each supplier, including product category data, monthly fulfillment data, monthly order quantities and monthly sales data from January 2020 to September 2020. The dataset includes 417 suppliers and 90 product categories. Typically, a supplier provides multiple product categories to the retailer, with each category containing at least one stock-keeping unit (SKU). We also collected registration information of suppliers from Tianyancha.com, an officially licensed enterprise credit agency in China. Combining all these data, we constructed panel data at the supplier-category-month level, including 5,592 observations.

3.2 Measures

3.2.1 Dependent and independent variables

Lead Time. Lead time is a critical aspect of suppliers’ operational performance in livestream selling, which reflects suppliers’ timeliness in fulfilling orders. Following the literature (Wan, 2022), we measured it using the average monthly lead time for different SKUs within a specific product category j from a particular supplier i during a month t, denoted as LeadTime_ijt. Specifically, the lead time for each order of a certain SKU is calculated by the amount of time between the order placed by the online retailer and the order received at the retailer’s warehouse (Blackburn, 2012; de Treville et al., 2014).

Livestream Selling Usage. We used the ratio of GMV generated through livestream selling to the overall GMV of category j from supplier i in month t as a metric to measure the extent of livestream selling usage, which is denoted as LSU_ijt. GMV refers to gross merchandise value, a widely used proxy for total sales on an e-commerce platform.

Product Variety. Product variety is typically defined as the number of distinct product variations (DeHoratius and Raman, 2008). Growing product variety complicates the operational processes, potentially resulting in longer lead times (Sorkun, 2019). Following previous studies (Shou et al., 2017; Sweeney et al., 2022), we measured it using the total number of SKUs for category j from supplier i, which is denoted as ProductVariety_ij.

Order Fill Rate. Order fill rate is defined as the ratio of the amount of orders satisfied by suppliers to the total amount of orders placed by retailers (Qi et al., 2023). It is a well-recognized measure of a supplier’s order fulfillment capabilities. Following the literature (Wan, 2022), we calculated it as the ratio of the completed quantity to the order quantity for category j from supplier i in month t, denoted as OrderFillRate_ijt.

3.2.2 Control variables

Order Quantity. We incorporated order quantity as a control variable due to its positive relationship with lead time. Larger order quantities require suppliers to allocate more time for product preparation and distribution. The variable OrderQuan_ijt is quantified by the total order quantities for category j placed by the online retailer to supplier i in month t.

Firm Size. We included firm size because larger firms may possess more resources to enhance operational capabilities and achieve better IT-enabled performance, potentially resulting in shorter lead times (Chuang et al., 2019). Following previous studies (Qi et al., 2023; Upadhyay and Sriram, 2011), firm size is measured by a supplier’s registered capital (FirmCapital_i), a publicly disclosed indicator used by authorities to evaluate the scale of a firm’s operations.

Firm Age. We also considered firm age (FirmAge_i) as it reflects a company’s resources and experience, which may have an impact on suppliers’ operational performance (Randall et al., 2006). Firm age is operationalized as the number of years since the firm was established.

To control for any latent category-specific time-invariant factors, we incorporated fixed effects (FE) for the categories into our analysis. For example, products of different sizes, like electric toothbrushes versus refrigerators, may have different lead times due to different packing requirements. Moreover, we controlled for month-specific time-variant FE to capture changes in other factors over time (Han et al., 2022). Table 1 reports the descriptive statistics of variables and Table 2 shows correlations between them.

4. Analysis and results

4.1 Econometric models

We applied a FE model to estimate coefficients in each equation. We chose this approach because it is a widely used method in panel data analysis to mitigate the endogeneity problems caused by both observable and unobservable heterogeneity (Xiao et al., 2023). Numerous studies in supply chain management and livestream selling have applied the FE model, e.g. Xiao et al. (2023) and Wan (2022). Inspired by these studies, we integrated the FE model with least squares regression to construct our econometric model. We first developed Model 1 in Equation (1) to investigate the impact of livestream selling on suppliers’ lead time. The estimated model is specified as follows:(1) $L e a d T i m e_{i j t} = α_{0} + α_{1} L S U_{i j t} + C o n t r o l s + μ_{j} + η_{t} + ϕ_{i j t},$ where the subscript $i$ refers to the supplier, $j$ represents the category, and $t$ denotes the month. The $α_{1}$ accounts for the impact of livestream selling on suppliers’ lead time. $α_{0}$ is the intercept, $μ_{j}$ is the category-level FE, $η_{t}$ is month-level FE, and $ϕ_{i j t}$ indicates the error term. Controls include order quantities, firm size and firm age, which are used to control category-specific and firm-specific characteristics on the lead time. Specifically, FE are included via dummy variables of categories and months. The robust standard errors for each model are clustered based on the combination of the supplier and the category.

To estimate the moderating effects of product variety and order fulfillment capability, we included interaction terms in Model 2 and Model 3, as outlined in Equations (2) and (3). The estimated models are specified as follows:(2) $L e a d T i m e_{i j t} = β_{0} + β_{1} L S U_{i j t} + β_{2} L S U_{i j t} \times P r o d u c t V a r i e t y_{i j} + β_{3} P r o d u c t V a r i e t y_{i j} + C o n t r o l s + μ_{j} + η_{t} + ϕ_{i j t},$ (3) $L e a d T i m e_{i j t} = γ_{0} + γ_{1} L S U_{i j t} + γ_{2} L S U_{i j t} \times O r d e r F i l l R a t e_{i j t} + γ_{3} O r d e r F i l l R a t e_{i j t} + C o n t r o l s + μ_{j} + η_{t} + ϕ_{i j t} .$

In both models, we applied a natural logarithmic transformation to ProductVariety_ij, OrderQuan_ijt and FirmCapital_i to better align their distributions more closely with the normal distribution (Chen et al., 2022a). To mitigate potential multicollinearity issues and ensure the variables are on comparable scales, we standardized all variables (Qi et al., 2023; Xu et al., 2023). Multicollinearity can be addressed by centering the variables, which involves subtracting the mean from each variable (e.g. Choi et al., 2021; Li et al., 2021). Following this, all variables exhibit a variance inflation factor (VIF) below the specified threshold value (Aiken and West, 1991), as shown in Table 3.

4.2 Results

We summarized the regression results in Table 4, including the models with only main effects, main effects with control variables, main effects with both control variables and FE, main effects with a single hypothesized interaction term and the full model with all hypothesized interactions. The significance and signs of variables for livestream selling usage are consistent across columns (1)–(3). Thus, we explained the findings based on the results in column (3), which includes FE and control variables. The positive and significant coefficient of livestream selling usage in column (3) (α = 0.4813, p < 0.01) indicates that suppliers’ lead time will be longer with the increasing use of livestream selling. Thus, H1 is supported. The coefficient reveals that one standard deviation rise in livestream selling usage leads to suppliers needing an extra 3.60% of the time to complete orders.

The results of the moderating effects are reported in columns (4) and (5). The interaction term between livestream selling usage and product variety in column (4) is positive and significant (β = 0.3085, p < 0.05), providing support for H2. This finding indicates that livestream selling will lead to a longer lead time with more product variety. Conversely, in column (5), the interaction term between livestream selling usage and order fill rate is negative and significant (γ = −0.5627, p < 0.01), which supports H3. As such, livestream selling will lead to a shorter lead time with a higher order fill rate. The full model in column (6) demonstrates consistent results, supporting all hypotheses.

Figures 2 and 3 provide a more detailed visualization of the moderating effects of product variety and order fulfillment capability. Following Sweeney et al. (2022), we used the 25th and 75th percentile values for both moderators as a reasonable high/low level, along with the coefficients in column (6) to create the figures. As depicted in Figure 2, suppliers with a higher level of product variety are more likely to experience negative impacts on their lead times due to livestream selling. In contrast, Figure 3 shows that suppliers with a higher order fill rate are less susceptible to the adverse impact of livestream selling on their lead time.

5. Robustness tests

5.1 Two-stage instrumental variable model

Omitted variable bias arises when factors correlated with lead time are excluded from the regression equation. Theoretically, while control variables and fixed effect models can help reduce this bias, the risk of endogeneity still exists because unobservable factors are hard to incorporate. For example, streamers’ efforts in a live video could affect demand variability (Lu et al., 2023), thereby influencing lead time. The close collaboration between suppliers and online retailers will empower suppliers to more effectively manage the risks inherent in fulfilling orders (Qi et al., 2023). To address this endogeneity concern, we applied the instrumental variable (IV) method. Specifically, we utilized the average usage of livestream selling (i.e. Average_LSU_ijt) associated with categories from the same supplier (excluding the focal category) as our IV. This Hausman-type IV is commonly used in the literature to satisfy both relevance and exclusion conditions (Hausman, 1996; Xu et al., 2023; Yu et al., 2023). When retailers decide which products to sell via live stream, they take into account not only their marketing strategy and current market demand but also category-specific characteristics. Their decision to use livestream selling for products in a category and the extent of its use often refer to the decisions for products in other categories from the same supplier. Such mimic decision-making behavior among retailers can be observed in prior studies (Gopalakrishnan et al., 2023; Yu et al., 2023). Therefore, the relevance condition is deemed satisfied since the average usage of livestream selling in other categories likely affects the usage in the focal category. Regarding the exclusion condition, the livestream-selling usage in other categories does not directly affect the outcomes of the livestream-selling usage in a focal category. Hence, the average usage of livestream selling can be treated as exogenous and is not correlated with the error term of suppliers’ lead time. In sum, we contended that the average usage of livestream selling in other categories excluding the focal category of a supplier serves as a valid IV.

We conducted a two-stage least squares regression using the IV. In the first stage, we incorporated the IV along with control variables and FE for categories and months to derive the predicted values for the endogenous variable (i.e. livestream selling). In the second stage, we used the fitted value of livestream selling usage (i.e. LSU_IV_ijt) and the interaction terms to analyze lead time. The findings from the IV estimation and the regressions that included instrumented livestream selling are reported in Table 5. In column (1), the results of the first-stage regression reveal a statistically significant relationship between the IV and livestream selling (β = 0.6424, p < 0.01), which verifies the relevance condition statistically. Further, the F-statistics in the first-stage regression exceed 10 (F = 149.85), which indicates that our instrument is not weak (Stock and Yogo, 2002).

The results of the second-stage regression in columns (2)–(5) are consistent with our main results presented in Table 5. Moreover, our instrument successfully passes the under-identification and weak identification tests. Kleibergen and Paap (2006) suggest that if the Kleibergen-Paap rk LM statistic rejects the under-identification null hypothesis (i.e. its p-value is below 0.1), the under-identification test is considered to be passed. In our results, the p-values of the Kleibergen-Paap rk LM statistic for each model are all well below 0.1, unequivocally indicating that our instrument is not under-identified. Stock and Yogo (2002) offer several critical values, such as 10%, 15%, 20% and 25% maximal IV size, which reflect varying levels of tolerance for inference biases in IV estimation. A lower critical value corresponds to a higher F statistic value, with 10% indicating stricter control to minimize bias, while 25% allows for larger potential biases in the estimation. In our results, the Cragg-Donald Wald F statistic for each model exceeds the critical value of 10%, indicating that our instrument is not weak. Therefore, our IV is both exactly and strongly identified in the regression with IV.

5.2 Two-stage Heckman model

The objective of this study is to examine how online retailers’ livestream selling practices directly impact suppliers’ lead time. When exploring these effects, a key empirical challenge lies in the endogeneity associated with a retailer’s decision to adopt livestream selling for a specific product category, which may lead to self-selection bias due to unobservable factors. Hence, we utilized a two-stage Heckman model to correct this selection bias (Heckman, 1979; Klöckner et al., 2022; Xu et al., 2023).

In the first stage, we used a Probit model to estimate the probability of retailers adopting livestream selling for each category for each supplier. The dependent variable in this choice model is a selection criterion, which is assigned a value of 1 if retailers adopt livestream selling for a supplier’s certain category in the respective month, and 0 otherwise. We included an additional exogenous variable, Number of New Consumers, that serves as an exclusion restriction. This variable is measured by the average number of new consumers who placed their initial orders within each category for each supplier during the last month. It influences the retailer’s decision to adopt livestream selling for a supplier’s certain category in the choice model but does not directly affect the ultimate lead time in the second stage. Specifically, the retailer’s choice to adopt livestream selling for a supplier’s certain category is primarily determined by recent competitive advantages in the market, which could be reflected in the number of new consumers prior to the focal month (Chen et al., 2022b); hence, this variable can directly affect the dependent variable in the choice model. Conversely, the number of new consumers from the last month only affects the number of orders for that month, making it unlikely to influence lead time for the focal month. We also included category-related and supplier-related factors that could affect the choice model, such as product variety, order quantity of last month, firm size and firm age. Endogeneity is accounted for by computing the inverse Mills ratio (IMR) using estimates obtained from the first stage. The IMR is calculated as ϕ(βX)/Φ(βX), where βX is the vector of predictions taken from the probit regression, ϕ(⋅) and Φ(⋅) are the density and cumulative distribution functions of the standard normal distribution, respectively (Bose and Leung, 2019). The IMR serves as a critical component of the Heckman model, employed in the outcome equation to mitigate selection bias (Xu et al., 2023).

In the second stage, we tested our hypotheses with the IMR as an additional control for the endogeneity. As shown in Table 6, the IMR is not statistically significant and the results remain highly consistent with the original analysis, suggesting that self-selection was not a concern and our main results are robust.

5.3 Extended analysis

In this section, we conducted additional analyses to evaluate the robustness of our results. Firstly, we excluded samples that did not generate sales through livestream selling within the time window. Table 7 summarizes the results, which are consistent with our main results.

Next, it is worth considering that online retailers in China commonly engage in a well-known shopping festival, similar to Singles’ Day or Black Friday, which occurs annually in June. As consumer purchase behavior and lead time during this period may be greatly impacted by factors other than livestream selling, we excluded samples from June. The results in Table 8 are consistent with our main results.

Additionally, we included FE of the geographic location of suppliers to control for unobservable region-specific factors and geographical differences among suppliers. Using the regional classification established by the National Bureau of Statistics of China, suppliers were categorized into the Northeast, East, West and Central regions within China based on their geographic location. The differences in transportation development across these regions could affect suppliers’ lead times. We introduced FE for these regions using three dummy variables, and the results in Table 9 remained consistent with our main findings. We also compared the Adjusted R² from our main results in Table 4 with those that controlled for the unobservable region-specific factors in Table 9. The improvement in Adjusted R² for each model exceeds 1.5% and is statistically significant. This indicates that the models’ explanation power has increased after correcting the unobservable region-specific factors.

Finally, to validate our findings, we conducted additional estimations using the population-averaged generalized estimating equation (GEE). GEE provides enhanced flexibility in modeling intricate and unbalanced panel data, particularly in management and organizational research (Ballinger, 2004; Shah et al., 2017). It provides robust standard errors even if both the distribution of dependent variables and correlation structures are not accurately specified (Datar et al., 1993; Zou et al., 2023). Given the highly unbalanced nature of our data and its unidentifiable distribution, we employed this method as an alternative for estimating our models. The results in Table 10 are consistent with our main results.

Taken together, all these results in Tables 7–10 indicate the robustness of our main results and provide support for our hypotheses.

5.4 Controlling the impact of COVID-19 pandemic

Our dataset spans from January 2020 to September 2020, encompassing the coronavirus disease 2019 (COVID-19) pandemic period. The government’s stringent controls may have an impact on suppliers’ lead time. During this time, the Chinese government implemented a severe lockdown in Wuhan, which could greatly affect the firms in that city. However, none of the suppliers in our dataset are located in Wuhan or Hubei Province, which mitigates the COVID-19 pandemic’s influence on our findings.

To further control the pandemic’s effects on our results, we constructed a new model in this section. We first removed data from February, the month most heavily impacted by the COVID-19 pandemic, to better isolate its effect on lead time. Additionally, we incorporated the stringency index and the government response index into our model to account for the COVID-19 pandemic’s influence on lead time (Hale et al., 2021). The stringency index measures the extent of closures and containment in each region k, denoted as Stringency_k. The government response index refers to the extent of measures taken by the government to curb the spread of the COVID-19 pandemic in each region k, denoted as GoverSupp_k. Specifically, we calculated the monthly average of the stringency index and the government response index from January to September 2020 for each region where the supplier is located. In this way, we controlled the COVID-19 pandemic’s impact. The results presented in Table 11 indicate that our findings remain consistent with previous analyses after excluding data related to the COVID-19 pandemic.

6. Discussion and conclusions

6.1 Discussion of findings

In today’s hypercompetitive business environment, a growing number of online retailers are adopting a novel consumer-centric practice known as livestream selling to improve consumer shopping experiences. Although this practice brings significant convenience for consumers, it increases impulse purchases and introduces significant demand variabilities into the supply chain, creating an increasingly volatile and dynamic environment for suppliers. In this study, we developed three hypotheses to investigate how changes in consumer behaviors caused by livestream selling impact the operational performance of upstream suppliers and how suppliers’ capabilities, such as product variety and order fulfillment, help them obtain improved performance in this context. To test the hypotheses, we employed multiple econometric models based on a unique dataset at the supplier-category-month level from a leading online retailer in China.

Our empirical results showed that livestream selling adversely affects suppliers’ operational performance by extending their lead times. The finding is associated with retailers’ inventory management strategy in response to demand variability caused by consumer impulse purchases in livestream selling (Marzolf et al., 2024; Rumyantsev and Netessine, 2007). The extended lead time potentially introduces a negative effect on the retailer’s inventory level, as the traditional inventory theory indicated (Rumyantsev and Netessine, 2007). Our results thus demonstrated that the lead time, as a key factor affecting retailers’ inventory management, may experience great changes in the context of livestream selling. Retailers are required to further refine their inventory models by considering the impact of livestream selling on lead time from a supply chain perspective so that all sides benefit. Additionally, previous studies have pointed to the negative effects of other sales channels (e.g. home shopping networks), such as high stockout rates and order cancellation ratios (Lee et al., 2023). Consistent with their findings, our study also shows that livestream selling adversely affects supplier lead times and that the causes of these adverse effects are related to consumers’ impulse buying behavior.

Furthermore, we observed that suppliers’ capabilities (i.e. product variety and order fulfillment capability) can help them obtain improved performance in livestream selling. Specifically, we confirmed that suppliers with lower product variety tend to experience a shorter lead time in the context of livestream selling. This highlights a potential downside of a wider product variety, that is, increasing demand variability, which aligns with previous research indicating the negative effects of product variety on demand forecasting (Wan and Sanders, 2017; Zhou and Wan, 2017). Contrary to the role of product variety, we found that suppliers with a stronger order fulfillment capability are likely to have a shorter lead time in the context of livestream selling. This finding underscores the importance of order fulfillment capability in effectively managing the demand variability in the supply chain (Qi et al., 2023; Vaidyanathan and Devaraj, 2008), emphasizing the strategic significance for suppliers to invest in this capability (Randall et al., 2006).

6.2 Research implications

We make several distinct contributions to the existing literature. First, this study contributes to the literature by investigating how changes in consumer behaviors caused by livestream selling impact the operational performance of upstream suppliers. Although the development of CCSCM is a new trend with the fast development of digital technology, there is a lack of empirical study on how consumer behaviors change influence the supply chain partner’s activities and performance (Esper et al., 2020). Most of the extant literature on livestream selling has revealed its impact on consumers’ purchase behaviors (e.g. Lu and Chen, 2021; Wongkitrungrueng and Assarut, 2020; Zhou et al., 2022). Our study suggests that such changes in consumer behaviors will adversely affect suppliers’ operational performance in terms of lead time. This finding highlights that a shift in consumer orientation in CCSCM should extend to all tiers of the supply chain, rather than focusing solely on those closest to the consumer (Esper et al., 2020).

Moreover, traditional inventory theory views lead time as a critical factor affecting retailers’ inventory levels (Rumyantsev and Netessine, 2007). Our study, therefore, contributes to the theory by uncovering the changes in this key factor of the theory, i.e. lead time, in the context of livestream selling. Since these changes in lead time can be attributed to retailers’ responses to customer demand variability in livestream selling, our research further contributes to the application of inventory theory by emphasizing the necessity for retailers to refine their inventory models to consider the impacts of livestream selling on lead time from a supply chain perspective. This deepens our understanding of traditional inventory theory in the context of livestream selling.

Furthermore, this study contributes to the literature by indicating that suppliers should improve their order fulfillment capability to alleviate the adverse impact due to consumer behavior change in livestream selling. While previous research indicates that order fulfillment capability is essential for enhancing e-procurement performance in traditional supply chains (Vaidyanathan and Devaraj, 2008), our study reveals its growing significance in addressing demand variability and ensuring a satisfactory consumer experience in livestream selling. Moreover, our study underlines that order fulfillment capability is a critical success factor for suppliers in both traditional and CCSCs.

Finally, this study contributes to the literature by denoting that suppliers should carefully extend product variety in livestream selling. Although previous research has demonstrated the negative impact of product variety on suppliers’ operational performance (Sweeney et al., 2023; Wan and Sanders, 2017; Zhang et al., 2007), our results suggest that such a negative effect will be strengthened in the context of livestream selling due to the changes in consumer behaviors. The discovery reveals that the inherent tension between reducing product variety for better operational efficiency and increasing product variety for higher sales becomes more intricate in the context of livestream selling. This underscores the necessity of providing different capabilities to enhance product variety for a traditional and CCSC.

6.3 Managerial implications

Our findings provide important managerial implications for online retailers and suppliers using livestream selling to build a CCSC. While retailers strive to adopt livestream selling to improve consumer experiences, how upstream suppliers respond to the changing consumer behavior has become a major concern for the current supply chain. We demonstrated that livestream selling leads suppliers to encounter an increasingly volatile environment with higher demand uncertainty, potentially impairing their operational performance. Therefore, retailers should enhance their order decision-making strategy to create a mutually beneficial situation for both parties in the context of livestream selling. At the same time, suppliers must adjust their current capabilities to adapt to these changes in livestream selling.

Our study demonstrated that the retailer’s order decisions in response to customer demand variability in livestream selling can negatively affect suppliers’ lead time. Hence, retailers should consider the changes in suppliers’ lead time when optimizing the inventory model from a supply chain perspective. More specifically, they can develop smooth order patterns to reduce the variability of order decisions in this context. This method may ensure less variable lead time while compensating retailer’s inventory level (Boute et al., 2007).

Our study also indicated that suppliers need to improve order fulfillment capability to respond effectively to changing consumer behavior in livestream selling. This enhancement allows them to more effectively manage the increasing order variability and time pressure from retailers, while also meeting the desired consumer experiences. Thus, suppliers should consider further prioritizing improving their order fulfillment capabilities in livestream selling. More specifically, these capabilities could be customized according to the intensity of livestream selling usage of products.

However, our study also demonstrated that suppliers need to carefully extend product variety in livestream selling. A higher product variety leads suppliers to encounter greater challenges in fulfilling orders on time during livestream selling. Therefore, suppliers should put more effort into improving their flexibility to hedge the impact of complexity from product variety. For instance, suppliers can enhance their capabilities to produce in smaller quantities, adjust output levels and modify the production mix (Esper et al., 2020; Shekarian et al., 2020). Furthermore, to reconcile the conflicting goals of suppliers’ operational efficiency and consumer experience, suppliers can collaborate with retailers to identify an appropriate level of product variety.

Finally, our results also highlighted that an effective CCSCM needs a shift in consumer orientation to be applied throughout all levels of the supply chain, not just solely to retailers. Thus, commensurate strategies that emphasize interactions and relationships within a B2B supply chain context should be valued in response to livestream selling. For instance, retailers and suppliers could work together to create a more effective method for sharing information and streamline the procurement process to enhance their response to livestream selling.

6.4 Limitations and future research

Our results should be interpreted with several limitations, which open up intriguing directions for further investigation. Firstly, this study examines the impact of livestream selling through quantitative analysis. Nevertheless, livestream selling creates an interactive communication environment for both consumers and sellers, offering rich textual data that could be used for a qualitative examination of human behaviors and processes within supply chains, as well as their effects on performance. Therefore, future research could explore the dynamics of livestream selling using ethnographic techniques, text mining or netnography (Rynarzewska and Giunipero, 2024). For example, researchers can use these methods to analyze the conditions in livestream selling that influence the changes in consumer behavior.

Secondly, this study explores how retailers’ responses to the demand variability, driven by consumer impulse purchases in livestream selling, influence suppliers’ lead time. Future investigations could delve deeper into how such demand variability affects suppliers’ inventory levels, turnover rates and costs. Moreover, subsequent research could expand on our results and traditional inventory theory to analyze retailers’ optimal order and pricing decisions using analytical models. In light of our findings, future research could also examine how other shifts in consumer behaviors in livestream selling may contribute to different outcomes. For instance, there is a need to investigate whether consumer impatience for extended delivery times (Paluzzi et al., 2024) and feelings of regret associated with purchase decisions made in an urgent buying environment (Lee et al., 2023; Lin et al., 2024) may lead to higher product return rates.

Thirdly, this study highlights the detrimental consequences of the disconnect between retailers and suppliers in the context of livestream selling. The latest research suggests that the implementation of advanced technologies, including chatbots, artificial intelligence (AI) and IoT, is likely to result in more efficient, responsive and intelligent operations (Durach and Gutierrez, 2024). For example, AI-driven chatbots can offer immediate updates on inventory and efficient capacity planning through real-time tracking, thereby enabling rapid responses to sudden changes in demand (Verma, 2023). It is, therefore, worth exploring how these technologies help to improve the coordination between retailers and suppliers and their capabilities (i.e. agility and flexibility) in the context of livestream selling. To further strengthen their partnerships, future studies should also focus on examining how issues in buyer-supplier relationships – such as commitments, trust, powers, opportunism and resource dependence – affect their interactions.

Lastly, this study’s sample is limited to cultural context. It would be interesting to investigate whether livestream selling produces varying effects across different countries with diverse cultural backgrounds. Moreover, our study focuses on the household appliance category. Future studies could broaden the range of products to encompass other categories, including fast-moving and highly substitutable items (Adamopoulos et al., 2021).

This research was supported by the Fundamental Research Funds for the Central Universities [UIBE (ZD3-11)].

Figure 1

The platform supply chain with livestream selling

[Figure omitted. See PDF]

Figure 2

Moderating effects of product variety

[Figure omitted. See PDF]

Figure 3

Moderating effects of order fill rate

[Figure omitted. See PDF]

Table 1

Descriptive statistics

	Unit	Mean	Std. Dev	Min	Max
(1) LeadTime_ijt	Day	13.382	7.061	3.958	39.057
(2) LSU_ijt	Ratio	0.022	0.033	0	0.164
(3) ProductVariety_ij	Unit	82.605	105.241	2	625
(4) OrderFillRate_ijt	Ratio	0.874	0.164	0.248	1
(5) OrderQuan_ijt	10⁵Unit	0.264	0.639	0.00032	4.222
(6) FirmCapital_i	10⁴RMB	25150.24	80603.33	10	601573.1
(7) FirmAge_i	Year	10.837	7.213	1	39

Source(s): Created by the authors

Table 2

Correlation matrix

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
(1) LeadTime_ijt	1
(2) LSU_ijt	0.081^***	1
(3) ProductVariety_ij	0.047^***	0.204^***	1
(4) OrderFillRate_ijt	−0.142^***	0.108^***	0.011	1
(5) OrderQuan_ijt	0.100^***	0.181^***	0.220^***	−0.092^***	1
(6) FirmCapital_i	−0.003	0.055^***	−0.01	0.072^***	0.078^***	1
(7) FirmAge_i	0.011	0.080^***	0.139^***	0.050^***	0.015	0.447^***	1

Note(s): ^*p < 0.1; ^**p < 0.05; ^***p < 0.01

Source(s): Created by the authors

Table 3

VIFs for each variable

	(1)	(2)	(3)	(4)
	Model 1	Model 2	Model 3	Full model
LSU_ijt	1.08	1.18	1.13	1.22
LSU_ijt × ProductVariety_ij		1.12		1.12
LSU_ijt × OrderFillRate_ijt			1.07	1.07
ProductVariety_ij		1.16		1.17
OrderFillRate_ijt			1.08	1.08
OrderQuan_ijt	1.07	1.18	1.09	1.20
FirmCapital_i	1.33	1.36	1.36	1.38
FirmAge_i	1.25	1.31	1.25	1.31
Mean VIFs	1.18	1.22	1.16	1.19

Source(s): Created by the authors

Table 4

Estimated results of lead time

	(1)	(2)	(3)	(4)	(5)	(6)
	Base model	Controlmodel	Model 1	Model 2	Model 3	Full model
LSU_ijt	0.5716^***	0.3667^**	0.4813^***	0.3285^**	0.6180^***	0.4665^***
LSU_ijt	(0.1415)	(0.1472)	(0.1401)	(0.1441)	(0.1431)	(0.1477)
LSU_ijt × ProductVariety_ij				0.3085^**		0.3298^**
LSU_ijt × ProductVariety_ij				(0.1268)		(0.1297)
LSU_ijt × OrderFillRate_ijt					−0.5627^***	−0.5840^***
LSU_ijt × OrderFillRate_ijt					(0.1470)	(0.1470)
ProductVariety_ij				0.5139^**		0.4698^**
ProductVariety_ij				(0.2169)		(0.2119)
OrderFillRate_ijt					−0.8494^***	−0.8330^***
OrderFillRate_ijt					(0.1362)	(0.1362)
OrderQuan_ijt		0.5081^**	0.4132^**	0.2463	0.2843	0.1298
OrderQuan_ijt		(0.2037)	(0.1940)	(0.2092)	(0.1762)	(0.1886)
FirmCapital_i		0.5288^***	0.1411	0.1873	0.2183	0.2561
FirmCapital_i		(0.1860)	(0.1887)	(0.1909)	(0.1873)	(0.1886)
FirmAge_i		−0.1973	−0.1506	−0.2309	−0.1295	−0.1997
FirmAge_i		(0.1743)	(0.1631)	(0.1650)	(0.1653)	(0.1667)
Constant	13.3823^***	13.3823^***	13.3823^***	13.3305^***	13.4432^***	13.3902^***
Constant	(0.1655)	(0.1637)	(0.1400)	(0.1413)	(0.1389)	(0.1402)
N	5,592	5,592	5,592	5,592	5,592	5,592
Adjusted R-square	0.0064	0.0166	0.2371	0.2415	0.2508	0.2551
FE for months	No	No	Yes	Yes	Yes	Yes
FE for categories	No	No	Yes	Yes	Yes	Yes

Note(s): Standard errors are reported in parentheses. ^*p < 0.1; ^**p < 0.05; ^***p < 0.01. All variables are standardized. FE refers to fixed effect

Source(s): Created by the authors

Table 5

Estimated results using a two-stage IV model

	(1)	(2)	(3)	(4)	(5)
	${L S U}_{i j t}$	Model 1	Model 2	Model 3	Full model
${A v e r a g e_L S U}_{i j t}$	0.6424^***
${A v e r a g e_L S U}_{i j t}$	(0.0330)
${L S U_I V}_{i j t}$		0.8775^***	0.7543^***	1.0894^***	0.9695^***
${L S U_I V}_{i j t}$		(0.2305)	(0.2430)	(0.2439)	(0.2567)
${L S U_I V}_{i j t} \times {P r o d u c t V a r i e t y}_{i j}$			0.4374^**		0.4295^**
			(0.1965)		(0.2004)
${L S U_I V}_{i j t} \times {O r d e r F i l l R a t e}_{i j t}$				−1.0022^***	−1.0052^***
				(0.2508)	(0.2545)
ProductVariety_ij			0.4843^*		0.4477^*
ProductVariety_ij			(0.2761)		(0.2698)
OrderFillRate_ijt				−0.9233^***	−0.9068^***
OrderFillRate_ijt				(0.1678)	(0.1673)
OrderQuan_ijt	0.0801^***	0.3210	0.1536	0.1434	−0.0113
OrderQuan_ijt	(0.0147)	(0.2253)	(0.2433)	(0.2038)	(0.2183)
FirmCapital_i	0.0117	−0.1338	−0.1163	−0.0662	−0.0531
FirmCapital_i	(0.0218)	(0.2365)	(0.2356)	(0.2348)	(0.2333)
FirmAge_i	−0.0066	−0.1022	−0.1679	−0.0453	−0.1050
FirmAge_i	(0.0227)	(0.1956)	(0.1978)	(0.2003)	(0.2013)
Constant	0.0959^***
Constant	(0.0168)
N	4,144	4,144	4,144	4,144	4,144
FE for months	Yes	Yes	Yes	Yes	Yes
FE for categories	Yes	Yes	Yes	Yes	Yes
F-test	149.85
Kleibergen-Paaprk LM statistic		203.188^***	205.701^***	205.648^***	209.974^***
Cragg-Donald-Wald F-statistic		2670.619	1341.674	1301.939	874.089
Endogeneity test (p-value)		0.0320	0.0338	0.0012	0.0017

Note(s): Standard errors are reported in parentheses. ^*p < 0.1; ^**p < 0.05; ^***p < 0.01. All variables are standardized. FE refers to fixed effect. IV refers to an instrumental variable

Source(s): Created by the authors

Table 6

Estimated results using a two-stage Heckman model

	(1)	(2)	(3)	(4)
	Model 1	Model 2	Model 3	Full model
LSU_ijt	0.7017^***	0.3798	0.8593^***	0.5457^**
LSU_ijt	(0.2304)	(0.2481)	(0.2346)	(0.2526)
LSU_ijt × ProductVariety_ij		0.9662^***		0.9633^***
LSU_ijt × ProductVariety_ij		(0.2626)		(0.2605)
LSU_ijt × OrderFillRate_ijt			−0.5961^**	−0.6196^**
LSU_ijt × OrderFillRate_ijt			(0.2673)	(0.2806)
ProductVariety_ij		0.3682		0.3131
ProductVariety_ij		(0.3880)		(0.3803)
OrderFillRate_ijt			−0.8652^***	−0.8309^***
OrderFillRate_ijt			(0.2204)	(0.2184)
OrderQuan_ijt	−0.0346	−0.1564	−0.0976	−0.2077
OrderQuan_ijt	(0.3101)	(0.3155)	(0.2834)	(0.2872)
FirmCapital_i	0.0886	0.1366	0.1583	0.1934
FirmCapital_i	(0.2854)	(0.2847)	(0.2834)	(0.2813)
FirmAge_i	−0.3205	−0.3998	−0.2503	−0.3201
FirmAge_i	(0.2480)	(0.2502)	(0.2523)	(0.2531)
IMR	−1.1549	−0.9909	−0.9281	−0.8196
IMR	(1.2305)	(1.2420)	(1.2298)	(1.2454)
Constant	13.6900^***	13.5955^***	13.7270^***	13.6442^***
Constant	(0.2542)	(0.2605)	(0.2557)	(0.2612)
N	2,247	2,247	2,247	2,247
Adjusted R-square	0.2701	0.2823	0.2847	0.2963
FE for months	Yes	Yes	Yes	Yes
FE for categories	Yes	Yes	Yes	Yes

Note(s): Standard errors are reported in parentheses. ^*p < 0.1; ^**p < 0.05; ^***p < 0.01. All variables are standardized. FE refers to fixed effect

Source(s): Created by the authors

Table 7

Estimated results excluding categories without sales from livestream selling

	(1)	(2)	(3)	(4)
	Model 1	Model 2	Model 3	Full model
LSU_ijt	0.4302^***	0.2660^*	0.5597^***	0.3971^***
LSU_ijt	(0.1433)	(0.1462)	(0.1467)	(0.1501)
LSU_ijt × ProductVariety_ij		0.3199^**		0.3379^**
LSU_ijt × ProductVariety_ij		(0.1294)		(0.1324)
LSU_ijt × OrderFillRate_ijt			−0.5275^***	−0.5486^***
LSU_ijt × OrderFillRate_ijt			(0.1491)	(0.1491)
ProductVariety_ij		0.5730^**		0.5236^**
ProductVariety_ij		(0.2465)		(0.2408)
OrderFillRate_ijt			−0.8659^***	−0.8445^***
OrderFillRate_ijt			(0.1471)	(0.1470)
OrderQuan_ijt	0.3909^**	0.2106	0.2614	0.0948
OrderQuan_ijt	(0.1965)	(0.2131)	(0.1786)	(0.1926)
FirmCapital_i	0.1543	0.2033	0.2330	0.2735
FirmCapital_i	(0.2006)	(0.2024)	(0.1992)	(0.2002)
FirmAge_i	−0.1824	−0.2764	−0.1553	−0.2395
FirmAge_i	(0.1732)	(0.1760)	(0.1755)	(0.1776)
Constant	13.5645^***	13.5227^***	13.6302^***	13.5866^***
Constant	(0.1519)	(0.1524)	(0.1506)	(0.1513)
N	5,063	5,063	5,063	5,063
Adjusted R-square	0.2399	0.2453	0.2542	0.2593
FE for months	Yes	Yes	Yes	Yes
FE for categories	Yes	Yes	Yes	Yes

Note(s): Standard errors are reported in parentheses. ^*p < 0.1; ^**p < 0.05; ^***p < 0.01. All variables are standardized. FE refers to fixed effect

Source(s): Created by the authors

Table 8

Estimated results controlling for the impact of shopping festivals

	(1)	(2)	(3)	(4)
	Model 1	Model 2	Model 3	Full model
LSU_ijt	0.3986^***	0.2618^*	0.5498^***	0.4117^***
LSU_ijt	(0.1382)	(0.1435)	(0.1419)	(0.1480)
LSU_ijt × ProductVariety_ij		0.3004^**		0.3313^**
LSU_ijt × ProductVariety_ij		(0.1261)		(0.1314)
LSU_ijt × OrderFillRate_ijt			−0.6734^***	−0.6971^***
LSU_ijt × OrderFillRate_ijt			(0.1490)	(0.1446)
ProductVariety_ij		0.4633^**		0.4223^**
ProductVariety_ij		(0.2137)		(0.2083)
OrderFillRate_ijt			−0.8771^***	−0.8655^***
OrderFillRate_ijt			(0.1372)	(0.1366)
OrderQuan_ijt	0.3998^**	0.2461	0.2487	0.1056
OrderQuan_ijt	(0.1963)	(0.2125)	(0.1749)	(0.1869)
FirmCapital_i	0.1143	0.1558	0.2021	0.2358
FirmCapital_i	(0.1875)	(0.1898)	(0.1863)	(0.1875)
FirmAge_i	−0.1893	−0.2590	−0.1701	−0.2298
FirmAge_i	(0.1643)	(0.1655)	(0.1671)	(0.1678)
Constant	13.0884^***	13.0390^***	13.1504^***	13.0987^***
Constant	(0.1384)	(0.1397)	(0.1378)	(0.1389)
N	4,944	4,944	4,944	4,944
Adjusted R-square	0.2289	0.2327	0.2452	0.2490
FE for months	Yes	Yes	Yes	Yes
FE for categories	Yes	Yes	Yes	Yes

Note(s): Standard errors are reported in parentheses. ^*p < 0.1; ^**p < 0.05; ^***p < 0.01. All variables are standardized. FE refers to fixed effect

Source(s): Created by the authors

Table 9

Estimated results controlling for regional effects

	(1)	(2)	(3)	(4)
	Model 1	Model 2	Model 3	Full model
LSU_ijt	0.5288^***	0.4299^***	0.6463^***	0.5456^***
LSU_ijt	(0.1378)	(0.1391)	(0.1417)	(0.1439)
LSU_ijt × ProductVariety_ij		0.2298^**		0.2537^**
LSU_ijt × ProductVariety_ij		(0.1160)		(0.1191)
LSU_ijt × OrderFillRate_ijt			−0.4933^***	−0.5127^***
LSU_ijt × OrderFillRate_ijt			(0.1415)	(0.1412)
ProductVariety_ij		0.2130		0.1936
ProductVariety_ij		(0.1923)		(0.1901)
OrderFillRate_ijt			−0.7526^***	−0.7510^***
OrderFillRate_ijt			(0.1300)	(0.1303)
OrderQuan_ijt	0.4669^***	0.3819^**	0.3523^**	0.2710^*
OrderQuan_ijt	(0.1628)	(0.1754)	(0.1503)	(0.1609)
FirmCapital_i	−0.0943	−0.0668	−0.0097	0.0138
FirmCapital_i	(0.1728)	(0.1783)	(0.1729)	(0.1779)
FirmAge_i	−0.1981	−0.2215	−0.1770	−0.1950
FirmAge_i	(0.1599)	(0.1630)	(0.1621)	(0.1649)
Constant	13.3823^***	13.3437^***	13.4357^***	13.3953^***
Constant	(0.1333)	(0.1342)	(0.1338)	(0.1346)
N	5,592	5,592	5,592	5,592
Adjusted R-square	0.2590	0.2603	0.2696	0.2710
FE for months	Yes	Yes	Yes	Yes
FE for categories	Yes	Yes	Yes	Yes
FE for regions	Yes	Yes	Yes	Yes
Changes in Adjusted R-square (compared to Table 3 without FE for regions)	2.19%	1.88%	1.88%	1.59%
F-statistics	54.0161	46.4514	47.0429	39.8627
p-value	0.0000	0.0000	0.0000	0.0000

Note(s): Standard errors are reported in parentheses. ^*p < 0.1; ^**p < 0.05; ^***p < 0.01. All variables are standardized. FE refers to fixed effect

Source(s): Created by the authors

Table 10

Estimated results using the GEE model

	(1)	(2)	(3)	(4)
	Model 1	Model 2	Model 3	Full model
LSU_ijt	0.402^***	0.345^***	0.599^***	0.547^***
LSU_ijt	(0.125)	(0.128)	(0.132)	(0.135)
LSU_ijt × ProductVariety_ij		0.227^*		0.262^**
LSU_ijt × ProductVariety_ij		(0.118)		(0.121)
LSU_ijt × OrderFillRate_ijt			−0.586^***	−0.606^***
LSU_ijt × OrderFillRate_ijt			(0.148)	(0.152)
ProductVariety_ij		0.275		0.244
ProductVariety_ij		(0.168)		(0.165)
OrderFillRate_ijt			−0.873^***	−0.875^***
OrderFillRate_ijt			(0.127)	(0.127)
OrderQuan_ijt	0.193	0.066	0.075	−0.049
OrderQuan_ijt	(0.153)	(0.166)	(0.144)	(0.155)
FirmCapital_i	0.025	0.088	0.108	0.171
FirmCapital_i	(0.156)	(0.160)	(0.155)	(0.158)
FirmAge_i	−0.270^*	−0.325^**	−0.270^*	−0.324^**
FirmAge_i	(0.145)	(0.147)	(0.144)	(0.146)
Constant	14.326^***	14.334^***	14.311^***	14.344^***
Constant	(0.367)	(0.363)	(0.359)	(0.356)
N	5,592	5,592	5,592	5,592
Wald Chi-square	1,314	1,341	1,394	1,403
FE for months	Yes	Yes	Yes	Yes
FE for categories	Yes	Yes	Yes	Yes

Note(s): Standard errors are reported in parentheses. ^*p < 0.1; ^**p < 0.05; ^***p < 0.01. All variables are standardized. FE refers to fixed effect

Source(s): Created by the authors

Table 11

Estimated results control the impact of the COVID-19 pandemic

	(1)	(2)	(3)	(4)
	Model 1	Model 2	Model 3	Full model
LSU_ijt	0.4073^***	0.2653^*	0.5158^***	0.3784^***
LSU_ijt	(0.1360)	(0.1386)	(0.1385)	(0.1417)
LSU_ijt × ProductVariety_ij		0.3309^***		0.3458^***
LSU_ijt × ProductVariety_ij		(0.1236)		(0.1255)
LSU_ijt × OrderFillRate_ijt			−0.3961^***	−0.4196^***
LSU_ijt × OrderFillRate_ijt			(0.1362)	(0.1362)
ProductVariety_ij		0.3860^*		0.3391^*
ProductVariety_ij		(0.2075)		(0.2034)
OrderFillRate_ijt			−0.8166^***	−0.8035^***
OrderFillRate_ijt			(0.1340)	(0.1336)
OrderQuan_ijt	0.3928^**	0.2537	0.2994^*	0.1738
OrderQuan_ijt	(0.1780)	(0.1927)	(0.1673)	(0.1803)
FirmCapital_i	0.0918	0.1266	0.1701	0.1967
FirmCapital_i	(0.1815)	(0.1849)	(0.1800)	(0.1828)
FirmAge_i	−0.1385	−0.1954	−0.1235	−0.1697
FirmAge_i	(0.1507)	(0.1543)	(0.1524)	(0.1557)
Stringency_k	0.0429	0.0331	0.0397	0.0315
Stringency_k	(0.0898)	(0.0897)	(0.0894)	(0.0893)
GoverSupp_k	0.0505	−0.0445	0.0740	−0.0107
GoverSupp_k	(0.5563)	(0.5298)	(0.5469)	(0.5219)
Constant	12.9400^***	12.8829^***	13.0100^***	12.9525^***
Constant	(0.1364)	(0.1375)	(0.1367)	(0.1379)
N	5,196	5,196	5,196	5,196
Adjusted R-square	0.2084	0.2126	0.2209	0.2249
FE for months	Yes	Yes	Yes	Yes
FE for categories	Yes	Yes	Yes	Yes

Note(s): Standard errors are reported in parentheses. ^*p < 0.1; ^**p < 0.05; ^***p < 0.01. All variables are standardized. FE refers to fixed effect

Source(s): Created by the authors

Word count: 8287

Show less

Impact of livestream selling on suppliers’ operational performance: an empirical study on an online retailer

Content area

Abstract

Full text

1. Introduction

2. Literature review and hypotheses development

2.1 Livestream selling

2.2 Impact of livestream selling on supplier’s lead time

2.3 Moderating effect of product variety

2.4 Moderating effect of order fulfillment capability

3. Research methodology

3.1 Data

3.2 Measures

3.2.1 Dependent and independent variables

3.2.2 Control variables

4. Analysis and results

4.1 Econometric models

4.2 Results

5. Robustness tests

5.1 Two-stage instrumental variable model

5.2 Two-stage Heckman model

5.3 Extended analysis

5.4 Controlling the impact of COVID-19 pandemic

6. Discussion and conclusions

6.1 Discussion of findings

6.2 Research implications

6.3 Managerial implications

6.4 Limitations and future research