Revenue recognition and channel stuffing in the Taiwanese semiconductor industry

This paper examines whether a new revenue recognition standard, namely International Financial Reporting Standard 15 Revenue from Contracts with Customers (IFRS 15), incentivizes Taiwanese semiconductor companies to employ channel stuffing (CS) to meet or beat their earnings targets. By using a sample comprising 1399 firms listed on the Taiwanese Stock Exchange and over-the-counter markets between 2013 and 2019, we find that compared with Taiwanese firms in other sectors, semiconductor companies in Taiwan are more likely to employ CS to meet their earnings targets. Moreover, Taiwanese semiconductor companies have become more likely to use CS to meet their earnings targets since the implementation of IFRS 15. The obtained results pass a series of robustness checks and may help the potential driving forces for CS to be understood.


Introduction
International Financial Reporting Standard 15 (hereafter IFRS 15) Revenue from Contracts with Customers became effective on January 1, 2018. Some researchers argue that one objective of the new revenue standard is to curb earnings management and accounting fraud by providing clearer guidance on revenue recognition for all entities with contracts with customers (Napier and Stadler 2020). However, others suggest that this new standard involves a significant number of judgments, which encourage more aggressive revenue recognition approaches (Veysey 2020) and thus leave more latitude for earnings management. This paper aims to provide evidence in this debate by examining the effect of the implementation of IFRS 15 on firms' channel stuffing (CS) activities. 1 Specifically, we investigate whether semiconductor companies (hereafter, SEMI companies) in Taiwan have become more likely to employ CS to meet or beat earnings benchmarks since the adoption of IFRS 15.
We examine the impact of IFRS 15 on SEMI companies' earnings management behavior because such companies involve numerous customer contracts, especially with diverse or constantly changing terms. The effect of IFRS 15 on these firms could be significant (Ernst and Young 2019). We examine the association between CS and SEMI firms' earnings benchmarks because the consequences of failing to meet earnings target could be severe. 2 We use a sample set of SEMI companies from Taiwan because Taiwan's semiconductor manufacturers accounted for more than 60% of total revenue of the global semiconductor industry in 2020 (The Wall Street Journal 2021), and are thus of interest in their own right.
Prior to 2018, SEMI companies used one of three revenue recognition approaches: the sell-in, sell-through, or combination approach. In the sell-in approach, revenue is recognized upon the transfer of products to distributors and product return and pricing adjustment accruals are recorded. In the sell-through approach, revenue is recorded only when a sale is made by a distributor to end customers. In the combination approach, revenue is recognized according to a combination of the sell-in and sell-through approaches.
Under the new revenue recognition rule, semiconductor manufacturers no longer need to differentiate between the sell-in and sell-through methods. The crucial decision that these companies must make is how to estimate probable sales returns when determining the overall consideration that they are entitled to receive. Thus, the new revenue recognition rule may accelerate revenue recognition and consequently is considered to be less conservative than the old rule (Wagenhofer 2014).
Earnings management activities can take many forms. This study mainly focuses on one method of earnings management, namely CS. Firms in the semiconductor manufacturing sector often experience rapid product obsolescence, declining prices over product life cycles, and expected industry downturns (Rasmussen 2013). These problems may lead to product returns and pricing adjustment uncertainties for sales to distributors. Most SEMI companies rely heavily on distribution channels to sell their products and maintain long-term and mutually beneficial partnerships with these channels; thus, CS is the most crucial tool used by firms to manage earnings. Consequently, we hypothesize that SEMI companies in Taiwan are more likely than other companies to use CS to meet their earnings targets (H1).
Moreover, the implementation of IFRS 15 may accelerate revenue recognition, thereby providing managers with a stronger incentive to engage in CS activities. CS is considered to be at least partially responsible for several incidents of financial fraud in the Taiwanese semiconductor industry. Therefore, we hypothesize that SEMI companies have become more likely to engage in CS activities to reach or beat their earnings targets since 2018 (H2).
We collect data from 2013 to 2019 for all listed firms on the Taiwan Stock Exchange (TSE) and over-the-counter (OTC) markets. The final dataset comprises 8986 firm-year observations for 1399 distinct firms. The number of firmyears (firms) for non-SEMI and SEMI companies are 8111 (1258) and 875 (141), respectively.
We find that for the overall sample if firms' current-year earnings are lower than their previous year's earnings, the probability of them engaging in CS increases by 6.3%, which is significant at the 1% level. Furthermore, when SEMI companies miss their earnings target, the probability of them adopting CS to manage their earnings is 4.5% higher than that of non-SEMI companies adopting CS to manage their earnings. This result is significant at the 5% level and consistent with H1.
In addition, we find that when missing their earnings benchmarks, SEMI companies are approximately 10% more likely to employ CS to manage their earnings after the implementation of the new revenue rule. This result is significant at the 10% level. To ensure the robustness of our results, we conduct three sets of additional tests. Specifically, we use an alternative measure of CS, an alternative measure of incentivization to engage in CS, and an alternative probability model to rerun all the performed tests. The test results obtained with the alternative variables and probability model match those obtained with the original variables and probability model.
The possible contributions of this study to earnings management literature are twofold. First, our results contribute to the debate on the impact of a new revenue recognition rule on firms' earnings management activities. Using a sample of large European companies, Napier and Stadler (2020) suggest that IFRS 15 has changed the philosophy of recognition, not only to provide a fairer representation of corporate revenues but also to inhibit the use of revenues for earnings management purpose. By contrast, our evidence shows that the adoption of IFRS 15 may trigger Taiwanese SEMI companies to more actively manage their earnings through CS. Further, studies of real effects rarely address specific changes in accounting standards. To fill this void, this study examines how accounting standard changes that require different recognition rules to be adopted can have real effects.
Second, Napier and Stadler (2020) call for future research "to undertake … other quantitative studies to assess how far IFRS 15 has enhanced users' understanding of companies' activities and overall business model". Further, Cohen et al. (2008) argue that understanding the potential driving forces for real earnings management, such as CS, and their effects on firms' operations is crucial. Our study answers their calls by using a setting in Taiwan and find that the implementation of IFRS 15 drives managers of Taiwanese SEMI companies to more actively engage in CS to reach their earnings expectations.
The remainder of this paper is organized as follows. "Previous studies and development of hypotheses" section presents relevant background information and the developed hypotheses; "Research design" section describes the research design; "Sample selection and descriptive statistics" section describes the sample selection and presents the descriptive statistics of key variables; "Empiricial results" section presents the findings of this study; "Additional tests" section describes the additional tests conducted; and "Conclusions" section provides the conclusions of this study.

Legacy and new revenue recognition standards
The legacy revenue recognition rules reflected the "sell-in" approach. Under this method of revenue recognition, revenue can be recognized when economic benefits are likely to flow to the companies and the significant risks and rewards of goods ownership are likely to be transferred to the buyer. Under these rules, SEMI companies can recognize revenue when the collection of the consideration is probable, the associated costs incurred can be measured reliably, and the companies have neither continuing involvement with nor effective control over the goods sold. In general, the aforementioned criteria are met after products have been shipped and delivered to distributors because the title and risk are transferred to the distributors at this point. 3 Many contractual arrangements, however, contain specific provisions, such as right of return and sales discounts or other rebates, which may render expected revenue uncertain. In these cases, if SEMI companies are unable to reasonably estimate potential returns or price adjustments, they must defer the recognition of sales revenue until the return period lapses or distributors resell the products to end customers (i.e., sales through the manufacturer's distribution channel), at which point the uncertainty regarding price concessions is resolved. This recognition method is called the "sellthrough" approach. Finally, SEMI companies may use one method for some distributors and another method for other distributors depending on whether the final selling price is fixed or variable. This method is called the "combination" approach.
The new revenue recognition standard, namely IFRS 15, became effective in 2018. According to this new standard, revenue is typically recognized as the net expected level of sales returns provided that the seller can reliably estimate the level of returns on the basis of an established historical record or other relevant evidence. 4 The new and legacy accounting guidelines for revenue recognition are largely the same if sales returns can be reliably estimated. However, if the sales returns cannot be reliably estimated, the legacy guidelines require the deferral of revenue recognition for some retail and consumer entities. By contrast, the new guidance requires companies to recognize the revenue and estimate the impact of returns by using a probability-weighted approach or most likely outcome, whichever is most predictive. Thus, under IFRS 15, revenue is recognized when a high probability exists that no significant reversal will occur when the uncertainty is resolved.
Elimination of the contingent limitation, a major change in revenue recognition, may result in revenue being recognized earlier under the new standard than under the legacy standard; thus, SEMI companies can recognize revenue upon the delivery of products to distributors rather than after the selling of goods to end customers. In addition, IFRS 15 may provide strong incentives for SEMI companies to engage in CS because distributors may be pushed to buy excess units so that SEMI companies can recognize the revenue immediately.

Hypothesis development
Earnings management activities can take many forms, including granting generous sales price discounts (Jackson and Wilcox 2000;Roychowdhury 2006), offering lenient credit terms (Chen 2020;Tung et al. 2008), and decreasing the cost of goods sold through overproduction or discretionary expenditures (Roychowdhury 2006). Most such earnings manipulation activities are aimed at meeting or beating earnings benchmarks (Chu et al. 2019;Habib and Hossain 2008).
The current study mainly focuses on one method of earnings management, namely CS, and expects that Taiwanese SEMI companies are highly likely to adopt CS to meet or beat their earnings benchmarks, especially since the adoption of IFRS 15, for several reasons.
First, firms in the semiconductor manufacturing sector often experience rapid product obsolescence and declining prices over product life cycles. These problems contribute to product return and pricing adjustment uncertainties for sales to distributors. Most SEMI companies rely heavily on distribution channels to boost their businesses and maintain long-term and mutually beneficial channel partnerships. Therefore, CS is the most common method used by such companies to manage earnings. Furthermore, as discussed in Sect. 2.1, IFRS 15 may accelerate the timing of revenue recognition, thereby providing managers with strong incentives to engage in CS. Moreover, CS is considered at least partially responsible for several instances of financial fraud in the Taiwanese semiconductor industry. 5 Therefore, the following hypotheses are proposed: H1 Taiwanese SEMI companies are more likely than Taiwanese non-SEMI companies to use CS to meet their earnings targets.
H2 Taiwanese SEMI companies have become more likely to use CS to meet their earnings target since the implementation of IFRS 15.

Identification of firms engaging in CS
Following Jackson and Wilcox (2000) and Tung et al. (2008), we use a two-step procedure to identify firms that are likely to employ CS to manage their earnings. First, we calculate the account receivable turnover (ARTO) and inventory turnover (INVTO) ratios for the third and fourth quarters of the year. 6 Then, we determine the quarterly change in these turnover ratios (i.e., ARTO_ch and INVTO_ch) by subtracting the ratio for the fourth quarter (ARTO_q4 or INVTO_q4) from that for the third quarter (ARTO_q3 or INVTO_q3).
We use the ARTO and INVTO ratios for the third quarter as benchmarks to determine the turnover ratios for the fourth quarter when unexpected CS is not conducted. We expect that if firms grant unusual credit terms in the fourth quarter, the average account receivable for the fourth quarter will increase to a considerable extent; thus, ARTO_q4 will be lower than ARTO_q3. Similarly, if firms employ CS to boost their earnings, the average levels of inventory in the fourth quarter are expected to decrease; thus, INVTO_q4 will be higher than INVTO_q3. Consequently, we expect ARTO_ch and INVTO_ch to be positive and negative, respectively, if firms conduct CS to manage their earnings.
We also use an alternative method, which involves using the firm-years of ARTO_ch and INVTO_ch adjusted for their respective benchmarks, to divide firms into CS and non-CS firms. The benchmarks are determined by the median value for each industry and year. The adjusted ARTO_ch and INVTO_ch values are denoted as ARTO_ch_adj and INVTO_ch_adj, respectively. Finally, firms are considered to engage in CS if their ARTO_ch_adj (INVTO_ch_adj) value is positive (negative).

Measurement of the economic impact of CS on earnings
Following Dechow et al. (2003), Marquardt and Wiedman (2004), and Tung et al. (2008), we use Eq.
(2) to capture the economic impact of CS on earnings for each industry year: where △AR it and △SALES it are equal to AR t − AR t−1 and SALES t − SALES t−1 for each firm i, respectively; β 1 represents the expected change in the gross account receivable for a given level of sales change. The residual in Eq. (2) is the change in the gross account receivable, which cannot be explained by the model and thus represents the estimated economic impact of CS on reported earnings.

Empirical models
To test whether the managers of SEMI companies are more likely to meet or beat their earnings targets through CS after 2018, we estimate the following equation: CS is represented using two variables: CS1 and CS2. The variable CS1 (CS2) is a dummy that is coded as 1 if a firm's ARTO_ch and INVTO_ch (ARTO_ch_adj and INVTO_ch_adj) are positive and negative, respectively, and zero otherwise. The variables ARTO_ch, INVTO_ch, ARTO_ch_adj, and INVTO_ ch_adj are defined in Sect. 3.1.
INCENT represents the management incentive to engage in CS. We use MISS and EBAR as proxies for the aforementioned incentive. MISS is a dummy that is coded as 1 if the earnings estimated before considering an unexpected change in the account receivable are lower than the earnings reported The ARTO ratio equals the net sales divided by the average gross account receivable. The IVTO ratio equals the cost of goods sold divided by the average inventory.
for the previous year, and zero otherwise. EBAR is a dummy, coded as 1 if a firm's earnings before including an unexpected change in the account receivable are lower than the reported earnings in the current year, and zero otherwise. We expect that SEMI companies are more likely to engage in CS if their current-year earnings before including the impact of CS on reported earnings miss their earnings benchmarks, namely the reported earnings for the current year or the reported earnings for the previous year.
We add several control variables that may affect a firm's level of CS activities but are unrelated to our research question, including △INV, △LIQ, OCF, SIZE, and MB, to Eq. (3). △INV is equal to the ratio of the change in inventory between the third and fourth quarters to the total assets at the beginning of the year. △INV is included because firms that have inventory build-up may offer an extend credit line at the end of the year to prevent products from becoming unusable and avoid high inventory holding costs. Thus, the coefficient of △INV is expected to be positive.
Further, △LIQ is the ratio of the change in working capital (excluding inventory) between the third and fourth quarters to the total assets at the beginning of the year. We include △LIQ in Eq. (3) because firms with higher liquidity are less likely to engage in CS; thus, the coefficient of △LIQ is expected to be negative. OCF is equal to the ratio of the operating cash flows to the total assets at the beginning of the year. Because CS involves offering major discounts to distributors or customers to promote products, we expect firms engaging in CS to exhibit low operating cash flows (Das et al. 2012). SIZE is the natural logarithm of the total assets. Larger firms tend to be less volatile and under more scrutiny (Baker et al. 2019); therefore, we expect the coefficient of SIZE to be negatively related to CS. MB is the ratio of the market value to the book value of equity at the end of the year. Firms with considerable growth opportunities generally exhibit upward performance trends and thus are unlikely to engage in CS. Finally, YEAR_dummies and IND_dummies are dummy variables that are used to control for year and industry fixed effects, respectively.

Sample selection and descriptive statistics
Our sample comprises all listed firms on the TSE and OTC markets between 2013 and 2019. Taiwan adopted IFRSs in 2013; thus, the data collection started in 2013. The year 2019 is the final year of data collection because this study was conducted in 2019. We exclude firms that minimally rely on distributors to sell their products and obtain 10,349 initial firm years. 7 The numbers of missing values for measuring the quarterly change in inventory turnover ratios and the quarterly change in liquidity ratios are 378 and 924, respectively. In addition, the number of missing observations for the market-to-book ratio is 61. The final dataset comprises 8986 firm-year observations for 1399 firms. 8 The descriptive statistics of key variables for the entire sample are presented in Panel A of Table 1. The mean of CS1 (CS2) is 0.160 (0.153), which indicates that on average, CS is noted for approximately 16% (15%) of the firm years. Furthermore, the means of MISS and EBAR are 0.487 and 0.520, respectively, which suggests that about half of the firms in the final sample miss their earnings targets. The means of ∆INV and ∆LIQ are 0.003 and − 0.014, respectively, which indicate that the average changes in the inventory and liquidity between the third and fourth quarters account for 0.3% and − 1.4% of total assets, respectively.
To compare CS and financial performance between SEMI and non-SEMI companies, we determine the means of key variables for the two groups (Panel B of Table 1). As presented in Panel B of Table 1, the average CS1 values for the SEMI and non-SEMI companies are 0.202 and 0.154, respectively. Moreover, the difference in the means of CS1 between the aforementioned two groups is 4.8%. This result is significant at the 1% level and suggests that the percentage of SEMI companies engaging in CS is 4.8% higher than that of non-SEMI companies engaging in CS, and this difference is significant. Moreover, similar results are found for CS2. The difference in the mean CS2 values between the SEMI and non-SEMI companies is 4.0%. This result is significant at the 1% level. We also find that the difference in the means of MISS (EBAR) between the SEMI and non-SEMI companies is 2.8% (7.3%), which is significant at the 10% level (1% level). Overall, these findings suggest that a higher portion of SEMI companies than non-SEMI companies miss their earnings targets and engage in CS.
The correlation coefficients between key variables are presented in Table 2. The correlation between CS1 and CS2 is 0.660. This result is significant at the 5% level and is expected because CS1 and CS2 are proxies for the same managerial decision. Furthermore, the correlation between MISS and CS1 (CS2) is equal to 0.061 (0.049). The results are significant at the 5% level. Similarly, EBAR is positively and significantly correlated with CS1 and CS2. The aforementioned findings and the results presented in Panel B of Table 1 suggest that before controlling for other factors, the proxy variables of missing the earnings benchmark are positively correlated with CS variables, consistent with H1.
Finally, as expected, ∆INV (∆LIQ) is positively (negatively) and significantly correlated with CS variables at the 5% level.

Empiricial results
To test H1, we use ordinary least squares to estimate Eq. (3), and the corresponding results are presented in Table 3. Panel A of Table 3 presents the results for the full, SEMI, and non-SEMI samples. Column (1) of Panel A of Table 3 reveals that the coefficient of MISS is 0.063 (t = 7.53) for the entire sample. This result suggests that for the full sample if firms' current-year earnings are lower than their previous year's earnings (i.e., MISS = 1), the probability of firms engaging in CS increases by 6.3%. This result is significant at the 1% level. Furthermore, columns (2) and (3) of Panel A of Table 3 indicate that the coefficients of MISS are 0.085 (t = 2.90) and 0.060 (t = 6.93) for the SEMI and non-SEMI samples, respectively. The aforementioned results are significant at the 1% level. As indicated by the coefficients of MISS, SEMI companies are more likely than non-SEMI companies to employ CS to boost their earnings if they miss their earnings target. This finding is consistent with H1. To ensure that the difference in the coefficients of MISS for the SEMI and non-SEMI samples is statistically significant, we add two additional variables into Eq. (3), namely SC and SC_MISS, and rerun the regression. SC is a dummy variable that is coded as 1 for data of SEMI companies and 0 otherwise. SC_MISS represents the interaction between SC and MISS. Untabulated results show that the coefficient of SC_MISS is 0.045 (t = 2.49). This result suggests that SEMI companies are 4.5% more likely than non-SEMI companies to engage in CS if they miss their earnings target, which is significant at the 5% level.
To test H2, we run regression by using only the observations of SEMI companies for the years before (PRE-IFRS15) and after the implementation of IFRS 15 (POST-IFRS15). Results are presented in Panel B of Table 3. As presented in column (1) of Panel B of Table 3, the coefficient of MISS for PRE-IFRS15 is 0.023 (t = 0.72), which is positive as expected, but not significant at any conventional levels.
By contrast, the coefficient of MISS for POST-IFRS15 is 0.105 (t = 2.21). The results suggest that after the promulgation of IFRS 15, SEMI companies are more likely to adopt CS to boost their earnings to meet or beat their earnings targets; thus, H2 is supported.
Furthermore, to test whether the incremental positive effect of MISS on CS1 after the promulgation of IFRS 15 is significant, we add to the regression a dummy, POST, coded as 1 for POST-IFRS15 observations and zero otherwise, and an interaction term between POST and MISS, POST_MISS. Untabulated result reveals that the coefficient of POST_MISS is 0.097 (t = 1.67), which suggests that the Table 3 Results for the regression of CS on missing the earnings benchmark and control variables For variable definitions, please refer to Sect. 3. *, **, and *** denote significance at the 1%, 5%, and 10% levels, respectively (two-tailed test). The SEMI and non-SEMI samples are observations for semiconductor and non-semiconductor manufacturing companies, respectively. PRE-IFRS15 and POST-IFRS15 represent the observations for years before and after the implementation of IFRS 15, respectively results reported in column (2) of Panel A of Table 3 are mainly driven by the POST-IFRS15 observations. The result also indicates that when missing their earnings benchmarks, SEMI companies are more likely to employ CS to manage their earnings after the promulgation of IFRS 15. The increased marginal probability after the adoption of IFRS 15 is approximately 10%, which is significant at the 10% level. 9

Alternative measure of CS
To check whether our results are robust to an alternative measure of CS, we use CS2 as the dependent variable and rerun all the tests. When we use full sample, untabulated results show that the coefficient of MISS is 0.046 (t = 5.64), suggesting that MISS is positively and significantly associated with CS2. Furthermore, for the SEMI and non-SEMI samples, the coefficients of MISS are 0.060 (t = 2.08) and 0.045 (t = 5.25), respectively. We also add an interaction term between MISS and SEMI companies to the regression model and find that the coefficient of this interaction term is 0.043 (t = 2.42), which is significant at the 5% level. Finally, the coefficients of MISS for the PRE-IFRS15 and POST-IFRS15 observations are 0.023 (t = 0.93) and 0.080 (t = 1.74), respectively. Thus, the results obtained using CS2 are qualitatively similar to those obtained using CS1, which suggests that our results remain robust for a different measure of CS.

Alternative measure of incentive to engage in CS
To examine whether our results are valid for an alternative measure of missing the earnings benchmark, we use EBAR instead of MISS as the independent variable and CS2 as the dependent variable and then rerun all the tests. Untabulated results suggest that the coefficient of EBAR is 0.070 (t = 8.34) for the full sample. In addition, the coefficients of EBAR are 0.098 (t = 3.23) and 0.067 (t = 7.74) for the SEMI and non-SEMI sub-samples, respectively. The coefficient of the interaction term between EBAR and SEMI is 0.028 (t = 2.39), which is significant at the 5% level. The coefficients of EBAR are 0.048 (t = 1.48) and 0.142 (t = 3.14) for the PRE-IFRS15 and POST-IFRS15 samples, respectively. Moreover, the coefficient of the interaction term between POST and EBAR is 0.077 (t = 1.60), which is significant at the 10% level under the one-tailed test. Overall, qualitatively similar results that support our hypotheses are obtained when using EBAR instead of MISS as the independent variable.

Alternative probability model
Equation (3) is essentially a linear probability model. To ensure that our results are robust to different model specifications, we use a logit model and rerun all the tests. 10 The results obtained with the logit model are presented in Table 4. To facilitate the interpretation of the parameter estimates of the logit model, the average partial effect (APE) for key explanatory variables is also provided in Table 4. Column (1) of Table 4 indicates that the coefficient of MISS for the entire sample is 0.514 (z = 7.65), which corresponds to an APE of 0.067 (z = 7.67). This result suggests that the average probability of engaging in CS increases by 6.7% if firms miss their earnings targets. This result is significant at the 1% level and very close to the corresponding result reported in Table 3. Furthermore, columns (2) and (3) of Table 4 reveal that the APE of MISS on CS1 is 0.063 (z = 7.02) and 0.095 (z = 3.19) for the non-SEMI and SEMI samples, respectively. These results are qualitatively the same as the corresponding results presented in Table 3. Moreover, column (4) of Table 4 indicates that the APE of the interaction term SC_MISS on CS1 is 0.036 (z = 2.28). This result suggests that compared with non-SEMI companies, SEMI companies are more likely to engage in CS when they miss their target. The average difference in the probability of engaging in CS between the aforementioned groups is 3.6%, and this result is significant at the 5% level; thus, H1 is supported.
Columns (5) and (6) of Table 4 reveal that the APE of MISS on CS1 is 0.024 (z = 0.73) and 0.122 (z = 2.54) for the PRE-IFRS15 and POST-IFRS15 samples, respectively. The aforementioned result for the PRE-IFRS 15 sample (POST-IFRS15 sample) is not significant (significant at the 5% level). The aforementioned results are qualitatively similar to the corresponding results presented in Table 3. Finally, we add the interaction term POST_MISS to the logit model, use SEMI companies as the test sample, and rerun the tests. The corresponding results are presented in column (7) of Table 4. The APE of POST_MISS on CS1 is 0.116 (z = 1.91). This finding suggests that SEMI companies are more likely to engage in CS after 2018, and the average increased marginal probability is 11.6%. This result is significant at the 10% level and supports H2. Overall, these findings suggest that our results are robust to an alternative probability model.

Conclusions
IFRS 15 has been in effect since 2018. This study uses a set of Taiwanese semiconductor companies (SEMI companies) listed on the TSE and OTC markets between 2013 and 2019 to examine how this new revenue recognition rule affects managers' incentives to adopt a particular form of real earnings management, namely CS, to meet or beat their earnings targets.
Prior to 2018, SEMI companies applied one of three revenue recognition approaches: the "sell-in," "sell-through," or "combination" approach. In the sell-in approach, firms recognize revenue when products are delivered to distributors and product return and pricing adjustment accruals are recorded. In the sell-through approach, revenue is recorded only when a sale is made by the distributor to end customers. Finally, depending on the terms of the sales contract, revenue can be recognized using a combination of the sell-in and sell-through approaches.
Under IFRS 15, SEMI companies do not need to distinguish between the sell-in and sell-through methods. The most important decision that these companies must make is how to estimate probable sales returns when determining the overall consideration that they are entitled to receive. Thus, the new rule may accelerate revenue recognition (Wagenhofer 2014).
Most SEMI companies rely heavily on distribution channels to increase their product sales and maintain long-term and mutually beneficial channel partnerships; thus, CS is the most valuable tool used by firms to manage earnings. Therefore, we argue that Taiwanese SEMI companies are more likely than Taiwanese non-SEMI companies to use CS to meet their earnings targets. Furthermore, the promulgation of IFRS 15 may accelerate the timing of revenue recognition and thus provide managers a strong incentive to engage in CS. Consequently, we propose that SEMI companies have been more likely to engage in CS activities to reach or beat their earnings targets after 2018.
We find that when firms' current-year earnings are lower than their previous year's earnings, their probability of engaging in CS increases by 6.3%. This result is significant at the 1% level. Moreover, when SEMI companies miss their earnings targets, the probability of them using CS to manage their earnings is 4.5% higher than that of non-SEMI companies using CS to manage their earnings. This result is significant at the 5% level.
Furthermore, when missing their earnings benchmarks, SEMI companies are more likely to employ CS to manage their earnings since the promulgation of IFRS 15. The increased marginal probability since IFRS 15 came into effect is approximately 10%, and this result is significant at the 10% level. We conduct three sets of additional tests to ensure the robustness of our inferences. Specifically, we use an alternative measure of CS, an alternative measure of incentive to engage in CS, and an alternative probability model and rerun all the tests. We find that the test results remain the same after replacing the original measures and model with the alternative measures and model, respectively.

Conflict of interest
On behalf of all authors, the corresponding author states that there is no conflict of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.