The diversity of divestiture– stock market reactions around the announcements of divestiture programs

The appointment of a new CEO often results in significant changes in corporate strategy (Weisbach 1995). Research indicates that CEO tenure correlates with economic investment cycles and increases agency problems (Pan et al. 2016). Moreover, new CEOs tend to reverse prior decisions, especially in response to performance issues (Haynes et al. 2002; Weisbach 1995). By announcing a divestiture program, a new CEO signals to stakeholders that they are taking decisive action to enhance firm performance and move beyond the previous administration’s legacy. This signal might be amplified when the new CEO opts for a program over a single transaction, indicating a commitment to comprehensive strategic change.

The presence of new blockholders, shareholders holding significant stakes, can also alter corporate governance dynamics, prompting management to reconsider strategic options (Bethel and Liebeskind 1993; Jensen and Meckling 1976). Blockholders often demand efficiency and may pressure firms to divest underperforming assets. When a new blockholder joins the firm, their influence can signal to the market a shift towards more rigorous corporate oversight, suggesting that management may be more inclined to pursue stand-alone divestitures, which may be perceived as less comprehensive and more reactive.

Financially distressed firms may turn to divestitures as a means of restructuring and regaining stability (Dranikoff et al. 2002; Ravenscraft and Scherer 1987). The announcement of a divestiture program during financial distress serves as a strong signal of management’s proactive efforts to address underlying issues and improve operational efficiency. This signal is critical for reassuring investors that the firm is taking necessary steps to restore health and stability.

According to learning theory, firms with prior divestiture experience are likely to develop the skills and knowledge necessary to navigate future divestitures more effectively (Levitt and March 1988). Erl et al. (2023) show that this experience can lead to positive expectations about the outcomes of divestitures, fostering confidence in management’s ability to implement considerable strategy changes. However, if a firm has recently divested, it may signal to the market that a comprehensive program is not required, leading to a preference for stand-alone transactions that capitalize on existing knowledge.

Divestitures often occur in waves within industries, impacting firms’ strategic decisions (Brauer and Wiersema 2012; McNamara et al. 2008). During such waves, firms may be hesitant to initiate divestiture programs due to concerns about signaling opportunism or following peers without clear strategic rationale. The signals sent by divestiture announcements can differ significantly depending on whether the firm is operating in a divestiture wave. In a wave, a stand-alone divestiture may be perceived as a necessary adjustment to stay competitive, whereas a divestiture program could signal that the firm is engaging in an extensive overhaul, potentially raising concerns about the firm’s stability and management’s ability to handle multiple changes simultaneously.

Following a manual analysis by classifying all divestiture program announcements and their coverage by secondary sources, three program rationales can be defined: financial motives, refocusing, and streamlining. These rationales are in line with prior literature on divestitures (Berger and Ofek 1999; Brauer 2006; Brauer and Schimmer 2010; Hamilton and Chow 1993; Montgomery et al. 1984). Programs that are undertaken out of financial motives divest businesses or assets in response to debt or liquidity concerns, e.g., a struggling industrial conglomerate that tries to reduce its high leverage.² Firms that divest with a financial rationale may also intend to refocus through their program. In such a case, refocusing is considered a secondary rationale, with the primary rationale still being the financial motive (see Fig. 2, dotted bar segments). Refocusing programs divest businesses or assets to exit specific industries or geographies and consequently increase the focus of the business portfolio, e.g., an industrial conglomerate exiting its chemicals business to focus on its engineering core. Prior divestiture research has also referred to refocusing as downscoping (e.g., Johnson 1996) and framed divestitures that follow a refocusing rationale as strategic (Montgomery et al. 1984). Streamlining programs divest businesses or assets as part of a portfolio review in a “housecleaning fashion as a means of ridding the firm of unwanted or undesired units” (Montgomery et al. 1984: 833) or with the intention of “clearing the decks” (Lee and Madhavan 2010: 1352). In contrast to refocusing programs, streamlining programs do not imply exiting an industry or geography, e.g., a hotel conglomerate that announces to review and divest part of its hotel portfolio. In addition, programs may be linked to or directly follow a previous acquisition. Thus, the presented rationale may be a direct consequence of the acquisition and involve both acquired and previously owned units, e.g., a firm may refocus through both acquisitions and divestitures, a firm may divest to streamline its portfolio, or to reduce its debt levels following a significant acquisition.

Divestiture programs are announced in various contexts, often aligned with the program rationale. Three primary announcement types could be identified: sole announcements, part of an acquisition, and part of a restructuring program (see Fig. 3). Sole announcements involve the firm solely disclosing a divestiture program, while acquisitions and restructuring may incorporate divestiture plans. Initial announcements outline specific goals, such as realizing EUR 1.0 billion from divesting non-core assets. Indications involve the CEO or CFO hinting at potential divestitures or initiating a portfolio review. Program updates occur when the firm refines or expands the initial announcement. Divestiture decisions may also be part of broader restructuring efforts, addressing debt or profitability concerns. Timing-wise, divestiture programs are often disclosed alongside quarterly or annual results, with 48% occurring on the same day or as part of a results announcement. Table 7 in the Appendix provides some exemplary announcements and shows the largest divestiture programs by relative and absolute values in terms of expected proceeds and sales.

The dependent variable for the first research question describes a firm’s choice between a divestiture program, a stand-alone divestiture, and non-divesting. Thus, it draws a clear distinction to stand-alone divestitures. Follow-up announcements to a prior announcement were excluded if the initial announcement of the program is included in the study to avoid sample bias. For stand-alone divestitures, an initial list of sell-offs undertaken by the sample was obtained from SDC.³ Transactions needed to be flagged as a divestiture deal by SDC, be announced between 1997 and 2014 and eventually be completed. Further, we excluded the sale of non-operational assets such as property or buildings and divestitures, where the holding is not reduced to a minority holding, e.g., the formation of a 50/50 joint venture. These criteria were applied based on the SDC fields deal type and deal synopsis. To ensure that only substantial divestitures are included, we required each transaction to divest at least 5% of firm size.⁴ The median of the divested firm size of the 101 firms in the sample takes a value of 0.42% for all 2,641 transactions for which size was available. This confirms that most divestitures in the sample are rather small in size and supported setting a size threshold. To avoid the inclusion of divestitures that were part of the identified divestiture programs, we excluded divestitures in the year before and the two years after a divestiture program. Overall, we identified 152 divestitures that fulfilled the criteria. They were undertaken over 142 firm periods.

CEO turnover. The variable takes a value of 1 if the firm’s top executive changed in the focal or the previous year.⁵ To identify management changes, we built an executive database for the sample by systematically searching primary (e.g., press releases) and secondary sources (e.g., financial press, newswires) using the Factiva database and filtering for the subject “management moves”.

Financial distress variables. We captured financial distress based on three measures: dividend cut, negative net income, and stock underperformance. Dividend cut takes a value of 1 if the dividend per share in the past fiscal year is lower than in the year before (e.g., Berger and Ofek 1999; Owen et al. 2010). Negative net income takes the value of 1 if a firm has reported a negative income in the past fiscal year (e.g., Feldman et al. 2016; Lang et al. 1995). Divestiture literature has previously considered prior stock performance as an indicator of financial distress (e.g., Owen et al. 2010). Stock underperformance is measured by calculating the cumulative excess returns for the two years prior to the focal period and ranking the firms.⁶ The excess returns are calculated by deducting the return of a reference index from the realized return. Given that the STOXX Europe 600, from which the sample was drawn, did not offer return data for the entire observation period, we take the S&P Europe 350 as our reference index. The bottom quartile of firms was considered to underperform.⁷

Divestiture wave. The variable takes a value of 1 if the focal period is within an industry divestiture wave. In case no divestiture wave was identified for the industry and period, the variable is 0. Further, we differentiated in the same manner whether the focal period is before, at, or after the peak of a wave. Industries are defined based on the four-digit Standard Industrial Classification. To identify industry divestiture waves, we followed the methodology proposed by Brauer and Wiersema (2012).⁸

In line with prior literature, we include several control variables in the analysis and lagged them by one period: Firm current ratio, firm size and firm leverage. Firm current ratio is the level of slack resources and is calculated as current assets over current liabilities (Feldman et al. 2016). Firm size is measured as the natural logarithm of total assets (Bergh and Sharp 2015; Brauer et al. 2017). Firm leverage is operationalized as total debt scaled by total assets (Berger and Ofek 1999; Dickerson et al. 1997; Haynes et al. 2002). We also consider year-dependent effects through year fixed effects (Brauer et al. 2017; Haynes et al. 2002).⁹

The dependent variable for the investor reaction consideration is the divestiture program market returns. Cumulative abnormal returns (CAR) on the three days surrounding a divestiture program announcement are used to measure the market reaction to such events. We account for potential confounding effects in two ways. Divestiture programs are often announced in the context of and on the same day as other events such as results presentations, restructuring programs, or acquisitions. Results presentations and restructuring program announcements are accounted for through control variables in the main analysis. Acquisition announcements that include a divestiture program announcement were excluded. To eliminate further confounding effects, we adhered to the methods suggested by McWilliams and Siegel (1997): We excluded confounding events in the five-day window around the announcement date using Factiva to screen the press coverage of a firm. After excluding acquisitions and confounded announcements, 144 out of 168 announcements remained in the final sample. We also winsorized at the 2.5% and 97.5% levels based on the CAR (-1, + 1) (e.g., Owen et al. 2010).

The independent variables describe the program characteristics that were obtained from primary and secondary coverage of divestiture programs through Factiva.

Program value. The actual program value was calculated as the relative share to be divested. Based on the sales to be divested or the expected proceeds, the share was measured relative to total sales or enterprise value at the end of the latest prior fiscal year. For program updates, the increase in sales to be divested or expected proceeds was measured. For programs where no value was provided, the variable was set to a value of 0. As shown in Table 1, 68% of all announcements in the OLS sample stated a program value. On average, the announced programs divested 14.5% of firm size, and the median size was 9.8%. Thus, many of the programs likely altered a firm’s operations significantly.

Program length dummy. Announced program length was measured as a binary variable that takes the value of 1 if a timeline was provided. For the descriptive statistics, shown in Table 1, the length was measured in years. 60% of all announced divestiture programs in the sample provided a timeline and were, on average, scheduled to run for 1.49 years or 18 months.¹⁰

All information points were available for 40 observations or 28% of the regression sample. Many divestiture programs are announced concurrently with other events. The restructuring program dummy variable takes a value of 1 if the divestiture program was announced alongside other restructuring measures, such as e.g., workforce reductions or cost-cutting. Three controls were included to account for the fact that nearly half of all divestiture programs were announced on the same day as a firm’s results. Net profit negative takes a value of 1 in the case negative results were reported. Net profit change is the change in net profit for the reported period relative to the previous comparable period. We capped the variable at -100% and + 100%. In case a firm changed from profit to loss, the variable was set to -100%. Coherently, when it changed from loss to profit, it was set to + 100%. The variable takes a value of 0 if no results were reported. In addition, we included the firm control variables outlined above.¹¹ The descriptive statistics and the correlation matrix of the variables are reported in Table 2, while the distribution of firms by industry and country is provided in Table 9

As hypothesized in Hypothesis 1, the announcement of a divestiture program is closely related to a change in the top management. 45% of all divestiture program announcements were preceded by a CEO change in the previous or focal period. This is a significantly higher proportion than for stand-alone divestitures or non-divesting periods (see Table 3). Programs with a financial rationale followed a CEO turnover more often than refocusing or streamlining programs (see Table 10). The multinomial logit regression, as presented in Table 4, supports this finding. CEO turnover is a highly significant determinant of a divestiture program announcement in the base model (b = 0.60, p = 0.000) and all other models. CEO turnover increases the average probability of a divestiture program announcement by 0.05 (p_AME = 0.000). For stand-alone divestitures, no significant effect of CEO turnover can be determined.

The analysis does not show a significant effect of a new blockholder on the probability of either a divestiture program or a stand-alone divestiture announcement. Thus, Hypothesis 2 which postulates that a new blockholder has a considerably antecedent of stand-alone divestitures is not supported. The proportion of stand-alone divestiture announcements preceded by a new blockholder amounted to 35%, only slightly more than the 33% for non-divesting periods, and 34% for divestiture program announcements. The proportion does not significantly differ across program rationales (see Table 10). Accordingly, the multinomial logit model shows no significant effects for the arrival of a new blockholder (see Table 4).

Hypothesis 3 assumes that firms engage in divestiture programs rather than stand-alone divestitures when financially distressed. Divestiture program announcements were preceded in 60% of all instances by a financial distress event. This is a significantly higher proportion than the 39% for stand-alone divestitures and non-divesting periods. All three financial distress events (dividend cut, negative net income, stock underperformance) exhibit the highest proportions for divestiture programs (see Table 3). When differentiating by program rationale, intuitively, the proportion was highest for financially motivated programs at 79%. The proportion for refocusing programs was 46%, thus, significantly lower than the 60% for all programs (see Table 10). Accordingly, as shown in Model 1 of Table 4, financial distress is a significant predictor in the multinomial logit regression (b = 0.87, p = 0.000). It increases the average probability of a divestiture program announcement by 0.07 (p_AME = 0.000). Model 2 shows that dividend cuts and stock underperformance have a significant impact on the occurrence of a program. The average probability of a program is increased following a dividend cut by 0.04 (b = 0.49, p = 0.08, p_AME = 0.09), stock underperformance by 0.05 (b = 0.58, p = 0.008, p_AME = 0.005). The effect of negative net income is 0.03 (b = 0.39, p = 0.11, p_AME = 0.14) and slightly below statistical significance. In contrast to the significantly explanatory power for program announcements, the probability of stand-alone divestitures is not found to increase following any of the three financial distress events.

Hypothesis 4 postulates that firms with high levels of divestiture experience would be less likely to initiate a divestiture program due to limited opportunities to divest and instead engage in stand-alone divestitures. The proportion of prior experience was similar for divestiture program announcements and stand-alone divestitures with 80% and 81%, respectively. Both are significantly higher than the 70% for non-divesting periods. The proportion of experience in t-2 and t-3 is similar for both divestiture programs and stand-alone divestitures, at 49–51%. The proportion of experience in t-1 is lower for divestiture programs compared with stand-alone divestitures at 48% vs. 54% (see Table 3). Differences in proportion between different program rationales lack significance (see Table 10). The regression as presented in Model 1 of Table 4 shows that experience in the previous three years does not increase the average marginal probability of divestiture program announcements. As argued for stand-alone divestitures there is a significant effect of 0.06 (b = 0.79, p = 0.001, p_AME = 0.003). When differentiating experience by year of occurrence (see Model 3), only distant experience (in t-3) is a significant predictor of a divestiture program announcement (b = 0.14, p = 0.08). In contrast, for stand-alone divestitures experience is a significant positive predictor in t-2 (b = 0.22, p = 0.003) and t-3 (b = 0.19, p = 0.011). Experience in t-1 does not show a statistical effect (b = 0.02, p = 0.72).

In line with the assumptions postulated in Hypothesis 5, the proportion tests presented in Table 3 show significantly higher proportions for divestiture programs in contrast to non-divesting periods before (5.5% vs. 2.8%) and at the peak (4.1% vs. 1.5%) of an industry divestiture wave. After the peak, the proportion was significantly lower (3.4% vs. 7.0%). Given the generally low occurrence of industry divestiture waves within the sample, interpretation of the sub-samples by program rationale should be considered carefully. Results show that occurrence is highest for programs with a financial rationale at and before the peak of a wave (see Table 10). While the coefficient is negative, the occurrence of a wave itself, as examined in Model 1 of Table 4, is no significant indicator of a divestiture program (b = -0.21, p = 0.27). However, the differentiation by timing (see Model 4) is as postulated by Hypothesis 5. It shows that after its peak, an industry wave is a significant negative predictor of a divestiture program announcement. The average probability of a divestiture program announcement decreases by 0.07 (b = -1.02, p = 0.011, p_AME = 0.020) after the peak of an industry wave. Further, the same holds for stand-alone divestitures, for which the average probability decreases by 0.05 (b = -0.77, p = 0.08, p_AME = 0.14) after the peak of an industry wave.

To test for the robustness of the analysis, we ran two alternate models of the regression in addition to the main model. Given the panel nature of the data, a panel model recommends itself. A Hausman test rejected a fixed effects estimator. Thus, a multinomial logit model was fitted using STATA’s gsem command in connection with a latent variable at the firm level to capture the random effect. However, the full models could not be fitted, and computation was only possible for the base model without year effects. Results were robust compared with the mlogit regressions reported in the Results section. The same holds for unreported xtlogit regressions for a choice between stand-alone divestitures and divestiture program announcements. Moreover, we added to our year fixed effects several additional fixed effects. We do not find that firm, industry, or country fixed effects impact our results.¹⁵

To further assure the robustness of our analyses, we performed several variations concerning the definition of variables and included additional controls. First, to avoid the inclusion of divestitures that were part of a divestiture program in the stand-alone divestiture sample, the main regression excluded divestitures in the year before and the two years after an announced divestiture program. When loosing this constraint to one year before and only one year after, 162 divestitures are included in the analysis, and results remain robust. Second, we replaced divestiture experience with dummy variables. The experience dummy over all three previous years is positive and significant for both divestiture programs and stand-alone divestitures. When differentiating by year, all three dummies are positive and significant for stand-alone divestitures. For divestiture programs, none of the individual dummies is significant. Third, we included the previous year’s industry sales growth as a control variable to account for the industry environment (e.g., Haynes et al. 2002). However, given its low explanatory power, the variable was not included in the reported analysis. The results remain robust. Fourth, we differentiated new blockholder in passive and non-passive blockholders. We define blockholders as passive if they were a financial company without activist record, e.g., asset management firms, institutional funds, banks, and insurance companies. We categorize investors as non-passive if they were strategic or activist investors, e.g., activist funds, private equity funds, non-financial companies, or private investors. We employed Factiva to identify traces of activism for all financial companies. Differences for passive and non-passive blockholders between divestiture programs, stand-alone divestitures, and non-divesting periods are not significant.