Governance through exit: Pension fund reform impact on real earnings management of portfolio companies

Abstract

Theoretical models predict that the threat of outside blockholder exit can mitigate agency problems and force managers to undertake actions that would maximize firm value in the long run. We examine whether the institutional blockholder exit threat curbs managerial misbehavior and short-termism reflected in real earnings management. Our study exploits a natural experiment—the Polish pension fund reform of 2013 that encouraged pension funds to trade more actively and imposed a real threat of exit on their portfolio companies. Using a difference-in-differences approach, we provide evidence that the reform significantly decreased the level of real earnings management in “treated” companies, that is, companies with open-ended pension funds (OFEs) playing the role of blockholders. The effect was more significant for firms in a multiple blockholder setting, firms under common ownership, and firms with higher insider’s stakes. Moreover, we confirmed that treated companies that decreased real earnings management in the post-reform period experienced the increased long-term operating performance.

1 Introduction

Theoretical models predict that outside blockholder exit threat can mitigate agency problems by constraining value-destroying managerial decisions and forcing managers to undertake actions that would maximize the firm value in the long run (Admati and Pfleiderer, 2009; Edmans, 2009). Agency problems include, among others, earnings management that helps managers obtain private benefits at the cost of shareholders. This problem particularly refers to the so-called real earnings management (REM)—a form of earnings manipulation stemming from real actions such as cuts in R&D. REM is perceived by most researchers (e.g., Roychowdhury, 2006; Cohen and Zarowin, 2010; Zang, 2012; Mellado-Cid et al., 2018) as much more detrimental for firm value than “traditional” accrual-based earnings management (AEM).

Existing studies indicate that since the last two decades firms have shifted their way of earnings management from AEM to REM (Habib et al., 2022). Revised accounting and tax regulations, tighter accounting standards and better auditing quality increase the probability of detection of AEM, make AEM more costly and lower the benefits of earnings management (Cohen et al., 2008; Zang, 2012). Evidence from the field by Graham et al. (2005) indicates that managers prefer REM as it draws less attention from regulatory scrutiny and the auditors. Moreover, a recent study by Amin and Cumming, (2021) documents the significance of REM in the emerging markets context and suggests the importance of institutional blockholders in constraining the REM.

Strong evidence shows that long-term investors holding relatively large stakes (such as pension funds) are mostly interested in long-term value creation, thus mitigating managers’ value-destroying actions, including REM (Roychowdhury, 2006; Koh, 2007; Zang, 2012; Sakaki et al., 2017; Amin and Cumming, 2021).

In this paper, we use a Polish pension fund reform announced on December 6, 2013 (effective February 1, 2014) as a natural experiment to test whether the blockholder (i.e., pension fund) exit threat curbs managerial misbehavior and managerial short-termism reflected in REM. We use the reform as we believe that it encouraged more active trading and imposed an exit threat on portfolio companies of pension funds. We offer several arguments supporting this conjecture.

First, the so-called open-ended pension funds (OFEs),¹ introduced to the Polish pension system in 1999,² were the most prominent group of outside blockholders in Polish listed companies in 2013, holding about 28% of their capitalization and being informed investors (Bohl et al., 2009). Exit theory states that the strength of an exit threat depends on the size of the stake held by the blockholder, determining his motivation for information acquisition, processing and impounding it into prices (Edmans, 2009, 2014).

Second, the reform completely rebuilt the investment policy of OFEs, transforming them from balance funds to equity funds. At the same time, the reform substantially increased the flexibility of their investment strategies, allowing for much higher involvement in international stock markets (gradually increasing the limit from 5% to 30%) and suspending the mechanism used by the Polish financial market supervisor (KNF) to evaluate the performance of open-ended pension funds to “punish” underperforming funds that were generating returns below the given benchmark.

Before 2014, both the limit and the benchmark caused the portfolio structures of OFEs to be quite similar, with smaller funds “mimicking” the behavior of market leaders, as this strategy helped them obtain similar returns and avoid extra costs. The reform indirectly encouraged OFEs to partly replace their domestic holdings with foreign stocks and rebalance their remaining portfolios of domestic companies. The domestic rebalancing was also forced to some extent by another change. Before 2014, contributions to OFEs were obligatory for the majority of Polish employees. The reform suspended the mandatory character of these contributions allowing employees to choose between two tiers of the system: a pay-as-you-go system operated by a state-owned entity (the so-called first pillar) or OFEs managed by private financial institutions (second pillar). This, in turn, extremely limited the inflow into the pension funds, forcing them to manage their portfolios more actively.

Third, the reform caused a side effect that could be important for other blockholders. The selling activity of OFEs increased the stock market liquidity, which could trigger subsequent actions by other blockholders, thereby increasing the “intensity” of the exit threat. We believe that all these circumstances make the reform a perfect laboratory for studying the relationship between blockholder exit threat and managerial misbehavior.

Using a difference-in-differences (DiD) approach, we observe that our treated companies (i.e., companies with at least one OFE holding at least a 5% stake at the end of 2013) significantly decreased real earnings management after the reform compared with control companies.

We observe that the effect was more significant for firms in a multiple-blockholder setting, firms under common ownership, and firms with managerial short-term concerns. This finding entitles us to claim that the observed change in REM is likely to be a direct consequence of the exit threat. It is because this observation aligns with theoretical models predicting that the following three factors should increase the strength of governance through the exit: number of blockholders (Edmans and Manso, 2011), number of blocks held by blockholders (Edmans et al., 2019) and wealth-performance sensitivity (Edmans, 2009, 2014).

To strengthen our conclusions, we then document that the observed change in REM levels holds primarily for the companies likely to engage in earnings manipulations (the so-called suspect companies). We treat as suspect companies those that just meet or beat earnings targets, are overvalued, or face high insider trades (Burgstahler and Dichev, 1997; Park and Park, 2004; Roychowdhury, 2006).

To bolster our arguments that the exit mechanism is the driver of the observed relationships, we divided our treated companies into two groups using managerial entrenchment as a criterion. Entrenched managers are relatively resistant to governance through voice but are still vulnerable to exit threats, particularly when their wealth is related to stock price. We observed that REM levels in our treated companies with entrenched managers decreased after the reform of 2013 and that this relationship holds only for the subgroup of companies with higher managerial (insider) exposure to stock prices.

A series of robustness tests were conducted in our study. First, we repeated our regressions for a smaller sample of companies using propensity score matching, finding the results to be qualitatively unchanged. Next, a parallel trend analysis of REM levels was conducted for our treated and control companies that confirmed the primary results. Moreover, dynamic effects analysis revealed that the blockholder exit threat stemming from pension funds reform had a permanent impact on REM for treated firms. We also controlled for alternative explanations of the observed decrease in REM for our treated companies, including a possible voice effect triggered by increased liquidity (Maug, 1998), a substitution effect between accrual-based earnings management and REM likely to occur after the reform, and family control over the firm. Finally, we extended our basic specification with an analyst coverage control variable and adopted a firm fixed effect model. All these robustness tests confirmed our primary results.

In additional analysis, we referred to the value implications of REM (Cohen and Zarowin, 2010; Vorst, 2016). Using the operating cash flow and the return on assets we confirmed that treated companies that decreased REM in the post-reform period experienced the increased operating performance three years after the implementation of the reform.

We contribute to the literature on blockholder governance mechanisms, finding the results predicted by exit theory. Focusing on a specific group of blockholders—pension funds—we confirm that the so-called pressure-resistant institutional investors (Brickley et al., 1988) can effectively monitor managers through various channels. We also contribute to the literature on earnings management by providing evidence of the effectiveness of one of several possible mechanisms used to curb value-destroying earnings manipulation. To the best of our knowledge, this paper is the first to adopt a natural experiment to test the relationship between an exit threat and REM.

The closest work to ours is a study by Dou et al. (2018), revealing a positive relationship between the blockholder exit threat and financial reporting quality. There are, however, several differences between our studies. First, Dou et al. (2018) test the relationship between exit threat and financial reporting quality. The authors hypothesize that financial reporting quality should increase with the rise in the intensity of the exit threat because the exit threat improves governance through greater manager–shareholder alignment, which in turn is expected to improve financial reporting quality (i.e., better financial reporting quality is an outcome of better governance). We focus on REM as a value-destroying form of earnings manipulation. Second, Dou et al. (2018) do not distinguish between various types of blockholders. We concentrate on pension funds as long-term institutional investors. Moreover, they use a different way of measuring the intensity of exit threats, based on proxies that capture both the liquidity and dispersion of blockholders, whereas we adopt a natural experiment. In studies that use exit threat proxies based on liquidity, it is difficult to eliminate the possible effect of increased liquidity on intervention, which was noticed by Maug (1998). Any measure of the intensity of an exit threat constructed as a combined measure of liquidity and blockholder presence may capture both effects. A natural experiment, in contrast, enables the analysis to be extended by incorporating possible intervention as a separate effect. Last, we also document the positive value implications of cuts in REM triggered by blockholder exit threat.

Edmans and Holderness (2017) call for the empirical investigation of blockholder voice and exit governance mechanisms based on experiments. Moreover, among potential avenues for further research on blockholders and corporate governance, they explicitly encourage researchers to consider the non-U.S. background and include various interactions in a multiple-blockholder setting. Our paper directly addresses this call in all these dimensions.

The remainder of the paper is organized as follows. Section 2 presents the details of the Polish pension fund reform and the literature review. Section 3 is devoted to hypotheses development. In Section 4, we present our research design and in Section 5, we describe our data sources and our sample. Section 6 presents our baseline results. Section 7 extends our study by implementing additional tests increasing the likelihood of our baseline results being the consequence of exit instead of voice mechanism. In Section 8, we provide a series of robustness checks. The last section concludes.

2 Institutional background and literature review

2.1 Institutional background—Polish pension system reforms

In 1999, Poland substantially changed its pension system by replacing the formerly defined benefit pay-as-you-go (PAYG) system with a notional defined contribution PAYG pillar, which was still managed by a government agency (ZUS). Additionally, a second mandatory pillar was created³ based on fully funded defined contributions managed by a group of OFEs. Initially, 21 such funds were granted licenses, but their number gradually decreased (due to mergers) to thirteen OFEs operating in 2013.

Since 1999, the Polish pension system has faced two main changes (reforms), with the last one implemented in December 2013. The reform of 2013 completely remodeled the investment policy of OFEs. When introduced in 1999, OFEs were forced to invest mostly in Polish treasury bonds. Investing in stocks, especially abroad, was limited (Zalewska, 2006). OFEs were allowed to invest up to 40% of their portfolio in shares traded in the domestic regulated market (Warsaw Stock Exchange). Moreover, a 5% limit was imposed on any foreign investments, including foreign treasury bonds. One of the most significant changes implemented in the reform of 2013 was a ban on investments in treasury bonds. All Polish treasury bonds (amounting to about half of the aggregated OFE portfolio) held by OFEs at the end of 2013 were redeemed and their cash equivalent was converted into “I owe you’s” (IOUs) and transferred to the first pillar.⁴ This transformed OFEs in one day (Monday, February 3, 2014) from relatively passive balanced funds into equity funds.

To avoid a rapid sell-off of domestic shares, OFEs were forced to hold at least 75% of their portfolios in domestic shares in the first year after the reform (2014). The lower bound was gradually reduced to 55% in 2015, 35% in 2016, and 15% in 2017. Starting in 2018, neither lower nor upper limits are placed on investing in stocks listed on the domestic regulated market. Simultaneously, the upper limit for foreign investment was gradually increased from 10% in 2014 to 30% in 2016.

These changes completely rebalanced OFE portfolios as well as their investment policies, forcing them to invest mostly in shares and encouraging them to invest abroad.

Figure 1 presents the aggregated portfolio structure of OFEs in 2013–2016. Treasury bonds accounted for almost 50% ($47.7 billion) of aggregated portfolios at the end of 2013 and only 0.8% in 2014 before they disappeared totally in 2016.⁵ The stake of domestic stocks in the aggregated OFE portfolio rocketed from 40% ($40 billion) in 2013 to almost 80% ($39 billion) at the end of 2014 due to the cancellation of Polish treasury bonds, gradually decreasing to about 75%–76% in subsequent years. The proportion of foreign stocks increased from 1.4% ($4.2 billion) in 2013 to 3.8% ($5.6 billion) in 2014 to 7.6% ($10.7 billion) at the end of 2015.

Fig. 1

OFE aggregated portfolio structure in 2013 and 2014-2016 (billions of USD)

Since 1999, the performance of each OFE has been assessed by the Polish financial market supervisor (KNF) according to specific metrics. A minimum rate of return was required equal to half of the average 3-year returns of all OFEs. Underperforming funds were penalized. They were required to cover the deficit (the difference between the minimum required return and the actual return) for all their members. That solution was criticized as it limited OFE flexibility, forcing smaller funds to mimic the market leaders. Consequently, the portfolio structures—and thus the returns of all OFEs—were similar. In 2014, this mechanism was eliminated to increase internal competition among OFEs. Moreover, funds were no longer allowed to use marketing tools to attract new members. As a consequence, OFEs could rely only on the results they would generate to induce new inflows.

Starting in 2014, the second pillar of the Polish pension system (OFE) was no longer considered obligatory for young new workers. Contributions for new workers (still 19.52% of gross salary) were transferred to the first pillar unless they applied for a further retransferring of a portion of their contribution (slightly increased from 2.3% to 2.92%) to a chosen OFE. Moreover, the second pillar stopped being mandatory for current workers, who needed to submit an application if they wanted their contribution (2.92%) to continue being retransferred to a chosen OFE (opt-in).⁶ That change almost eliminated new flows into the second pillar.

Another significant change was connected with future pensioners’ safety. To avoid the risk of “unfortunate timing” of retiring in a period with bearish stock markets that would hardly affect pensioners’ future benefits, OFEs were obliged to gradually shift accumulated savings to the first pillar starting 10 years before the planned retirement of a given member. The change meant that each OFE was required to transfer 1/10 of members’ capital annually⁷ to all members who were going to reach retirement age within 10 years. The mechanism, called a “slider” (or “zipper”), caused substantial outflows during the first years after the implementation of the reform. We conjecture that changes limiting pension funds’ inflows, increasing their outflows, or lowering their liquidity altogether increase the exit threat for their portfolio companies.

Taken together, after the implementation of the reform, OFEs became active equity funds competing with each other and the new regulations encouraged OFEs to trade more actively and shortened their investment horizon.

Figure 2 presents the evolution of the portfolio turnover for a sample of pension funds (OFEs), other domestic institutions, and foreign institutional investors. Figure 2 shows that portfolio turnover of OFEs follows a similar trend to non–pension fund institutions (both, domestic and foreign) in the pre-reform period but shows a markedly upward turn in the post-reform period. The portfolio turnover for the average OFE institution has nearly doubled, and it would appear that OFEs’ investment horizon has shortened significantly over time. That is why we maintain that the new regulations’ driven active trading imposed a substantial exit threat on OFE portfolio companies.

Fig. 2

The portfolio turnover evolution for pension funds (“OFEs”) and for other institutions around 2013 pension fund reform

2.2 The agency perspective of real earnings management

Managers typically use two types of earnings manipulations: accrual-based earnings management and REM. Accrual-based earnings management is reflected in discretionary (unexpected) accruals stemming from “accounting choices”. Real earnings management stems from actual operational decisions such as cuts in research and development (R&D) spending; price discounts accelerating sales; overproduction decreasing the cost of goods sold; or cuts in selling, general, and administrative expenses. Managers use these techniques mainly to inflate current earnings to avoid presenting losses or to meet or beat earnings targets (Burgstahler and Dichev, 1997; Degeorge et al., 1999; Dechow and Skinner, 2000).

Most academics regard earnings management as detrimental because it helps managers obtain private gains at the cost of shareholders (Leuz et al. 2003; Gopalan and Jayaraman, 2012). In that sense, earnings management is an example of agency costs. Moreover, a consensus among researchers indicates that REM is much more detrimental because it represents a departure from optimal operational decisions, thus destroying a company’s long-term ability to generate earnings (Roychowdhury, 2006; Cohen and Zarowin, 2010; Badertscher, 2011; Francis et al., 2016 ). However, earnings management can also be used to convey forward-looking, value-relevant information by removing some of the noise in a truth-telling report of short-term earnings (Ronen and Yaari, 2008; Gunny, 2010). In a systematic review paper, Habib et al. (2022) refer to the destructive perspective of REM as agency-based “opportunistic REM” and the information-gathering view as “efficient REM”.

Various mechanisms can be used by shareholders to curb managerial misbehavior reflected in “opportunistic” earnings management.⁸ Many studies confirm that companies engage in less accrual-based earnings management with the increase of independent board members, the presence of audit committees, and severe scrutiny from auditors and institutional investors. Moreover, blockholders can prevent managers from engaging in real earnings management. Edmans (2009) suggests that blockholders can deter earnings manipulation because they can “see through” the numbers and will sell if high earnings are not backed up by strong fundamentals. Evidence demonstrates that companies engage in less earnings management when there are outside blockholders in their ownership structure (Dechow et al. 1996). Nevertheless, the role of various outside blockholders as potential monitors may differ as the funds do not form a homogeneous group and significantly vary in their investment horizons, strategies, and connections with portfolio companies, which determines their motivation for monitoring.

2.3 Outside blockholders and earnings management

Substantial evidence describes the relationship between ownership structure and earnings management (Dechow et al., 1996; Roychowdhury, 2006; Zang, 2012; Wang, 2014; Sakaki et al., 2017). Many studies confirm a negative association between outside blockholder (institutional) ownership and earnings manipulation. Analyzing a group of listed U.S. firms targeted by the SEC for allegedly overstating earnings, Dechow et al. (1996) find that those firms are less likely to have outside blockholders. Roychowdhury (2006) finds strong evidence of a negative correlation between the measures of REM and institutional ownership. Similarly, Zang (2012) provides evidence that institutional investors exert more pressure on firms to constrain real activities manipulation than accrual-based earnings management. All of these studies, however, don’t examine the specific role of various groups of outside blockholders (institutional investors). Ronen and Yaari (2008) point out that the importance of earnings and the demand for earnings management are sensitive to differences in blockholders’ investment horizons.

The impact of a diverse group of outside blockholders (particularly institutional investors) on earnings management can have the opposite effect. Bushee (1998) tests whether institutional investors create or reduce incentives for corporate managers to cut investment in R&D to meet short-term earnings targets. The author shows that managers do not cut R&D expenditures when institutional ownership is high. Nevertheless, the higher the proportion of “transient” institutional investors (short-term investors with highly diversified portfolios) in corporate ownership, the higher the probability that managers reduce R&D activities to boost current earnings (Bushee, 1998). This observation supports the view that the short-termism of some institutional investor groups pressures managers to behave myopically. Koh (2007) finds that long-term institutional investors constrain accruals management among firms that manage earnings to “meet or beat” earnings benchmarks. On the other hand, “transient” institutional ownership is positively associated with income-increasing accruals management among these firms (Koh, 2007).

Sakaki et al. (2017) find that the higher the ownership stability of pressure-insensitive institutional investors (investors with no direct business ties with their portfolio companies such as mutual funds or pension funds), the lower the firm’s ability to engage in REM. Similar results were obtained by Kałdoński et al. (2020) for companies listed on the Warsaw Stock Exchange (WSE) but only in a subsample of firms subject to capital market pressure (firms that have only single-class shares outstanding and no entrenched managers). No such relationship exists for less value-destroying accrual-based earnings management, which strengthens the monitoring role of stable institutional investors such as pension funds. Amin and Cumming (2021) investigate the relationship between blockholders and REM in a multiple-blockholder setting in eight Asian emerging markets. The authors observe that the presence of an institutional blockholder as the second-largest blockholder (in family-controlled companies) prevents managers from engaging in REM.

Recent studies have begun focusing on the channel used by outside blockholders to monitor managers and curb earnings management. For example, Dou et al. (2018) provide evidence that financial reporting quality measured with earnings management proxies (with higher levels of earnings management representing lower reporting quality) increases with the rise in blockholder exit threat. In their study, the exit threat grows with an increase in blockholder dispersion or competition (measured as an inverse of Herfindahl–Hirschman index) as well as the stock liquidity that strengthens the threat of exit.

Chung et al. (2018), find that accrual-based earnings management decreases with the increase of the stake held by short-term (especially domestic) institutional investors, contrary to many findings for U.S. markets (e.g., Koh, 2007). Chung et al. (2018) further claim a positive role of blockholder exit threat on reducing accrual-based earnings management in the Korean stock market, where active monitoring by voice is highly unlikely due to the predominant role of family-oriented chaebols. The authors interpret the results as a sign of “passive monitoring,” that is, monitoring by the threat of exit. Specifically, they claim that “blockholders with higher turnovers are more likely to credibly signal to the managers of their portfolio firms a potential sell-off when firm fundamentals deteriorate, prompting them to focus on shareholder value” (Chung et al. 2018). The authors, however, do not provide evidence on the value-destroying role of accrual-based earnings management in Korea.

3 Hypothesis development

3.1 Blockholder exit threat as a monitoring channel

Traditionally, literature on the monitoring role of blockholders focuses on the so-called voice channel, that is, monitoring based on direct intervention (Maug, 1998; Shleifer and Vishny, 1986). Shleifer and Vishny (1986) propose a model of effective monitoring by a large shareholder in a widely held company with atomistic shareholders, which is likely to face severe agency problems. Blockholder interventions may take various forms with direct observable actions, such as shareholders’ proposals, voting against managers’ proposals, or publicly expressing critiques of the managers’ actions (Denes et al., 2016; Gillan and Starks, 2000; McNulty and Nordberg, 2016). Interventions might also include unobservable private negotiations behind the scenes (McCahery et al., 2016).

However, instead of pursuing direct intervention, blockholders dissatisfied with underperforming managers can “vote with their feet” and simply sell their stocks. Admati and Pfleiderer (2009), and Edmans (2009), provide theoretical models of monitoring through “exits,” showing that even if blockholders cannot exercise their voice and intervene directly in a company’s operations, they can still govern through the “Wall Street Walk”—selling their blocks and driving down the stock price. These actions punish managers ex-post, thereby inducing them to maximize value ex-ante. What really matters is the threat of exit rather than the exit itself, as Edmans (2014) observes: “The threat of selling may be sufficient to induce the manager to maximize value, so that the actual act is not necessary” (Edmans, 2014, p. 25).

Blockholder exit threat might induce managers to take actions tending to maximize long-term firm value (Admati and Pfleiderer, 2009; Edmans, 2009) and discipline managers by eliminating departures from optimal operational decisions such as real earnings management (e.g., Roychowdhury, 2006; Cohen and Zarowin, 2010; Zang, 2012). Empirical research provides evidence that blockholder exit threat improves firm value (Bharath et al., 2013), and corporate performance (Edmans et al. 2013; Hope et al., 2017). This threat also helps mitigate potential agency problems by decreasing corporate cash holdings (Sun et al., 2021), reducing excess executive perks (Chen et al., 2024), and improving financial reporting quality (Dou et al., 2016).

A debate exists regarding whether the investment horizon of various groups of blockholders affects the strength of their potential exit threats. On the one hand, Döring et al. (2021) suggest that short-term institutional investors (particularly foreign institutional owners) are more likely to discipline managers through credible threats of exit. On the other hand, McCahery et al. (2016) argue that the exit threat from long-term dedicated investors (who tend to intervene prior to a potential exit) is “more meaningful” than the exit threat from short-term transient institutional investors. Long-term investors engage more in monitoring by intervention and do not exit as often as short-term investors by definition. Nevertheless, we believe that when long-term investors eventually decide to exit, it can be treated as a much more informative signal for other investors. Consequently, we maintain that encouraging a long-term institutional investor such as a pension fund to trade more actively makes governance through exit more credible and mitigates corporate misbehavior. Based on these arguments, we formulate the main hypothesis:

• H1. Pension fund blockholders’ exit threat is negatively associated with real earnings management in their portfolio companies

We test this hypothesis using the Polish pension fund reform of 2013, which imposed an exit threat on OFE portfolio companies. Using a difference-in-differences approach, we thus expect to find a significant decrease in the level of REM in “treated” companies (firms having at least one OFE blockholder in their ownership structure) compared to control group companies, after the implementation of the reform.

3.2 Determinants of the intensity (strength) of blockholder exit threat

The intensity (strength) of the exit threat rises with the number of blockholders, as the competition between blockholders in a multiple-blockholder setting results in more information being impounded into prices, thereby increasing the strength of a possible exit signal (Edmans and Manso, 2011; Cvijanović et al., 2022). Taking this into account we formulate the second hypothesis:

• H2. The effect of pension fund blockholders’ exit threat on real earnings management is stronger in firms with multiple blockholders

To test this hypothesis, we calculate a natural logarithm of the number of OFE blockholders who held at least 5% of firms’ shares in 2013 for each company from our sample. We expect to find a negative relationship between this variable and the level of REM after the implementation of the reform.

The effectiveness of the exit mechanism also increases with the number of blocks held by each blockholder in companies from the same industry because the so-called common ownership gives the blockholder the choice of which firms to sell upon a liquidity shock (Edmans, 2014; Edmans and Holderness, 2017; Edmans et al., 2019). Considering these factors, we formulate the third hypothesis:

• H3. The effect of pension fund blockholders’ exit threat on real earnings management is stronger in firms with blockholder common ownership

To test this hypothesis we calculate the natural logarithm of the number of same-industry peers block-held by the average cross-holding pension fund in the year 2013 expecting to find a negative relationship between this variable and the level of REM after the implementation of the reform.

Finally, the effectiveness of the exit threat also depends on managerial (insiders) short-term concerns (e.g., stock price-related wealth). According to Edmans (2009), the sensitivity of managers’ wealth on stock price is crucial for the exit to be effective governance mechanism. “When the manager is more concerned with the stock price, he is more concerned with the effect of blockholder selling if he shirks” (Edmans and Holderness, 2017, p. 577). Based on these considerations, we offer the fourth hypothesis:

• H4. The effect of pension fund blockholders’ exit threat on real earnings management is stronger in firms with higher insiders’ sensitivity to stock price

To test this hypothesis we divide our sample firms by insiders’ wealth sensitivity in the year immediately before the reform using three different proxies: the percentage of equity owned by all members of the management board, the percentage of equity owned by all members of the management board and the supervisory board,⁹ and an indicator variable for firms using stock-based compensation schemes. We expect to find stronger relationships in companies with higher levels of insider wealth sensitivity measures.

4 Research design

We employ a DiD design to test whether the increase in exit threat stemming from pension funds reform has differential effects on real earnings management (REM) for treated and corresponding control firms.

We estimate the following difference-in-differences regression specification:

$${REM}_{i,t}=\alpha +{\beta }_{1}{TREAT}_{i}+{\beta }_{2}{TREAT}_{i}\times {POST}_{t}+\sum {\beta }_{j}{CONTROLS}_{i,t}+{\alpha }_{t}+{\alpha }_{s}+{\varepsilon }_{i,t}$$

(1)

A firm is classified as a treated (TREAT) if it has at least one pension fund (OFE) holding at least 5% of the firm’s shares outstanding in the year immediately before the reform (i.e., 2013). In line with Edmans and Holderness (2017), we acknowledge that there is no theoretical basis for the commonly used 5% threshold—or indeed any threshold. However, in a field study by McCahery et al. (2016), institutional investors widely believed that investor’s equity stake size matters for the effectiveness of an exit threat. Most respondents claimed that the equity stake size should be at least 5% for the exit threat to be effective (McCahery et al., 2016). Moreover, in existing empirical studies on exit threat (e.g., Bharath et al., 2013; Hope et al., 2017) a blockholder is typically defined as a 5% shareholder, as well.

We require a 5% blockholding before the reform to ensure that treated firms do not choose whether to be treated and that the shock leads to an increase in outside blockholder liquidity and exit threats.

The POST indicator takes the value of 1 from 2014 to 2016, and 0 from 2011 to 2013. CONTROLS are general control variables, with α representing the intercept, α_t year fixed effects, α_s industry (sector) fixed effects, and ε_i,t representing the error term. We do not include POST indicators separately because we include year fixed effects.

Firms can manage earnings through operational activities that include the following (Roychowdhury, 2006):

• Sales manipulation, that is, accelerating the timing of sales or generating additional unsustainable sales through increased price discounts or more lenient credit terms. Such manipulation may cause a temporary increase in sales but also can lead to a drop in operating cash flow (OCF),

• Reduction of discretionary expenditures. Firms can reduce, for instance, selling and general expenses (SGE) and thus increase reported earnings,

• Overproduction. Increasing the production volume causes the allocation of fixed production costs to more units, which lowers the cost of goods sold (COGS).

Following previous research, we calculate abnormal operating cash flows (ABOCF), abnormal discretionary expenses (ABSGE),¹⁰ and abnormal production costs (ABPROD) as proxies for deviations in real operations from industry-year “norms” indicating REM. Abnormal levels are calculated as residuals from models proposed by Roychowdhury (2006):

$$\frac{{OCF}_{i,t}}{{A}_{i,t-1}}={\alpha }_{0}+{\alpha }_{1}\times \frac{1}{{A}_{i,t-1}}+{\beta }_{1}\times \frac{{S}_{i,t}}{{A}_{i,t-1}}+{\beta }_{2}\times \frac{{\Delta S}_{i,t}}{{A}_{i,t-1}}+{\varepsilon }_{i,t}$$

(2)

$$\frac{{SGE}_{i,t}}{{A}_{i,t-1}}={\alpha }_{0}+{\alpha }_{1}\times \frac{1}{{A}_{i,t-1}}+{\beta }_{1}\times \frac{{S}_{i,t-1}}{{A}_{i,t-1}}+{\varepsilon }_{i,t}$$

(3)

$$\frac{{PROD}_{i,t}}{{A}_{i,t-1}}={\alpha }_{0}+{\alpha }_{1}\times \frac{1}{{A}_{i,t-1}}+{\beta }_{1}\times \frac{{S}_{i,t}}{{A}_{i,t-1}}+{\beta }_{2}\times \frac{{\Delta S}_{i,t}}{{A}_{i,t-1}}+{\beta }_{3}\times \frac{{\Delta S}_{i,t-1}}{{A}_{i,t-1}}{+\varepsilon }_{i,t}$$

(4)

In line with previous research, we multiply ABOCF and ABSGE by –1 so that higher proxies indicate higher REM. Finally, we sum the proxies to obtain the overall measure of REM.

We include several control variables used in prior research on REM and institutional investor monitoring. We control for general firm characteristics, which existing literature (e.g., Achleitner et al., 2014; Gopalan and Jayaraman, 2012) have proven to be related to REM in insider-based economies. These variables include company size, profitability, sales growth, and debt level. Following Roychowdhury (2006), we also control for institutional ownership. In some specifications, we additionally include institutional investors' characteristics representing their motivation and skills for monitoring. We control for institutional ownership concentration, portfolio turnover, portfolio weight, and multiple blockholding.

5 Data and sample overview

Our study is based on a sample of non-financial companies listed on the Warsaw Stock Exchange in 2011–2016. We begin the sample selection process by obtaining institutional ownership characteristics on all nonfinancial companies listed on the WSE over the research period. Our analyses were restricted to three years before (2011–2013) and three years after the reform (2014–2016) to limit concerns about the potential effect of confounding events over longer horizons.¹¹ We required that each firm exist both before and after the event. In addition, we require that each firm-year observation has the variables necessary to calculate our measures of REM.¹² Finally, we required the availability of data necessary to construct our control variables. Most data were derived from three data sources: the S&P Capital IQ database, Notoria Serwis (a Polish data provider), and the Amadeus database (Bureau Van Dijk – A Moody’s Analytics Company). The few missing data points on insider ownership were collected by-hand.

Our base sample includes 187 non-financial companies listed on the main market of WSE over the period 2011–2016 (1,122 firm-year observations). Table 1 presents the distribution of our sample by industry. The sample spans seven industries, though there is some concentration in capital goods. We classify approximately 48% of firm-years as treated firms. The summary statistics are presented in Table 2. The mean (median) firm in our sample has a return on assets (ROA) of 3.5% (3.4%), total assets of $102.6 (95.6) million, and institutional ownership of 25.4% (22.2%).

Table 1 Sample distribution by industry

Table 2 Summary statistics

Table 3 presents the Pearson correlation coefficients between our main variables.

Table 3 Pearson correlations matrix

6 Primary findings

6.1 Changes in REM around pension funds reform of 2013

We start our analysis by examining whether the pension fund reform of 2013 led to any significant reduction in REM (H1). Table 4 displays the estimation results for Eq. (1) using REM as a dependent variable.

Table 4 Blockholder exit threat and real earnings management

In Model 1, the coefficient on the interaction term TREAT × POST is −0.041 and significant at the 5% level (t-statistic 1.98). This finding indicates that the treatment firms experienced a statistically significant decrease in REM after the reform relative to control firms. Moreover, the economic significance is also meaningful. The decrease in REM of 4.1 percentage points for treatment firms represents 17.3% of one standard deviation of the full sample REM (23.7 percentage points).

Models 2 and 3 substitute the indicator variable with continuous variable to capture more precisely the strength of the treatment effect (H2 and H3). Exit theory predicts that the threat of disciplinary selling is stronger for firms under the multiple-blockholder structure (Edmans and Manso, 2011) and for firms under common ownership (Edmans et al., 2019). Therefore, we also use two continuous variables to capture the strength of the exit threat. Competition between blockholders enhances the exit threat by impounding more information into prices. Therefore, we use the logarithm of the number of pension fund blockholders who hold at least 5% of firm shares in year 2013 (Num_OFE). Finally, to gauge the importance of flexibility regarding which assets to sell upon a liquidity shock, motivated by existing literature on common ownership (He and Huang, 2017) we use the logarithm of the number of same-industry peers block-held by the average cross-holding pension fund (OFE_AvgNum) in the year preceding the reform.

In line with the results from Model 1, the results in Model 2 reveal a negative and significant (t-statistic −2.01) coefficient on Num_OFE × POST. Increasing the number of pension fund blockholders from the 25th to the 75th percentile is associated with a larger decrease in post-reform REM of 3.80 percentage points (i.e., −0.038 × 1.000). Model 3 substitutes the number of pension fund blockholders with the number of same-industry peers block-held by the average cross-holding pension fund. The association becomes negative after the reform, as seen by the negative and significant coefficient on OFE_AvgNum × POST.

Of the control variables, only two are statistically significant, and their signs are in line with our expectations. Larger (SIZE) and more profitable (ROA) firms are less likely to engage in REM. The coefficient on IO is negative and in line with a monitoring-based explanation but not significant.

However, the results of Models (1)–(3) show significant treatment effects consistent with exit theory. Thus, it is still possible that our findings capture information contained in other institutional monitoring proxies. To examine whether our primary results are driven by institutional investors’ incentives and monitoring skills, we control for additional institutional investment characteristics in the regression analyses as suggested by existing research.

Shleifer and Vishny (1986) argue that large and concentrated (institutional) holdings result in better monitoring, as they make the monitoring less costly and more beneficial. Moreover, Hartzell and Starks (2003) claim that more concentrated holdings lower coordination costs among investors and result in better monitoring. Consequently, we augment the regression model using the Hirschman–Herfindahl share concentration index (HHI_IO). The incentives for and effectiveness of monitoring managers also differ with institutional shareholders’ investment horizons (Bushee, 1998; Chen et al., 2007). That is why we add the investors’ portfolio turnover (TURNOVER) to the Eq. (1). The institutional shareholder’s portfolio weight in the sample firm is also included (PORTFWEIGHT). Fich et al. (2015) provide evidence that institutional shareholders have stronger monitoring incentives when firm’s stocks account for a larger percentage of their portfolios. We also control for institutions’ number of blockholdings (MULTIBLOCK), as information advantages and governance experience obtained from multiple blockholdings improve monitoring efficiency (Kang et al., 2018).

The results controlling for additional institutional monitoring proxies are reported in Models (4)–(6). We obtained results comparable to our primary specification for all treatment effects measures. The findings suggest a negative and significant association between all interaction variables and REM.

Overall, the results reported in Table 4 indicate that the pension fund reform led to a meaningful decrease in REM in firms with pension fund blockholding relative to other firms and provide support for exit threat as a governance mechanism mitigating REM (H1). Moreover, in line with our expectations, the observed effect of long-term institutional blockholders’ exit threat on REM is stronger in firms with multiple blockholders (H2) and blockholder common ownership (H3).

6.2 Analysis of suspect firms

Our findings thus far indicate that exit threat mitigates REM. However, to bolster the validity of our research results, we take a conservative approach in this section and do not equate firms’ deviation in real operations from industry norms with REM. Instead, we refer to them as REM only when managers’ incentives of manipulating earnings are present. We adopt this approach acknowledging that some firms can adopt a unique business model to strategically differentiate themselves from industry peers, which mechanically creates deviations in real operations.

To examine whether the effect of institutional blockholders’ exit threat on real earnings management is stronger in firms in which insiders have greater incentives to report better financial performance we estimate a multiple treatment effects variant model of Equation (1). In particular, we interacted the treatment dummy with dummies indicating whether the insiders have incentives to meet or beat earnings targets, as well as incentives arising from overvalued equity or insider selling. The results of these splits are reported in Table 5.

Table 5 Blockholder exit threat and real earnings management – controlling for incentives to engage in earnings manipulation

Previous research shows that earnings management is more severe in firm-years with reported earnings marginally above earnings targets (Burgstahler and Dichev, 1997; Degeorge et al., 1999). Moreover, results of other studies (Bushee, 1998; Gunny, 2010; Roychowdhury, 2006) suggest that benchmark-beating firms engage more in real operations manipulation than others. For this reason, we concentrate on these firm-years to increase the power of our tests, for which abnormal real operations are earnings-target-oriented.

Three earnings benchmarks (targets) commonly adopted by management are zero earnings, previous year’s earnings, and analysts’ earnings forecasts consensus. Hence, we identify treated firm-years suspected of managing earnings and create a dummy variable, BENCHBEAT. We set the variable BENCHBEAT equal to 1 if either net income divided by total assets is between 0 and 0.01 or the change in net income divided by total assets between year t − 1 and year t is between 0 and 0.01, or the firm just meets or beats the analysts’ earnings per share forecast consensus. Otherwise, the dummy variable is coded as 0. Consistent with our expectations, we find in Model (1) of Table 5 that the coefficient on TREAT × POST × 1 {BENCHBEAT=1} is negative and significant, with a p-value less than 0.10. We note that the coefficient on TREAT × POST × 1 {BENCHBEAT=0} is not significantly different from zero.This indicates that, the exogenous shock induced by pension fund reform decreases REM only in firms with short-term incentives to meet or beat earnings targets.

To increase the power of our test we also consider whether overvaluation-based incentives moderate the effect of an exit threat on REM. Several papers examine the association between overvaluation and earnings management and indicate that highly overvalued firms engage more in earnings management practices (Badertscher, 2011; Chi and Gupta, 2009).

To identify overvalued firms, we use a methodology proposed by Rhodes–Kropf et al. (2005). First, we decompose the M/B ratio into three components: firm-specific error, industry-level error, and long-run valuation error (LR_VB), which captures growth opportunities. In the next step, we sum the first two components to achieve total valuation error (TOT_ERR), which captures misvaluation. Next, we classify firm-years as having overvalued equity if TOT_ERR is positive. Badertscher (2011) suggests that the longer a firm is overvalued, the more likely it is to engage in real earnings management. Therefore, we construct a dummy variable OVERVALUED coded as one if TOT_ERR for a treated firm is positive in at least three consecutive years and 0 otherwise.

The results are presented in Model (2) of Table 5, which displays the average treatment effect of pension fund reform on REM for firms with overvalued equity (OVERVALUED=1) and non-overvalued firms (OVERVALUED=0). In line with our predictions, the coefficient on TREAT × POST × 1 {OVERVALUED=1} is negative and significant (t-statistic 2.40). However, in the absence of overvalued equity, firms with pension funds’ blockholdings are not affected by the reform and do not decrease REM.

Furthermore, we test whether a REM decrease induced by pension fund reform varies with insider selling. Park and Park (2004) suggest that managers inflate earnings before they sell their shares. We split our treatment sample into two groups and define an indicator variable equal to 1 depending on whether yearly insider sales are greater than insider purchases {INSIDERNETSELL=1} or not {INSIDERNETSELL=0}. Consistent with our previous findings on insiders’ incentives, the negative estimated treatment effect on REM is statistically significant only for treated firm-years with an INSIDERNETSELL dummy equal to one. This finding shows that the effects of pension fund reforms lowering REM are present when insiders have strong incentives to report better financial performance to sell their own shares at inflated prices.

In sum, the results reported in Table 5 show that insiders’ incentives to meet or beat earnings targets and incentives arising from overvalued equity or insider selling significantly negatively moderate the effect of pension fund reforms on REM. Thus, consistent with our main hypothesis, the threat of exit improves corporate decision-making and limits insiders’ misbehavior.

7 Exit versus voice

7.1 Insiders’ sensitivity to stock price

Exit theory predicts that exit threats will be more effective when insiders’ wealth is more sensitive to the stock price (Edmans, 2009). To test this prediction (H4), most existing research uses two measures reflecting the manager’s interest in the stock price: “wealth-performance sensitivity” proposed by Edmans et al. (2009) and “pay-performance sensitivity” computed using the methodology in Core and Guay (2002).

However, stock-based compensation schemes are uncommon in Poland, and stock prices remain important for insiders. The predominant ownership structure model in Poland is one with a large controlling investor (a family or an individual), which is often an active shareholder involved in the firm (Aminadav and Papaioannou, 2020; Gugler et al., 2014). In the majority of sample firms, management board members and supervisory board members are either large owners¹³ or act as a representative of large owners. Hope et al. (2017) claim that in settings where managers’ shareholdings are often much more valuable than their compensation, researchers cannot simply focus on managerial pay. Instead, researchers should use managerial shareholdings because managers holding a greater proportion of shares would care more about the stock price than managers holding a smaller share. Moreover, according to a McCahery et al. (2016) survey on the corporate governance preferences of institutional investors, 70% of respondents considered managerial equity ownership important or very important for the effectiveness of an exit threat.

To partition our sample firms by insiders’ wealth sensitivity in the year immediately before the reform, our tests use the three following variables. Two continuous variables represent the percentage of equity owned by all members of the management board (MB_OWNERSHIP) and all members of the management board and the supervisory board (MB&SB_OWNERSHIP), and an indicator variable represents firms using stock-based compensation schemes (STOCK_COMP).

As before, to examine whether the effect of institutional blockholders’ exit threat on REM is stronger in firms with higher insider sensitivity to stock prices, we estimate a multiple treatment effects variant model of Eq. (1). In particular, we interacted the treatment dummy with dummies indicating whether the firm insider sensitivity to stock prices is high or low. The results of these splits into firms with high and low firm insider sensitivity to stock price are reported in Table 6.

Table 6 Blockholder exit threat and real earnings management - the effect of insiders’ wealth sensitivity (IWS) to stock prices

First, we consider management board ownership. We classify treatment firm-year observations as highly sensitive to stock price (MB_OWNERSHIP_HIGH=1) if the observation is above the treated sample median in the year immediately before the reform. The other treated firm-year observations are classified as firms with low insider stock-related sensitivity (MB_OWNERSHIP_HIGH=0). We then estimate our baseline model, allowing the treated effect to differ among these two groups. As Model (1) of Table 6 indicates, the estimated effect of pension fund reform on REM is concentrated among firms with high pre-reform insider sensitivity to stock prices. For this group, the estimated DiD coefficient (TREAT × POST × 1 {MB_OWNERSHIP_HIGH=1}) is negative and statistically significant at the 1% level. This evidence is in line with our expectations and supports the conjecture that effectiveness of exit threat as governance mechanism reducing REM is stronger if the insiders’ wealth is more tied to the stock price.

In Model (2), which additionally includes shareholdings of supervisory board members, we substitute management board ownership with all insider ownership. As before, we split our treatment sample into two groups (MB&SB_OWNERSHIP_HIGH=1 and MB&SB_OWNERSHIP _HIGH=0) using the pre-reform median among treated firms. The estimated treatment effect on REM is negative and only statistically significant for treated firm-years with the MB&SB_OWNERSHIP_HIGH dummy equal to one. Next, in Model (3), we consider using stock-based compensation schemes as another proxy for insiders’ sensitivity to the stock price proxy. Our results essentially remain unchanged; however, the statistical significance is much lower than for the previous model specifications (i.e., the coefficient on TREAT × POST × 1 {STOCK_COMP =1} is negative and significant at the 10% level [t-statistic 1.70]). Overall, our results reported in Table 5 provide evidence that consistent with our prediction the effect of exit threat on REM is stronger in firms with higher insider sensitivity to stock price, which supports our last hypothesis (H4).

7.2 Insiders’ entrenchment

Exit theory states that non-controlling blockholders can improve corporate performance even it they are unable to intervene through “voice” (Edmans, 2009). Existing research provides evidence that in CEE countries firms controlled by large individuals very often use various control-enhancing mechanisms (CEMs), including pyramids and dual-class shares (Gugler et al., 2014).¹⁴ Using CEMs grants insiders the power to resist monitoring through intervention (Gompers et al., 2009); however, exit is likely to hold even among entrenched firms (Bharath et al., 2013). Thus, we attempt to use insider entrenchment to test whether governance through exit is operational in our sample. In particular, we predict that the effect of pension funds reform on REM is stronger in firms using dual-class shares.

To test this presumption, we split our treatment sample into two groups and define an indicator variable equal to 1 depending on whether the firm uses dual-class shares (DUALCLASS = 1) or not (DUALCLASS = 0). We then estimate our baseline model, allowing the treated effect to differ among these two groups. The results of this investigation are presented in Model (1) of Table 7.

Table 7 Blockholder exit threat, managerial entrenchment and real earnings management

As Model (1) of Table 7 shows, a negative and significant coefficient on the interactive variable TREAT × POST × 1 {DUALCLASS = 1} indicates that pension funds exit threat mitigates REM in dual-class companies. This result is not the case for the treated subgroup adopting single-class shares. These findings are in line with our expectations; however, we recognize that lower REM in dual-class firms could be an effect of institutional blockholders’ coalition, which can make intervention more successful even in entrenched firms (Amin and Cumming, 2021).

To further assess the exit-threat-based interpretation of our results, we test whether the reported treatment effect for firms using dual-class shares is more pronounced in firms with higher insiders exposure to stock prices. We classify a firm as being high exposed to the stock price if it has above-median management board ownership in the year immediately before the reform.

The results of these splits into firms with high and low insider exposure to stock price are reported in Models (2)–(3) of Table 7. Consistent with our expectation, we find a negative and significant coefficient on TREAT × POST × 1 {DUALCLASS = 1} only in the stock-price-sensitive firm subsample (Model 2). The difference in coefficients on TREAT × POST × 1 {DUALCLASS = 1} between subsamples (Models 2–3) is significant at the 1% level.

Taken together, the results reported in Table 7 support our presumption that governance through exit mitigates REM even in entrenched firms that are considered to be subject to less scrutiny from shareholders.

8 Robustness and additional tests

8.1 Propensity score matching

As the focus of our study is on the relationship between pension fund ownership and REM, we must consider a potential endogenous matching (selection) of pension funds and companies. Previous research suggests that investment decisions of institutions may be based on a clientele preference or regulations. For example, if pension funds are a subject to strict fiduciary restrictions and prefer to invest in large and liquid companies that are often well governed, any observed relationship between certain pension funds’ ownership characteristics and REM could be a result of differences in institutions’ investment strategy rather than exit threat outcome. In other words, pension fund ownership can mitigate REM activities, but the negative association between OFE blockholdings and REM can also occur when institutional investors choose to invest in certain types of firms exhibiting less managerial misbehavior (e.g., REM).

Thus, one of the major challenges of our identification strategy is the nonrandom assignment of firms to the treatment and control groups. Anything that attracts pension funds or discourages them to concentrate their shareholdings before the pension fund reform, which also affects REM after the reform, may bias our results. To address this concern, we use propensity score matching (PSM).

We match the treatment and control firms in the year 2013 (i.e., the last year of the pre-reform period) based on the firm characteristics affecting the allocation decisions of institutional investors. Accordingly, following Bushee (2001), we employ a set of variables associated with institutional ownership. We use firm size (MCAP) because some institutions may prefer or may be constrained to invest in large companies. Due to the liquidity preferences of institutional investors, we include a share volume turnover variable (SHARETURN). We also use the dividend yield (DYIELD) to reflect institutions’ preferences for firms paying dividends.

A dummy variable (WIG20) controls for institutional investors’ preferences for “blue-chip stocks” included in the WIG20 index, which contains stocks of the 20 largest and most liquid companies listed on the WSE. We include firms’ three-year average sales growth rates (SGR3Y) to control for investors’ preferences for growing firms. Regarding firm performance, we employ a market-adjusted rate of return (BHAR1Y) and a dummy variable (DPROF), which equals 1 for firms with a positive income and 0 otherwise. Lastly, we control for risk using a beta coefficient (BETA2Y),¹⁵ the standard deviation of weekly lognormal price returns (TRISK), and leverage (LEV).

We conduct one-to-one matching without replacement and require a minimum caliper distance of 0.01. PSM results in 516 firm-years of matched treatment and control firms. The results of PSM are presented in Table 8. Panel A reports the mean values of firm characteristics for the treatment and control groups, as well as p-values from t-tests of differences. No significant difference in overall propensity score indicates a successful matching. Among individual firm characteristics, only TRISK is different at the 5% level between both groups of firms.

Table 8 Blockholder exit threat and real earnings management: Propensity Score Matching (PSM)

After we ensure covariate balance along almost all firm characteristics, we validate our primary analysis using the matched sample. The results of this investigation are shown in Panel B of Table 8. The coefficients on all interactive variables are negative and statistically significant. In addition, the treatment effects obtained using this sample are in line with those reported in Table 4, implying that differences in firm characteristics are not likely to drive our findings on REM changes affected by pension fund reform.

8.2 Parallel trends assumption and dynamic effects of the event

The key identifying assumption of the DiD framework is that, in the absence of a treatment, the treatment and control firms would have followed the parallel patterns. To validate this parallel trend assumption, we ran a DiD regression analysis by interacting TREAT with an indicator variable for each year to examine the dynamics of the treatment effect. Table 9 reports the results of this examination.

Table 9 Blockholder exit threat and real earnings management: parallel trends analysis

The benchmark year is 2011 (i.e., the year t − 2). Compared to the benchmark period, we do not find significant treatment effects for years t − 1 to t = 0. The coefficients on TREAT × BEFORE (t = −1) and TREAT × BEFORE (t = 0) are not significantly different from zero. The difference between treatment and control groups appears in the year after the reform (t = 1) and continues to be significant in year 3. Both coefficients on TREAT × AFTER (t = 1) and TREAT × AFTER (t = 3) are negative and significant at least at the 5% level.

We also employ a placebo test to examine whether the parallel assumption holds. Maintaining the same treatment and control groups, we use 2014 as a placebo event and reestimate Eq. 1. The unreported results of this test indicate that there is no treatment effect of the placebo event in 2014. The placebo test, together with parallel trends regression, support a causal relationship between pension-fund-reform-related exit threat and a reduction in REM.

Finally, we conducted another investigation to identify whether the blockholder exit threat stemming from pension fund reform has a permanent or transitory impact on REM for treated firms. In particular, we examined the event window, allowing the post-event window to extend one, two, and three years after the reform. We then reestimated our baseline regression model, but now assessing the mean change in REM between the treated and control firms using these dynamic event windows.

Table 10 reports the results.¹⁶ We find that the effects of the blockholder exit threat are the highest in terms of magnitude and statistical significance in the first year after the pension fund reform. In Model 1, the coefficient on the interaction term TREAT × POST is −0.065 and significant at the 1% level (t-statistic 2.70). When we consider a two-year post-event window (Model 2) the average treatment effect decreases from −0.065 to −0.040. With a three-year post-event window (Model 3), the magnitude of the treatment effect increases slightly to −0.041. Taken together, while REM for treated firms decreases up to three years following the reform, the decreases are economically more pronounced and statistically more significant in the first post-event year.

Table 10 Blockholder exit threat and real earnings management: dynamic post-reform window analysis

Consistent with a monitoring mechanism, we find a permanent impact of blockholder exit threat stemming from pension fund reform on REM, though our estimated effect attenuates over time. One reason for this attenuation might be that after the second pillar stopped being mandatory, OFEs started aggressively competing to attract new members. Dasgupta and Piacentino (2015) show that when money managers compete for investor capital, the threat of exit loses credibility, weakening its governance role. On the other hand, it is also possible that in later years following the reform, alternative mechanisms came into force, reducing the disciplining effect of exit threat. For example, Kałdoński and Jewartowski (2020) suggest that under the relatively high level of book-tax conformity and considerable intensification of tax enforcement, the tax authority provides an effective monitoring mechanism for managerial misbehavior. Starting from 2015, Poland has experienced a radical improvement in tax enforcement¹⁷ that may mitigate REM and potentially weaken the significance of governance through exit.

8.3 Alternative explanations

A necessary condition to make exit theory credible is stock liquidity (Edmans, 2009). However, previous studies suggest that liquidity affects both the decision to acquire a block and the choice of governance mechanism. On the one hand, Maug (1998) proposes that liquidity induces new block formation and new blockholders are incentivized to engage more in monitoring through “voice.” On the other hand, an empirical study by Edmans et al. (2013) shows that liquidity-driven block formation reduces the likelihood of governing through intervention (voice).

In line with the existing literature, we also find that a liquidity increase induced by pension fund reform is associated with new block formation. After the 2013 reform, 22% of the treatment sample firms have block formation in new pension funds. To ensure that our results are not driven by the channel proposed by Maug (1988), we removed all firms from the sample with new block formation after the reform and replicated our primary analysis. We report the results of this investigation in Model 1 of Table 11, where we obtained similar findings. The significance and sign of the coefficient on TREAT × POST remain the same as in Table 4, confirming that our primary findings on treatment effects are less likely to be driven by enhanced intervention associated with new block formation by pension funds.

Table 11 Blockholder exit threat and real earnings management – alternative explanations and estimation method

Furthermore, Boone and White (2015) claim that to minimize transaction and monitoring costs, quasi-indexing institutions that are actively managed (e.g., pension funds) can prefer stocks with greater public information production. Quasi-indexers demand greater analyst coverage and promote richer information production. Irani and Oesch (2016) show that a reduction in analyst coverage leads managers to use less REM. Moreover, the loss of coverage results in greater accrual-based earnings management, indicating that analysts influence managers’ preferred mix of accrual and real activities manipulation. Hence, another alternative explanation for our results is that the reform of 2013 could have substantially changed the investment and trading strategies of pension funds resulting in more private information-gathering and less reliance on analyst services.

To address the concern that varying preferences for public versus private information production may drive our results, we augment Eq. (1) with two additional controlling variables. To investigate whether our findings on treatment effects might be biased by lesser demand for analyst services, we use analyst coverage (ANALYST), which we calculated as the natural logarithm of the total number of financial analysts following a firm. Furthermore, to control whether our findings are driven by the substitution effect between two earnings management methods, as suggested by Zang (2012), we follow Kothari et al. (2005) and compute abnormal accruals using the performance-adjusted modified Jones model. We obtain the same results as our primary analysis after controlling for ANALYST (Model 2 in Table 11) and accrual-based earnings management (Model 3 in Table 11). The estimated DiD coefficient (TREAT × POST) remains negative and statistically significant. Hence, it is unlikely that lesser analyst coverage or the earnings management methods mix affects our results.

Another concern for our identification strategy is that most treatment firms are family-controlled. Existing studies (Achleitner et al., (2014), Ma and Ma, (2024)) document that family firms use earnings management strategically and avoid this method of boosting earnings, which destroys the firm’s long-term value (i.e., REM). Therefore, to control for the potential effect of family control on our findings on the treatment effect, we add to Eq. (1) a family firm dummy (FF). Our results (Model 4) and inferences, however, do not change.

Our tests have controlled for many firm characteristics likely to affect REM. However, we acknowledge that the observed treatment effect may still arise because of omitted unobservable firm characteristics. To address this endogeneity concern, we also employ a firm fixed effects model (Model 5 in Table 11). This approach did not change our conclusions.

9 Additional analyses

9.1 Substitution between REM and AEM after pension fund reform of 2013

Although in Model (3) of Table 11 we preliminarily control whether our findings are driven by the substitution effect between two different earnings management methods, in this section we run a direct examination: does accrual-based earnings management increase for those firms that report lower REM following pension fund reform. In particular, we test whether AEM in treated firms exhibits an increase relative to that of control firms following the reform. Table 12 displays the results of our test. In Model (1) of Table 12, we use the same set of control variables as in Eq. (1). In the next step, we augment our primary model with additional institutional monitoring proxies (Model (2) of Table 11). The coefficients on the interactive term TREAT × POST are insignificant, indicating that accrual-based earnings management for treated firms did not change after pension fund reform.

Table 12 Blockholder exit threat and accrual-based earnings management

9.2 REM and long-term operating performance around pension fund reform of 2013

Our previous analyses provide evidence that the Polish pension fund reform of 2013 imposed a real threat of exit on their portfolio companies and decreased the level of real earnings management in treated firms. These results suggest that the threat of outside blockholder exit can mitigate agency problems and force managers to undertake actions that would maximize firm value in the long run. If this is the case, one would expect that the observed change in earnings management (i.e., the decrease in REM) would be beneficial for “treated” firms in the post-reform period¹⁸.

To test this conjecture, we examine the effects of the change in REM on future long-term firm performance. Following Gunny (2010), we measure future operating performance as industry-adjusted operating cash flow (AdjCFO) or industry-adjusted return on assets (AdjROA). AdjCFO (AdjROA) equals the difference between firm-specific CFO (ROA) and the median CFO (ROA) for the same year and industry (four-digit GICS). CFO represents operating cash flow scaled by lagged total assets. ROA is an income before extraordinary items scaled by lagged total assets. Since we have shown in Section 7.2 that treated firms experienced a statistically significant decrease in REM after the reform relative to control firms, we expect to observe an increase in firm performance following the reduced engagement in real activities manipulation in the post-reform period. We are particularly interested in the sign of the interaction term between REM and TREAT × POST. In our regression analysis, we follow Gunny (2010) and additionally include current-period performance measures to control for earnings and cash flow persistence. We also control for company size, market-to-book, stock return, insolvency risk, debt level and institutional ownership. Industry fixed effects and year-fixed effects are included as well.

Model (1) of Table 13 displays the estimation results with AdjCFO for year t +3 as a dependent variable (the results for t + 2 and t + 4 are similar)¹⁹. The coefficient on the triple interaction term TREAT × POST × REM is negative and significant at the 1% level, indicating that treated firms are more likely to experience a positive performance effect when they decrease real earnings management post-reform. The results of Model (2) show that after controlling for additional institutional monitoring proxies, our results remain unchanged. In further analysis, we test whether the observed results hold for using AdjROA for year t +3 as a dependent variable. The coefficients on TREAT × POST × REM of Model (3) and of Model (4) are significantly negative at the 5% level, supporting our previous findings between REM and future performance in treated firms post-reform. All in all, by documenting that future firm performance increases significantly when treated companies decrease real earnings management in the post-reform period, we provide evidence of the real benefits of exit threat as a governance mechanism. We interpret this result as consistent with the view that using exit threat for curbing managerial misbehavior and short-termism reflected in real earnings management is a value-enhancing monitoring activity.

Table 13 Blockholder exit threat, real earnings management and future operating performance

10 Conclusions

Recent studies on institutional investor monitoring provide evidence that outside blockholders can still govern even if they cannot exercise their voice and intervene directly in firm’s operations. The threat of share price decline arising from dissatisfied investors’ exits motivates management to constrain their value-destroying behavior and meet investor demand for maximizing value. The threat of exit by institutional investors has drawn increased attention in the U.S. setting. However, the literature on exits as a governance mechanism for other markets is rather scarce.

In this study, we exploited a natural experiment created by a Polish pension fund reform implemented in 2013 to examine whether blockholder exit threat curbs managerial misbehavior and short-termism reflected in REM. Using a DiD research design, we provide evidence that the reform significantly decreased the level of REM in “treated” companies, that is, companies with pension funds playing the role of blockholders. Moreover, consistent with the exit theory, the effect was more significant for firms with a multiple-blockholder setting, firms under common ownership, and firms with insiders’ wealth closely related to the stock price.

Our results are robust to alternative explanations of the observed effect, such as incentives to engage in earnings management by the pension fund portfolio firms or the possible impact of the “voice” channel used due to new block formations enabled by the increased liquidity induced by the reforms. Our conclusions also hold for a propensity score-matched sample and are robust to different choices of model specifications. We also document that treated companies that decreased real earnings management in the post-reform period experienced increased long-term operating performance. Overall, our results indicate that outside blockholders’ exit threats have a governance role in the Polish setting.

The findings of our study have several implications for managers, investors, and market regulators. First, even in markets where ownership is highly concentrated, managers should consider the possible feedback from informed outside blockholder trading when making decisions about misreporting. Second, non-blockholding (small) investors that rely more on public information such as reported earnings can make better investment analyses and decisions when they evaluate the impact of institutional blockholders on reporting quality and firm performance. Third, policymakers should be aware that reforms affecting pension funds’ investment policies, internal competition, and liquidity can also have a significant impact on channels through which institutions exert governance in their portfolio companies.

Appendix

1.1 Variable definitions

Variable	Variable Definition
Real Earnings Management Variables
REM	amount of real earnings management, which is the sum of ABSGE, ABOCF, and ABPROD for year t. SGE equals selling, general, and administrative expenses. CFO equals cash flow from operations. PROD is the sum of cost of goods sold and the change in inventory during the year. Each component of REM is estimated for each 4-digit GICS industry and year group. Prior to summing, ABSGE, ABOCF are multiplied by −1 so that higher levels of the variables proxy for higher levels of REM. The larger the amount of REM, the more likely the firm is engaging in real earnings management. See Roychowdhury et al. (2006) for complete details.
ABSGE	abnormal discretionary expenses for year t, measured as the product of negative one and the deviations from the predicted values of the corresponding industry-year regression model.
ABOCF	abnormal cash flow from operations for year t, measured as the product of negative one and deviations from the predicted values of the corresponding industry-year regression model.
ABPROD	abnormal production cost for year t, measured as the deviations from the predicted values of the corresponding industry-year regression model.
Institutional Investor Exit Threat Variables
TREAT	indicator variable coded as one if the firm has at least one pension fund (“OFE”) blockholder in year 2013, where blockholder is defined as holding at least the 5 % of the firm’s shares outstanding.
Num_OFE	natural logarithm of one plus the number of pension funds’ (“OFE”) blockholders in year 2013.
OFE_AvgNum	natural logarithm of one plus the number of same-industry peers block-held by the average cross-holding pension fund (“OFE”) in year 2013. See He and Huang (2017) for complete details.
POST	indicator variable coded as one for the years after the announcement of the pension funds reform in year 2013.
General Control Variables
SIZE	natural logarithm of total assets for year t.
ROA	return on assets for year t computed as net income before extraordinary items for year t scaled by total assets in year t − 1.
LOSS	indicator variable coded as one if net income before extraordinary items for year t is less than zero.
GROWTH	annual percentage change in sales for year t.
LEV	leverage ratio (long-term debt in year t, scaled by total assets in year t − 1).
IO	aggregate institutional ownership for year t.
Institutional Investor Monitoring Variables
HHI_IO	Hirschman-Herfindahl institutional ownership concentration index calculated as the sum (over all institutional investors) of the squared percentage owned of the firm’s shares outstanding in year t.
TURNOVER	firm-level weighted average three-year portfolio turnover rate by institutional investors for year t. Portfolio turnover is computed as the fraction of the investor’s portfolio that is no longer held at the end of the three-year period. See Derrien et al. (2013) for computing investor portfolio turnover.
PORTFWEIGHT	firm-level weighted average weight of the value of the equity investment in a firm in the institutional shareholder’s portfolio for year t. See Fich et al. (2015) for complete details.
MULTIBLOCK	firm-level weighted average multiple blockholding residual for year t, where residual is calculated from the regression of ln (1 + raw blockholding number) on the value of total equity holdings of the institutional investor. See Kang et al. (2018) for complete details.
Insiders’ Wealth Sensitivity to Stock Prices Variables
STOCK_COMP	indicator variable coded as one if the firm uses stock-based compensation (options, restricted stock etc.) in year 2013.
MB_OWNERSHIP	percentage of equity owned by all members of the management board, as well their families in year 2013.
MB_OWNERSHIP_HIGH	indicator variable coded as one if the percentage of equity owned by all members of the management board, as well their families in year 2013 is above sample median.
MB_OWNERSHIP_LOW	indicator variable coded as one if the percentage of equity owned by all members of the management board, as well their families in year 2013 is below sample median.
MB&SB_OWNERSHIP	percentage of equity owned by all members of the management board and the supervisory board, as well their families in year 2013.
MB&SB_OWNERSHIP_HIGH	indicator variable coded as one if the percentage of equity owned by all members of the management board and the supervisory board, as well their families in year 2013 is above sample median.
MB&SB_OWNERSHIP_LOW	indicator variable coded as one if the percentage of equity owned by all members of the management board and the supervisory board, as well their families in year 2013 is below sample median.
Insiders’ Entrenchment Variables
DUALCLASS	indicator variable coded as one if the firm uses dual-class shares in year 2013.
Incentives to Engage in Earnings Manipulation
BENCHBEAT	indicator variable coded as one if the firm just meets or beats zero earnings or last –year earnings or analyst EPS forecast consensus in year t, 0 otherwise. Just beating/meeting the zero benchmark (the last year earnings) are firm-years with net income before extraordinary items over lagged total assets between 0 and 1.0 percent (are firm-years with the change in the return of assets ratio is between 0 and 1 percentage point).
OVERVALUED	indicator variable coded as one if the total valuation error (TOT_ERR) that captures misvaluation is positive in at least three consecutive years. TOT_ERR is computed by decomposing MB ratio into firm-specific error, industry-level error, and long-run valuation error (LR_VB) that captures growth opportunities. Each component is estimated for each 4-digit GICS industry and year group. TOT_ERR is the sum of the first two components. See Rhodes-Kropf et al. (2005) for complete details.
INSIDERNETSELL	indicator variable coded as one if insiders’ sales are greater than insiders purchases in year t, 0 otherwise.
Other Firm Characteristics
ANALYST	natural log of 1 plus the number of analysts following the firm in year t.
AEM	abnormal accruals derived from the performance-adjusted modified Jones model for year t. The modified Jones model is estimated for each 4-digit GICS industry and year group. See Kothari et al. (2005) for complete details
FF	indicator variable coded as one if the firm is family controlled at the 25% threshold of control in year 2013.
Selection Equation Variables
MCAP	natural logarithm of market capitalization in year t.
SHARETURN	natural logarithm of the share volume turnover in year t. Share volume turnover is the ratio of total number of shares traded to number of shares outstanding.
DYIELD	value of dividends paid in year t per share of stock held divided by value of one stock in year t.
WIG20	indicator variable coded as one if the firm is included in the WIG20 index in year t.
SGR3Y	average sales growth over the prior three years in year t.
BHAR1Y	buy-and-hold market adjusted one-year rate of return in year t.
DPROF	indicator variable coded as one if the net income is positive in year t.
BETA2Y	market model beta estimated with up to 24 prior monthly returns in year t.
TRISK	standard deviation of weekly log-normal price returns over the past two years in year t.
Additional Analyses Variables
CFO	cash flow from operations for year t scaled by total assets in year t − 1.
AdjCFO	difference between firm-specific CFO and the median CFO for the same year and industry (four-digit GICS) for year t .
AdjROA	difference between firm-specific ROA and the median ROA for the same year and industry (four-digit GICS) for year t.
MB	market-to-book ratio for year t.
ZSCORE	Z-score is calculated following a modified version of Altman's Z-score that proxies for a firm's financial condition. Specifically, Z-score = 3.3(net income / assets) + 1.0(sales / assets) + 1.4 (retained earnings / assets) + 1.2 (working capital / assets) + 0.6(stock price × shares outstanding) / total liabilities.