Not on the same page: comprehensibility of MBS investment prospectuses

Hibbeln, Martin; Metzler, Ralf; Osterkamp, Werner Article — Published Version Not on the same page: comprehensibility of MBS investment prospectuses Review of Derivatives Research Provided in Cooperation with: Springer Nature Suggested Citation: Hibbeln, Martin; Metzler, Ralf; Osterkamp, Werner (2025) : Not on the same page: comprehensibility of MBS investment prospectuses, Review of Derivatives Research, ISSN 1573-7144, Springer US, New York, NY, Vol. 28, Iss. 2, https://doi.org/10.1007/s11147-025-09213-8 This Version is available at: https://hdl.handle.net/10419/323690 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/4.0/ Vol.:(0123456789) Review of Derivatives Research (2025) 28:9 https://doi.org/10.1007/s11147-025-09213-8 Not onthesame page: comprehensibility ofMBS investment prospectuses MartinHibbeln1 · RalfMetzler1· WernerOsterkamp1 Accepted: 20 April 2025 / Published online: 26 June 2025 © The Author(s) 2025 Abstract We investigate whether originators of mortgage-backed securities lower the comprehensibility of disclosure documents to obfuscate low security quality. Utilizing commonly used complexity proxies primarily based on the volume of the investment prospectus, we find no impact on the performance of European mortgage-backed securities. In contrast, focusing on the comprehensibility as a combination of prospectus volume and text readability, we find that originators obfuscate low security quality by lowering the comprehensibility of the investment prospectus, resulting in greater defaults and lower returns. Investors only partially price these dimensions of complexity: Since the financial crisis, investors have been demanding a significant risk premium for prospectus volume but not for text readability. These findings underscore the importance of assessing comprehensibility of disclosure documents beyond traditional complexity measures and highlight potential implications for originators, investors and regulators in communicating, understanding and mitigating the risks associated with complex financial instruments. Keywords Complexity· Disclosure· Prospectus comprehensibility· Security design· Text analysis JEL Classification D82· D83· G01· G12· G14· G23· M41 We thank Fabian Rendchen for his support with the data collection. We are grateful for helpful comments from our discussants Christine Bangsgaard and Johannes Kriebel. We also thank the participants of the Paris Financial Management Conference 2022, the Banking Research Workshop Münster 2022, and the Cardiff Fintech Conference 2022 for helpful comments. This work was supported by the German Research Foundation (DGF) [grant number 434130478]. * Martin Hibbeln mar[email protected] Ralf Metzler [email protected] 1 Mercator School ofManagement, University ofDuisburg-Essen, Lotharstr. 65, 47057Duisburg, Germany M.Hibbeln et al. 9 Page 2 of 37 1 Introduction “A 200-page prospectus that nobody ever reads and a one-page flyer with a pot of gold on it and a lot of fine print”—Martin Wheatley, Chief Executive of the Securities and Futures Commission of Hong Kong, about disclosure documents of complex derivative financial products (May 28, 2010). Complex securitizations experienced a significant surge in defaults during the financial crisis of 2007/08 (Griffin etal., 2014). The empirical literature finds evidence for a complexity channel in pre-crisis U.S. securitization markets: Originators strategically increased complexity to obfuscate low securitization quality, with “complexity” commonly quantified using volume-based proxies derived from the investment prospectus and the number of tranches (Furfine, 2014; Ghent et al., 2019). In contrast, we focus on a distinct dimension of complexity: the comprehensibility of the investment prospectus, as it is the main disclosure document and thus serves as the main information transmission tool of the originator toward investors. We measure prospectus comprehensibility as the combination of prospectus volume and text readability. While a recent strand of the literature studies the text readability of corporate disclosure documents in isolation (e.g., Hwang & Kim, 2017; Li, 2008; Lo et al., 2017), we contend that hard-to-read but short investment prospectuses should still be comprehensible for investors (and vice versa). However, for long prospectuses, low text readability may be a major obstacle for fully understanding the securitization’s legal terms. We contribute to the literature by showing that low prospectus comprehensibility is associated with greater defaults and lower returns on the European mortgage-backed security (MBS) market. Yet, prospectus comprehensibility is only partially priced after the financial crisis: While investors demand a significant risk premium for prospectus volume, they do not demand any risk premium for the readability of the text. We interpret the poorer performance of MBS with less comprehensible prospectuses as a strategic attempt by originators to extract value from investors by obfuscating the low quality of the sold securities. This interpretation is contingent on two premises: First, there must exist some investors who are not sophisticated enough to fully understand the legal terms set out in some of the particularly incomprehensible prospectuses. Although most MBS investors are institutions, we believe this condition is met for several reasons. Most importantly, there is variation in the sophistication of institutional investors, which is dependent on the resources that these institutions allocate to analyzing complex securitizations. Even larger institutions that do not lack the necessary expertise to properly assess the securitization quality may have opted not to spend these resources, over-relying on the inflated ratings of credit rating agencies (Coval, 2009). This view aligns with Martin Wheatley’s assessment, then Chief Executive of the Securities and Futures Commission of Hong Kong, regarding the disclosure documents of complex derivatives: “A 200-page prospectus that nobody ever reads and a one-page flyer with a pot of gold on it and a lot of fine print” (Cheng, 2010). To this point, many European institutions became insolvent or required bailouts due to severely underestimating the risks associated with their Not onthesame page: comprehensibility ofMBS investment… Page 3 of 37 9 purchased securitizations (to name a few: German Hypo Real Estate, Swiss UBS, and UK Royal Bank of Scotland). Second, there must exist some originators who are aware that some investors lack the necessary sophistication or the willingness to fully assess the securitization quality, especially if the prospectus is incomprehensible, and originators are willing to extract value from investors by designing less comprehensible prospectuses. Note that these originators may not (only) choose to actively decrease prospectus comprehensibility for low-quality securitizations but could also refrain from making efforts to increase prospectus comprehensibility for low-quality securitizations, thus not pruning the prospectus length and leaving the text structure in a more complex state. While we believe that this second condition is met, we make extensive efforts to rule out alternative explanations for our main results. We therefore test two alternative interpretations suggesting that low prospectus comprehensibility is merely incidental. However, we find no evidence linking low prospectus comprehensibility to the ex-ante riskiness of the underlying loans, the complexity of the securitization structure, or loan complexity. We thereby rule out that low prospectus comprehensibility is just a byproduct of complex securitization structures or loans. The complexity channel can be theoretically motivated through the existence of asymmetric information within securitization markets. As a result of asymmetric information, originators systematically selected poorly performing loans into securitizations (Jiang et al., 2014; Kruger, 2018; Purnanandam, 2011), reduced screening efforts for loans which were originated to be sold (for the originate-todistribute model, see Keys etal., 2010; Purnanandam, 2011), and monitored these loans less than those held on the originator’s balance sheet (Berndt & Gupta, 2009; Mian & Sufi, 2009; Wang & Xia, 2014). In the context of asymmetric information, the degree of securitization complexity represents the associated search costs that an investor needs to pay to assess the quality-adjusted price of the security (Ghent etal., 2019). Theoretical models show that strategically shrouding negative attributes of products can be the optimal seller behavior if unaware buyers exist in the market (Gabaix & Laibson, 2006). Furthermore, a reduction in the quality of goods is not observed by buyers if search costs outweigh the buyer’s utility for the good (Ellison & Ellison, 2009). Even though complexity is not necessarily a feature of low-quality products, products are more likely to be both more complex and of low quality if the demand for them is high, which was, e.g., the case for MBS prior to the financial crisis (Asriyan etal., 2022). Empirical studies for the pre-crisis U.S. MBS market have found evidence for a strategic obfuscation of poor securitization quality through the complexity channel: by raising investors’ search costs for low-quality securitizations, originators aimed to extract value from investors. Investors did not anticipate this behavior, as they did not demand a risk premium for buying these more complex and riskier securitizations (Furfine, 2014; Ghent etal., 2019). We specifically chose to study MBS because they are a prime example of complex financial products that are not easily understood by investors, making them particularly susceptible to strategic obfuscation and opportunistic behavior by originators in the context of asymmetric information. MBS also play a crucial role in global financial markets, serving as a key investment vehicle for institutions and a vital source of liquidity for mortgage lenders. Aggregate annual M.Hibbeln et al. 9 Page 4 of 37 issuance volumes of European MBS consistently exceed 100 billion Euros, underlining their economic importance. Furthermore, our focus on the European market allows us to investigate the impact of regulatory interventions on prospectus comprehensibility within a distinct regulatory environment. Specifically, the recent introduction of the EU Securitization Regulation (EUSR) provides a unique setting to assess how measures aimed at enhancing simplicity, transparency, and standardization affect the comprehensibility of prospectuses. We contribute to the research on complex financial securities by examining in detail the complexity of the associated disclosure documents. Specifically, we study the complexity of securitization investment prospectuses in three dimensions: type of complexity (text readability and prospectus comprehensibility), period (prevs. post-crisis), and regulatory setting (EU). In this context, we aim to answer three research questions: (I) Did originators strategically increase complexity in European MBS before the financial crisis to obfuscate low securitization quality (“complexity channel”), and was this behavior anticipated and therefore priced by investors? (II) Did originators and investors change their behavior after the financial crisis due to learning effects? (III) Do originators obfuscate securitization quality beyond the commonly used complexity measures by lowering prospectus comprehensibility, and is this behavior anticipated and therefore priced by investors? (I) Analogous to the findings of Furfine (2014) and Ghent etal. (2019) for the pre-crisis U.S. market, we investigate whether the complexity channel also existed on the European MBS market prior to the financial crisis. Building on this, we verify whether investors anticipated and therefore priced the complexity channel in European MBS. Although investors could not observe the existence of the complexity channel ex-ante, we can check from an ex-post perspective whether their pricing of complexity is consistent with originators using complexity for obfuscation. This is an important question because the U.S. and European securitization markets are fundamentally different in terms of their developments and market structure. Most notably, to date, it remains unclear to what extent originators used the originate-todistribute model in the European market. Albertazzi etal. (2015) find that Italian mortgage loans performed better if they were securitized, which they explain through originators’ reputational concerns, contradicting the concerns of the originate-to-distribute model. (II) Previous studies have not yet investigated whether the complexity channel persisted and whether investors changed their pricing behavior for the period after the financial crisis due to learning effects. In particular, because of the extensive defaults of complex securitizations in the U.S. during the crisis, investors might have recognized the high risk of these products and increased the risk premium they require for holding them. Anticipating this behavior of investors, originators might have, in turn, refrained from using complexity for obfuscation. Not onthesame page: comprehensibility ofMBS investment… Page 5 of 37 9 (III) To measure complexity, previous studies have commonly relied on proxies that account for the complexity of both (a) the securitized loan pool and (b) the securitization structure. These proxies are mostly derived from the volume of the investment prospectus (Furfine, 2014; Ghent etal., 2019). While the volume of the prospectus has been used frequently, the comprehensibility of the prospectus has not yet been investigated. This is surprising considering that the prospectus is the main disclosure document and thus the main information transmission tool of the originator toward investors. Originators might have specifically lowered prospectus comprehensibility for obfuscation, as this is relatively cheap compared to increasing structuring or loan pool complexity. To this point, it is ex-ante not clear if variation in prospectus comprehensibility is an active decision of the originator or rather simply an incidental byproduct of more general securitization complexity. While there is no research concerning the comprehensibility (as the combination of volume and text readability) of investment prospectuses, a recent strand of the literature investigates the text readability of corporate disclosure documents, mostly annual reports. These studies find that less readable and thereby less transparent annual reports are related to lower firm earnings, higher earnings volatility, and lower firm valuations by investors, suggesting that firms seek to obfuscate negative information while investors anticipate and thereby price this behavior (Hwang & Kim, 2017; Li, 2008; Lo etal., 2017). Consistent with these findings, Bonsall and Miller (2017) demonstrate that less readable financial disclosures are linked to less favorable credit ratings, higher rating disagreement, and increased costs of debt, underscoring the significant influence of disclosure readability on credit market outcomes. Regarding securitizations, Ertan etal. (2017) find that regulatorily-enforced transparency requirements lead to higher-quality loans being chosen for securitization. Confirming these results, Billio etal. (2023) find that the new EU Securitization Regulation (EUSR) adopted in 2017 is associated with significantly improved credit quality. Apart from our study, there exist two other articles studying the contents of securitization prospectuses: Zhang et al. (2023) show for U.S. securitizations issued before the financial crisis that textual contents in the risk-factor section predict subsequent losses and were not priced by investors. Debener etal. (2021) provide a first investigation of the text readability of securitization investment prospectuses; they find that low text readability impairs the ability of investors and credit rating agencies to correctly assess risk and induces greater secondary market price volatility. However, these studies neither investigate if originators lower text readability for obfuscation nor examine the level of investor pricing in relation to text readability. Methodologically, we first adopt Ghent etal. (2019) procedure regarding data items, variable construction, and sample selection to ensure that any differences in our pre-crisis results are attributable to market differences between the U.S. and European residential MBS (RMBS) markets rather than to methodological differences. Thus, we obtain the same Bloomberg data items but for European RMBS issued between 2003 and 2020. We construct one intensive measure M.Hibbeln et al. 9 Page 6 of 37 (Default) and one extensive measure (Internal Rate of Return) of security performance. We measure complexity pricing set by investors with the Credit Spread over the 3-month EURIBOR. Then, we regress these variables on deal-level complexity, together with various control variables and fixed effects for rating, year of issuance, country, and originator. In addition to proxying complexity by common measures, we use Fog Index to measure text readability, which is an established proxy for assessing the readability of financial documents (Bushee etal., 2018; Dyer etal., 2017; Li, 2008; Lo etal., 2017). We further exploit regulatorily induced cross-sectional variation in deal complexity stemming from the EUSR: On the deal-level, we use the deals’ prospectuses and documents to identify the level of adherence the respective deal has towards 48 unique regulation features relating to either simplicity, transparency, or standardization (STS). We provide broad insights into the existence, persistence, investor anticipation, regulatory influences on, and type of the complexity channel on the European RMBS market. When measuring securitization complexity with proxies commonly used in the empirical literature, we do not find any evidence in favor of originators using complexity for obfuscation. In contrast, low prospectus comprehensibility is significantly associated with greater security defaults and lower security returns, suggesting that originators obfuscated low securitization quality by lowering prospectus comprehensibility. The effects are economically meaningful: For a one standard deviation decrease in readability, the annual returns are on average 18 basis points lower when considering a prospectus with average prospectus length. For a prospectus where the prospectus length is one standard deviation above the average, this effect increases to 34 basis points. This “prospectus comprehensibility channel” does not persist for securities issued after the financial crisis, indicating that originators expected investors to either refrain from buying or demand a large risk premium for securities with low prospectus comprehensibility. However, the prospectus comprehensibility channel is only partially priced post-crisis: Investors demand a significant risk premium of 24 basis points for an additional 100 pages in prospectus length, which speaks in favor of learning effects based on the documented complexity channel for U.S. securitizations; in contrast, they do not demand a risk premium for text readability and comprehensibility of the investment prospectus. Recent regulatory interventions, most prominently in the form of the EUSR, have no significant effect on enhancing prospectus comprehensibility, even though reducing deal complexity was one of the regulations’ major goals suggesting that prospectus comprehensibility is a distinct form of complexity not yet addressed by regulators. 2 Structure oftheEuropean MBS market We study the European RMBS market, which is similar to the U.S. RMBS market in many respects. A typical deal follows the pay-through concept, meaning that prioritized senior tranches receive incoming cash flows from the underlying loans before subordinated junior tranches. Another similarity is that tranches are usually rated by more than one credit rating agency, except for the lowest-ranked equity tranches, which are often retained by the originators. Investors in these markets are Not onthesame page: comprehensibility ofMBS investment… Page 7 of 37 9 predominantly institutions such as banks, insurance companies, or funds. Similar to the U.S. market, the deal lead managers are large investment banks, which also include U.S. banks. However, there are also fundamental differences between the European and U.S. RMBS markets in terms of market structure and historical market development. Regarding the market structure, there is little government participation in the European securitization market, compared to the high relevance of government-sponsored entities in the United States (Altunbas etal., 2009). Furthermore, securitized loans on average are less risky in Europe, as subprime lending does not occur at the same scale as in the United States, and it is not common to grant no-documentation loans. Finally, in our sample of European RMBS, there is almost always a single loan pool underlying all tranches simultaneously rather than multiple loan pools underlying different series of tranches, as is often the case for U.S. RMBS deals (Ghent etal., 2019). In this regard, European RMBS are less complex. Regarding historical market development, in the United States, house prices, mortgage debt, and securitization volume increased simultaneously in the run-up to the financial crisis (Levitin & Wachter, 2012). In contrast, the pre-crisis housing boom in Europe was primarily funded by covered bonds, where the underlying loans remained on the originator’s balance sheet, with originators having an obligation to repay investors. MBS were generally less relevant for funding than in the United States, and the originate-to-distribute model was not a major contributor to the crisis in Europe (Wachter, 2015). Regardless of these differences, the general exponential growth of the European securitization market was similar to that of the U.S. market, albeit the European market lagged behind. While the U.S. private-label RMBS market collapsed from its peak at over $500 billion issuance volume in 2007 to approximately $250 billion in 2008 (Ghent etal., 2019), the European MBS market reached its peak in 2008 at almost €400 billion and collapsed to approximately €240 billion in 2009.1 The growth of securitization markets before the crisis was attributed to a high demand for safe securities from institutional investors (Altunbas et al., 2009), which was partly due to investors being rating-constrained, such as banks and insurance companies that are capital-constrained in their risk-taking through rating-based regulation. After the financial crisis, European and U.S. regulators identified an overreliance of investors on ratings in securitization markets (Coval etal., 2009). Since then, several regulations have been implemented in both markets. Most notably, in the context of securitization complexity, in 2014 and 2015, the European Banking Authority (EBA) and the Basel Committee on Banking Supervision (BCBS) proposed the concept of simple, transparent, and standardized (STS) securitizations for discussion (EBA, 2014; BCBS, 2015). The STS concept was implemented through the EU Securitization Regulation (EUSR), which was adopted in December 2017 and implemented in January 2019, with the goal of revitalizing the securitization market (EU, 2017). To achieve this goal, investors’ risk assessment should be improved by mitigating information asymmetry and reducing deal complexity. The EUSR contains legally binding minimum requirements, including basic STS features, as well 1 See FigureC.1 in the Online Appendix. M.Hibbeln et al. 9 Page 8 of 37 as optional requirements for deals to obtain the STS label. First evidence shows that investors tend to focus on the new quality label instead of the security design, but the latter is more important for originators’ behavior and ultimately for the underlying loan performance (Hibbeln & Osterkamp, 2024). We also exploit regulatorily induced cross-sectional variation in deal complexity stemming from the EUSR: On the deal-level, we use the deals’ prospectuses and documents to identify the level of adherence the respective deal has towards 48 unique regulation features relating to either simplicity, transparency, or standardization (STS). We then show that the EUSR had no significant effect on enhancing prospectus comprehensibility, even though reducing deal complexity was one of the regulations’ major goals. This suggests that prospectus comprehensibility is a distinct form of complexity, which is not significantly related to securitization and loan pool complexity and is not yet addressed by regulators. 3 Hypotheses There are several equilibrium considerations underlying our subsequent hypothesis development. On a perfect capital market with full information, originators cannot obfuscate low securitization quality through the strategic increase of complexity, thereby also making the pricing of complexity unnecessary. Deviating from this perfect capital market assumption, we assume that originators have the opportunity to obfuscate low securitization quality through strategically increasing complexity. This obfuscation behavior yields them a positive NPV, as they extract value from the investors in a zero-sum type of game. However, this value extraction only works if investors do not consider high complexity as a signal for obfuscation behavior, and thereby do not (fully) price complexity. If the investors simply recognize the possibility of complexity being used for strategic obfuscation, they will always price complexity regardless of the originator actually using it for obfuscation. This is because we assume investors ex-ante cannot distinguish when complexity is strategically used and when it is just incidental. Specifically, we develop hypotheses related to our three main research questions: (I) Did originators strategically increase complexity in European RMBS before the financial crisis to obfuscate low securitization quality (“complexity channel”), and was this behavior anticipated and therefore priced by investors? (II) Did originators and investors change their behavior after the financial crisis due to learning effects? (III) Do originators obfuscate securitization quality beyond the commonly used complexity measures by lowering prospectus comprehensibility, and is this behavior anticipated and therefore priced by investors? (I) For the pre-crisis U.S. securitization market, there is evidence for originators strategically obfuscating poor securitization quality through increased complexity. By raising investors’ search costs for low-quality securitizations, originators aimed to extract value from investors (Furfine, 2014; Ghent etal., 2019). Investors did not anticipate this use of complexity for obfuscation, meaning that they did not demand a higher risk premium for more complex securitizations. This investor pricing behavior is not directly related to originators actually using complexity for obfuscation, Not onthesame page: comprehensibility ofMBS investment… Page 15 of 37 9 variables are very high: the longest prospectus is 725 pages long, and the highest number of Glossary Terms is over 1000. Table 1 Summary statistics of deal complexity This table reports the summary statistics for our deal-level complexity variables. Unit of observation is deals. File Size is measured in megabytes. Variable definitions are given in Table8 obs mean sd min p25 p50 p75 max Deal Tranches 625 4.5 2.9 1 3 4 6 31 Pagesmpool 625 17.7 11.5 1 10 15 22 72 Pageswaterfall 625 17.3 8.0 2 12 16 20 64 Glossary Terms 625 315 158 0 223 290 406 1017 File Size (MB) 625 1.6 1.2 0.2 0.9 1.3 2.0 13.9 Total Pages 625 223 82 56 174 214 256 725 Fog Index 625 22.0 1.3 16.9 21.2 22.1 22.9 25.4 Table 2 Tranche performance and credit spread summary statistics This table presents tranche-level summary statistics for the performance variables Default and IRR, as well as the pricing variable Credit Spread. Unit of observation is tranches. Variable definitions are given in Table8. The presented rating categories refer to the rating at the time of tranche issuance Obs Mean sd p5 p25 p50 p75 p95 Panel A: Default AAA 817 1.84 AA 391 1.53 A 329 3.04 BBB 284 8.10 Total 1821 2.97 Panel B: IRR AAA 817 1.63 2.35 0.05 0.64 1.69 2.55 3.92 AA 391 1.79 1.24 0.30 1.07 1.62 2.46 3.35 A 329 2.19 1.34 0.28 1.34 1.98 2.93 4.44 BBB 284 3.20 2.10 1.02 1.93 2.50 3.90 7.20 Total 1821 2.01 2.03 0.13 1.09 1.86 2.77 4.70 Panel C: Credit Spread AAA 817 65.3 56.2 7.0 19.7 53.7 95.0 155.0 AA 391 131.7 99.6 17.0 40.8 115.0 200.0 300.0 A 329 149.2 124.5 24.0 50.0 110.0 201.9 400.0 BBB 284 204.2 179.8 45.0 78.3 150.0 253.6 600.3 Total 1821 116.4 118.2 10.0 35.0 80.1 150.3 375.0 M.Hibbeln et al. 9 Page 16 of 37 The highest Fog Index is 25.4. Using the original interpretation of Fog Index, this would imply that the reader is required to have 25.4years of formal education to understand this prospectus in the first reading. This interpretation is not particularly useful in our context, so we will refrain from interpreting Fog Index in that way and instead focus on the distribution of the variable. The mean of Fog Index (22.0) is slightly higher than the means observed in comparable finance and accounting studies (e.g., Dyer etal., 2017; Li, 2008; Lo et al., 2017; Loughran & McDonald, 2014), which report average Fog Index values for financial documents ranging from 18 to 21. While Fog Index does not greatly vary over time, we observe a notable cross-sectional variation. Figure2 exemplifies this observation: Panel A presents the subsection “Cash Collection Arrangements” for a prospectus with a low measured readability (Fog Index = 25.4). Panel B presents similar information for a prospectus with high a measured readability (Fog Index = 18.2). Table2 reports the summary statistics for our performance measures Default and IRR as well as Credit Spread. Compared to U.S. RMBS, we observe rare Defaults: Only 1.8% of AAA-rated tranches defaulted, whereas Ghent etal. (2019) report a 42% default rate for AAA-rated tranches. For initially BBB-rated tranches, the default rate was relatively high at 8.1%, considering that this is investment-grade, although it is still considerably lower than for BBB-rated U.S. RMBS (97%). A look at the IRR confirms this picture of relatively moderate European RMBS losses. The mean IRR increases along with worse initial ratings. This indicates that risk did not materialize to the extent that it was priced, which is in stark contrast to U.S. RMBS, as those securities had shrinking average IRRs with worse initial ratings. 4.4 Methodology To test our hypotheses regarding the complexity channel (H1a, H2a), we regress our performance measures Default and IRR on common complexity measures, as well as various controls and fixed effects: We estimate the probability that tranche i, which was issued at time t, defaults until the point in time T, which is the time of Default measurement (February 2021). Therefore, T–t gives the duration of observation in which a tranche default would be recorded by us. To model the probability of Default, we use the linear probability model.12 We use model (3) to estimate the continuous variable IRR by OLS regressions. We cluster standard errors at the deal level. (2) Pr (Defaulti,t,T=1|xi,j,t)=𝛼+Complexity � j,t 𝛽+Controls � i,j,t 𝛾+𝜓t+𝜓c+𝜓r+𝜓 o (3) E (IRRi,t,T|xi,j,t)=𝛼+Complexity � j,t 𝛽+Controls � i,j,t 𝛾+𝜓t+𝜓c+𝜓r+𝜓 o 12 We choose to model the probability of Default via the linear probability model because of our extensive use of fixed effects. Since our Default variable has a mean of approximately 3%, i.e., 97% of the tranches in our sample did not default, it can be considered a mild case of “rare events data” (King and Zeng, 2001). Timoneda (2021) investigates the statistical properties of various binary dependent variable models in the case of rare events data using Monte Carlo simulations. They conclude that the linear prob- Not onthesame page: comprehensibility ofMBS investment… Page 17 of 37 9 On the right-hand side of our regression equations, we include the common complexity measures described in Sect.4.2, which vary at deal level j. Furthermore, we include Deal Volume and Excess Spread as control variables that vary at the deal level j. We also include controls that vary at the tranche level i: In addition to the Subordination of the tranche, we also control for a dummy variable that equals one if at least two credit rating agencies issued diverging ratings at the time of tranche Fig. 2 Text readability examples of different prospectuses. This figure presents excerpts from the section “Cash Collection Arrangements” for two different investment prospectuses with Fog Index 25.4 (Panel A) versus 18.2 (Panel B) ability model outperforms other models such as logit in this regard. This finding is intuitive considering that the logistic model drops observations if there is no variation in the binary dependent variable within one of the fixed effect groupings. Nevertheless, as a robustness check, we estimate the default probability with a probit regression and obtain similar results. Footnote 12 (Continued) M.Hibbeln et al. 9 Page 18 of 37 issuance (Disagree Rating).13 We further control for the Credit Spread over the 3-month EURIBOR measured in basis points. All right-hand side variables are measured at the time of deal issuance t. Finally, we include various fixed effects, particularly for the year of issuance ψt, country of collateral ψc, rating at issuance ψr,14 and originator ψo.15 We provide detailed variable definitions in Table8. To test our complexity pricing hypotheses (H1b, H2b), we apply a similar regression model, except that Credit Spread is no longer among the control variables: To test our hypotheses regarding the prospectus comprehensibility channel (H3a, H3b, H4a, and H4b), we estimate similar models as Eqs.(2)–(4): In these models, we change the measurement of complexity to include both dimensions of prospectus comprehensibility, Total Pages and Fog Index, separately, as well as combined in an interaction term. Total Pages is an extensive measure of prospectus comprehensibility, as the prospectus length increases the time it takes an investor to fully understand the details of the deal. Fog Index, as a text readability proxy, can be interpreted as an intensive measure of prospectus comprehensibility, which raises the time the investors need to understand one given page of the prospectus. We include the interaction term Total Pages × Fog Index to quantify the effort required by an investor to understand the prospectus. While any given page of a prospectus with a high Fog Index might be difficult to understand, it will not require great effort to understand it if the prospectus is short. Conversely, even long prospectuses can be relatively easy to understand if the language used is clear (4) E (CreditSpreadi,t|xi,j,t)=𝛼+Complexity � j , t 𝛽+Controls � i , j , t 𝛾+𝜓t+𝜓c+𝜓r+𝜓 o (5) Pr(Default i,t,T =1|x i,j,t )=𝛼+TotalPages j,t 𝛽 1 +Fog j,t 𝛽 2 + TotalPages ×Fogj,t𝛽3+Controls� i , j , t 𝛾+𝜓t+𝜓c+𝜓r+𝜓 o (6) E(IRR i,t,T |x i,j,t )=𝛼+TotalPages j,t 𝛽 1 +Fog j,t 𝛽 2 + TotalPages ×Fogj,t𝛽3+Controls� i , j , t 𝛾+𝜓t+𝜓c+𝜓r+𝜓 o (7) E(CreditSpread i,t |x i,j,t )=𝛼+TotalPages j,t 𝛽 1 +Fog j,t 𝛽 2 + TotalPages ×Fogj,t𝛽3+Controls� i,j,t 𝛾+𝜓t+𝜓c+𝜓r+𝜓 o 14 Rating at issuance is defined as the best rating that one of the three rating agencies issued. Disagree Rating, therefore, indicates a worse rating being issued by another rating agency. 15 We extend the control variables of Ghent etal. (2019) to include originator fixed effects. We rationalize this choice in Appendix D of our Online Appendix. When regressing our complexity variables solely on originator fixed effects, we observe R2s of up to 80%, indicating that complexity mostly varies between originators. 13 Disagree Rating corresponds to Ghent etal. (2019)’s variable disagreetranche. As we lack the distinction of different loan groups, we cannot construct Ghent etal. (2019)’s control variable crosscollateralization. We further do not control for the total yearly issuance volume of the lead manager (leadtot) but instead control for originator fixed effects. Not onthesame page: comprehensibility ofMBS investment… Page 19 of 37 9 and concise. However, if both text readability is low and prospectus length is high, understanding the prospectus would be a major obstacle even for sophisticated investors. 5 Main results 5.1 Performance andpricing ofcommon complexity measures For all subsequent analyses, we split the sample into a pre-crisis and a post-crisis sample (tranches issued between 2003 and 2007 and those issued between 2008 and 2020). In Table3, we report the regressions of our extensive and intensive performance measures Default and IRR on the common complexity variables and analyze the pricing of common complexity through Credit Spread. For MBS issued before the financial crisis, we hypothesize that originators obfuscated poor securitization quality by increasing complexity (H1a). If this were the case, we would expect more complex MBS to perform worse, even after controlling for potential confounding factors that affect both securitization performance and complexity. We test this hypothesis in Table3, Panel A, columns 1–6. When we measure performance by using our extensive measure of Default (columns 1–3), we find almost no evidence in favor of this complexity channel: None of our common complexity measures are significantly related to higher tranche Defaults pre-crisis, except for a slight positive coefficient of Total Pages (column 1). When we instead measure performance by using the intensive measure of IRR (columns 4–6), we still do not observe a significant relationship between most of our common complexity variables and performance. The only common complexity variable associated with lower returns is Total Pages, which we will investigate in detail in the next section. Taken together, our results for Default and IRR do not support the existence of the complexity channel in the pre-crisis period (H1a) when proxying complexity with common measures. Disagree Rating is associated with significantly worse performance pre-crisis (columns 1–6), as Rating at Issuance fixed effects only include the best rating assigned. If a different rating agency issued a worse rating, the Default probability increases by more than 6 percentage points, indicating the usefulness of such information for default prediction. Beyond the credit rating, investors did not incorporate useful information in their pricing decision: Higher Credit Spread is not significantly associated with an increased Default probability (columns 1–3). Furthermore, an additional basis point of Credit Spread is associated with 0.7 basis points higher IRR. Therefore, the risk priced through the spread has not fully materialized, as would be indicated by a coefficient of 0. Next, we check if investors anticipated the originators using complexity for obfuscation and thus required a risk premium for complexity in the pre-crisis period (H1b). We cannot confirm this hypothesis based on columns 7–9, as there is almost no significant relation between Credit Spread and any of our common complexity variables. For Glossary Terms, we even find a weak negative relation. From an M.Hibbeln et al. 9 Page 20 of 37 ex-post perspective, the lack of common complexity pricing was the correct decision, as more complex MBS did not perform worse pre-crisis. Subsequently, we focus on tranches issued during 2008–2020 (“post-crisis”). We test if originators use the complexity channel post-crisis, despite perhaps fearing being punished by investors through increased risk premia (H2a). When estimating our regressions of Default and IRR on our common complexity variables for the post-crisis sample, we find little evidence for worse performance of complex MBS. Though the coefficients for Pagesmpool are statistically significant for both Default (column 10) and IRR (column 13), the economic significance is low, with an additional page of Pagesmpool reducing annual returns by just 1.3 bps. The only notable exception is Total Pages, which remains significantly negatively related to IRR (column 14). Except for Total Pages, we cannot confirm H2a (columns 10–15). Concerning the post-crisis investor pricing of complexity, investors may have become increasingly complexity-averse, thus starting to require a risk premium for more complex MBS (H2b). Even though we find little evidence that originators used complexity for quality obfuscation in Europe, investors may still apply the lesson they learned on the U.S. market, where the complexity channel did exist pre-crisis (., 2019). Moreover, regulators communicated that the high complexity of securitizations and their poor performance during a crisis are related, which may have lead investors becoming more complexity-averse (EBA, 2014). We find that in line with H2b, common complexity is generally priced post-crisis: A one standard deviation increase in PC1 results in a 23bp increase in Credit Spread (p < 1%, see column 18). Complexity related to the number of Glossary Terms is priced to a large extent, with 100 additional terms increasing the spread by approximately 15 bps (column 16). Total Pages is statistically significant in model (17), with a 100-page increase in prospectus length associated with a 24 bps increase in Credit Spread. Overall, this provides evidence in favor of investors starting to price common complexity after the financial crisis (H2b). Ex-post, this pricing decision is not justified by performance, considering that neither prenor post-crisis, the common complexity was related to significantly worse MBS performance. Still, from an ex-ante equilibrium perspective, investors cannot distinguish if complexity is just incidental or a strategic obfuscation behavior of the originator. In particular, only because the originators did not use common complexity measures for obfuscation in the past, this does not mean that they will not start using it at some point. Therefore, this pricing behavior of common complexity by investors can be rationalized. Contrary to the pre-crisis results, the coefficient for Disagree Rating is highly significant and positive post-crisis (columns 16–18). Investors now take deviating ratings into consideration and demand a risk premium of approximately 43 bps, which is rational considering that tranches historically performed worse if a deviating rating agency assigned a worse rating at tranche issuance. In summary, European originators did not significantly increase common complexity measures to obfuscate poor RMBS quality, either before or after the financial crisis, which is different from findings for pre-crisis issues in the U.S. The only exception is the variable Total Pages, where we find some indications for originators using prospectus length to obfuscate low securitization quality. While investors did not require a risk premium for common complexity before the crisis, they started Not onthesame page: comprehensibility ofMBS investment… Page 21 of 37 9 Table 3 Default, IRR, and Credit Spread regressions on common complexity measures Panel A: Pre-crisis issuances Default IRR Credit Spread (1) (2) (3) (4) (5) (6) (7) (8) (9) Deal Tranches − 0.001 (− 0.238) 0.015 (0.373) 4.116+ (1.845) Pagesmpool 0.003 (1.299) − 0.006 (− 0.721) 0.616 (1.410) Pageswaterfall − 0.000 (− 0.097) − 0.006 (− 1.046) − 0.083 (− 0.172) GlossaryTerms − 0.030 (− 1.319) − 0.091 (− 1.239) − 10.230* (− 2.190) File Size − 0.016 (− 1.616) 0.057 (1.489) − 2.730 (− 1.256) Total Pages 0.099+ (1.798) 0.058 (1.563) − 0.302* (− 2.011) − 0.379** (− 2.715) − 0.567 (− 0.046) − 14.711 (− 1.424) PC1 0.003 (0.130) − 0.063 (− 0.550) − 4.111 (− 0.734) DisagreeRating 0.067** (3.214) 0.064** (3.071) 0.064** (3.063) − 0.264** (− 2.722) − 0.254** (− 2.765) − 0.250** (− 2.693) 5.796 (0.846) 4.372 (0.644) 4.611 (0.679) Credit Spread − 0.008 (− 0.253) − 0.006 (− 0.178) − 0.007 (− 0.218) 0.007*** (3.894) 0.007*** (3.923) 0.007*** (3.929) Observations 607 607 607 607 607 607 607 607 607 Adjusted R20.226 0.228 0.226 0.347 0.350 0.346 0.450 0.448 0.447 Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Rating FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Year Issue FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Country FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Originator FE Yes Yes Yes Yes Yes Yes Yes Yes Yes M.Hibbeln et al. 9 Page 22 of 37 Table 3 (continued) Panel B: Postcrisis issuances Default IRR Credit Spread (10) (11) (12) (13) (14) (15) (16) (17) (18) Deal Tranches 0.008+ (1.952) − 0.001 (− 0.039) 2.696 (1.183) Pagesmpool 0.002* (2.551) − 0.013* (− 2.079) 0.117 (0.363) Pageswaterfall 0.004 (1.633) 0.003 (0.400) − 1.362 (− 1.553) GlossaryTerms − 0.021* (− 2.005) 0.019 (0.363) 15.389** (3.106) File Size − 0.005 (− 0.712) 0.112 (1.387) − 1.668 (− 0.379) Total Pages − 0.021 (− 1.153) − 0.010 (− 0.680) − 0.297+ (− 1.906) − 0.266* (− 2.087) 10.741 (0.893) 24.135* (2.075) PC1 0.003 (0.259) − 0.175+ (− 1.919) 22.771** (2.862) DisagreeRating 0.010 (0.725) 0.013 (0.923) 0.014 (0.948) − 0.109 (− 0.922) − 0.135 (− 1.199) − 0.127 (− 1.111) 42.359*** (4.585) 43.829*** (4.674) 42.983*** (4.605) Credit Spread − 0.005 (− 0.907) − 0.006 (− 1.030) − 0.007 (− 1.088) 0.011*** (16.479) 0.011*** (16.324) 0.011*** (16.439) Observations 1214 1214 1214 1214 1214 1214 1214 1214 1214 Adjusted R20.219 0.197 0.197 0.565 0.565 0.565 0.662 0.657 0.659 Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Rating FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Year Issue FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Not onthesame page: comprehensibility ofMBS investment… Page 23 of 37 9 Table 3 (continued) Panel B: Postcrisis issuances Default IRR Credit Spread (10) (11) (12) (13) (14) (15) (16) (17) (18) Country FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Originator FE Yes Yes Yes Yes Yes Yes Yes Yes Yes This table shows OLS estimates from regressing Default, IRR, and Credit Spread on common complexity variables and controls. IRR is measured in percentage; Credit Spread is measured in basis points. Unit of observation is tranches. Panel A reports the results for tranches issued in 2003–2007 (“pre-crisis”); Panel B reports the results for tranches issued in 2008–2020 (“post-crisis”). Variable definitions are given in Table8. Standard errors are clustered at the deal level. t statistics are given in parentheses. ***, **, *, and + denote statistical significance at the 0.1%, 1%, 5%, and 10% levels, respectively M.Hibbeln et al. 9 Page 24 of 37 pricing it afterward. This may be due to investors becoming complexity-averse due to learning effects. This may be as they (1) observed extensive defaults of complex U.S. securities during the financial crisis, and (2) got communicated by regulators that the high complexity of securitizations and their poor performance during a crisis are related. This pricing behavior, though ex-post inconsistent with realized risk, can be rationalized from an ex-ante equilibrium perspective, as investors of a given securitization cannot observe if its high complexity is strategically used for obfuscation or simply incidental. 5.2 Performance andpricing ofprospectus comprehensibility In this section, we analyze whether originators used prospectus comprehensibility, which is a combination of prospectus length and text readability per page, to obfuscate poor securitization quality and whether low comprehensibility was priced by investors. As in the previous section, we split the sample into a pre-crisis sample (2003–2007) and a post-crisis sample (2008–2020). In Table4, we report the regressions of our extensive and intensive performance measures Default and IRR on the prospectus comprehensibility variables and analyze the pricing of prospectus comprehensibility through Credit Spread. For these analyses, we demean Fog Index and Total Pages to facilitate interpretation. For MBS issued before the financial crisis, we hypothesize that originators lowered text readability in combination with increasing prospectus length to obfuscate poor securitization quality (H3a). If this was the case, we would expect less comprehensible MBS to perform worse. We test this hypothesis in Table4, Panel A, columns 1 and 2. The coefficients are significantly positive at the 5% level (column 3). Due to the demeaning, the coefficient for Fog Index gives the marginal effect that Fog Index has on Default for a prospectus with an average number of pages: an increase of one unit in Fog Index corresponds to an increase of 5.2 percentage points in Default probability for a prospectus with an average of 223 pages. For a prospectus where Total Pages is one standard deviation above the average (223 + 82 = 305 pages), this effect increases to 5.2 + 5.6 × 0.82 = 9.8 percentage points, which is of high economic significance. Conversely, the effect of Total Pages on Default is also statistically and economically significant: A 100-page increase in Total Pages is associated with a 7.7 percentage point increase in Default probability for a prospectus with an average Fog Index. These results provide the first evidence in favor of the existence of the pre-crisis prospectus comprehensibility channel (H3a). This finding is further supported by the regression of IRR on prospectus comprehensibility: In column 2, for a one standard deviation increase in Fog Index, IRR is 0.135 × 1.3 = 0.18 percentage points lower when considering a prospectus with average Total Pages. For a prospectus where the number of pages is one standard deviation above the average, this effect increases to –0.135 × 1.3 – 0.153 × 0.82 × 1.3 = –0.34 percentage points. The marginal effect of Total Pages on IRR is even stronger: For the average Fog Index of 22, a 100-page increase in Total Pages is associated with a 0.44 percentage point decrease in annual returns. In total, this provides strong evidence in favor of Not onthesame page: comprehensibility ofMBS investment… Page 31 of 37 9 cannot observe their performance over their entire lifespan. Defaults or large principal losses that lower the IRR might only occur in later periods of the tranche lifespan. To account for this measurement error, we construct the variable IRR 5Y, which is defined as the internal rate of return during a 5-year holding period after tranche issuance. Tranches for which we cannot observe the full five-year period of cash flows (those issued after 1st of May 2016) are dropped from our sample. We report the results in Tables E.2.1 and E.2.2. Panel A of Table E.2.1 only provides evidence in favor of the pre-crisis complexity channel for variables related to prospectus length (Glossary Terms and Total Pages). In contrast, and confirming our previous results, we find that low prospectus comprehensibility is negatively related to tranche IRR 5Y (Panel A of Table E.2.2). The coefficient magnitudes remain almost identical to those reported in the baseline IRR regression (Table4, column 2). Third, we check if our results are robust to an alternative readability measure. Initially, we use the Fog Index, as it is an established proxy to measure the readability of financial documents (Bushee etal., 2018; Dyer etal., 2017; Li, 2008; Lo etal., 2017). As an alternative, we compute the Smog Index, which is computed using only the fraction of complex words (words with three syllables or more scaled by the number of sentences) as an input. Our results remain completely robust to this change. Particularly regarding the pre-crisis prospectus comprehensibility channel, we now observe that an increase of one unit in Smog Index corresponds to an increase of 8.5 percentage points in Default probability for a prospectus with an average of 223 pages (column 3 of Table E.3). Fourth, we include year-quarter fixed effects instead of year fixed effects. We confirm the non-existence of the common complexity channel, and we still find strong evidence in favor of the pre-crisis prospectus comprehensibility channel (Table E.4.1 and Table E.4.2). Fifth, we rerun all our Default regressions using the probit model instead of the linear probability model, confirming our pre-crisis results.16 Sixth, we repeat our regressions regarding our common complexity measures (Sect. 5.1) without including originator fixed effects to allow for a direct comparison with Ghent etal.’s (2019) results regarding pre-crisis issues. Our results largely remain similar to our main results. There is still little evidence in favor of the common complexity channel, both pre-crisis and post-crisis (see Table F.1 and Table F.2). The exceptions are the variables Total Pages and Glossary Terms, which are associated with higher Default and lower IRR in the pre-crisis period (columns 5 and 7 in both Tables F.1 and F.2). Similarly, we observe no pricing of complexity for the pre-crisis period (see Table F.3, Panel A). For post-crisis pricing, we confirm our previous results, although the coefficient of PC1 decreases to 15, indicating that a one standard deviation increase in complexity is associated with a 15 bps increase in Credit Spread when taking variation between originators into account (see Table F.3, column 16). 16 As our Default variable has very low variation post-crisis (only 6 of 1257 tranches defaulted), we cannot run probit regressions for the post-crisis period. The pre-crisis probit results are available upon request. M.Hibbeln et al. 9 Page 32 of 37 7 Conclusion We investigate whether originators of European RMBS lower the comprehensibility of disclosure documents to obfuscate low securitization quality. Relying on commonly used, mainly volume-based complexity proxies, we find no evidence that originators increase complexity to obfuscate low securitization quality. However, we argue that the combination of text readability (which measures the difficulty of reading a given page) and prospectus length is decisive for measuring the level of difficulty in reading and understanding disclosure documents. Implementing this measure of comprehensibility, we find evidence for originators obfuscating low securitization quality. The effects are economically meaningful: For a one standard deviation decrease in readability, the annual returns are on average 18 basis points lower when considering a prospectus with average prospectus length. For a prospectus where the prospectus length is one standard deviation above the average, this effect increases to 34 basis points. Although investors did not anticipate originators using complexity for obfuscation prior to the financial crisis, they changed their pricing behavior thereafter, demanding a risk premium of up to 23 basis points for a one standard deviation increase in common complexity. The “prospectus comprehensibility channel”, however, is only partially priced: While investors demand a significant risk premium for prospectus length, they do not demand any risk premium for low text readability. The recent regulatory intervention in the form of the EUSR has no significant effect on enhancing prospectus comprehensibility, even though reducing deal complexity was one of the regulation’s declared goals, suggesting that prospectus comprehensibility is a distinct form of securitization complexity. Our results have important implications for originators, investors, and regulators. Originators should try to avoid designing long prospectuses, as prospectus length and the number of terms in the glossary are associated with a strong increase in Credit Spread. In addition, investors should consider not only the length of the prospectus but also its text readability because originators might be trying to obfuscate low-quality securitizations through both dimensions of comprehensibility. For investor protection, regulators should also consider explicitly enhancing prospectus comprehensibility. On average, the prospectus length more than doubled between 2003 and 2019 to approximately 300 pages without improving the measured readability per page, which highlights that the low comprehensibility of MBS investment prospectuses is still a major obstacle. Appendix Table7presents the distinction between our “common complexity” and “prospectus comprehensibility” measures. The common complexity variables are mostly volume-based measures derived from the investment prospectus. For prospectus Not onthesame page: comprehensibility ofMBS investment… Page 33 of 37 9 comprehensibility, we focus on the joint effect of text readability and prospectus length. Precise variable definitions are given in Table8 Table 7 Types of complexity Type of complexity Description Panel A: Common complexity measures Prospectus volume Proxies derived from the prospectus, that attempt to measure the complexity in the securitized loan pool and the securitization structure. These measures include the number of pages (total and of specific sections), the number of glossary terms, and the file size of the prospectus Number of tranches Number of tranches in a deal as a proxy for complexity in the securitization structure Panel B: Text readability and prospectus comprehensibility Text readability Text readability of an average page in the investment prospectus of a given deal, measured by Fog index (or Smog index). A higher Fog index indicates a lower text readability Text readability × Prospectus length The interaction of text readability and prospectus length (measured by the number of pages of the investment prospectus) can be a factor that influences prospectus comprehensibility beyond the individual measures. For example, if the investment prospectus is very short, institutional investors should have relatively little trouble investing the associated search costs to understand the contents, even if the prospectus text readability is low. However, for long prospectuses, low text readability may be a major obstacle for fully understanding the securitization’s legal terms. A higher value of the interaction term Fog#TotalPages indicates a lower comprehensibility M.Hibbeln et al. 9 Page 34 of 37 Table 8 Variable definitions Variable Definition Panel A: Performance (tranche or deal level) Default Binary variable that is equal to one if the current rating (measured in February 2021) signals a tranche default or there are any principal losses greater than zero recorded in Bloomberg. An S&P rating of CCC + or lower, a Moody’s rating of Caa1 or lower, or a Fitch rating of CCC or lower is defined as an indicator for default IRR Internal rate of return based on the cash flows toward the tranche until April 2021. The tranche is assumed to have been bought at par and any remaining principal left outstanding was paid back in full in June 2021. IRR is measured in percentage IRR 5Y Internal rate of return based on the cash flows that the tranche received until 5years after tranche issuance. The tranche is assumed to have been bought at par and any remaining principal left outstanding was paid back in full 5years after tranche issuance. IRR 5Y is measured in percentage Collateral Loss Share Deal-level principal losses (as observed in February 2021) divided by Deal Volume at time of deal issuance, measured in percentage Panel B: Pricing (tranche level) Credit Spread Credit spread over the 3-month EURIBOR measured at tranche issuance in basis points Panel C: Complexity (deal level) Deal Tranches The total number of tranches within a deal Pagesmpool The number of pages in the prospectus describing the pool of the underlying mortgage loans Pageswaterfall The number of pages in the prospectus describing the cash flow allocation to the tranches (waterfall mechanism) Glossary Terms The total number of terms in the glossary of the prospectus File Size The file size of the prospectus measured in megabytes Total Pages The total number of pages of the prospectus PC1 First principal component of the variables: Deal Tranches, Pagesmpool, Pageswaterfall, Glossary Terms, File Size, Total Pages. Factor loadings are as follows: Total Pages (0.55), Glossary Terms (0.51), Pagesmpool (0.41), File Size (0.38), Deal Tranches (0.32) and Pageswaterfall (0.14). PC1 is standardized to a standard deviation of one Fog Index Linear combination of the average number of words per sentence and the fraction of complex words (fraction of words with three syllables or more): Fog Index = (average sentence length + fraction of complex words) × 0.4 Smog Index Smog Index is calculated by: Smog Index =1.043 × √ complex words ×30 number of sentences + 3.1291 , where a “complex word” is defined as a word with three syllables or more STS Features The total number of simplicity, transparency, and standardization (STS) features that a specific deal adheres to. Takes a maximum value of 48 if a deal adheres to all 48 STS features Not onthesame page: comprehensibility ofMBS investment… Page 35 of 37 9 Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1007/ s1114702509213-8. Funding Open Access funding enabled and organized by Projekt DEAL. This study was funded by Deutsche Forschungsgemeinschaft (Grant no.: 434130478) by Martin Thomas Hibbeln. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/ licenses/by/4.0/. References Albertazzi, U., Eramo, G., Gambacorta, L., & Salleo, C. (2015). Asymmetric information in securitization: An empirical assessment. Journal of Monetary Economics, 71, 33–49. Altunbas, Y., Gambacorta, L., & Marques-Ibanez, D. (2009). Securitization and the bank lending channel. European Economic Review, 53, 996–1009. Asriyan, V., Foarta, D., & Vanasco, V. (2022). The good, the bad and the complex: Product design with imperfect information. American Economic Journal: Microeconomics, 15, 187–226. BCBS, and IOSCO. (2015). Criteria for Identifying Simple, Transparent and Comparable Securitisations. Basel Committee on Banking Supervision and International Organization of Securities Commissions. Available online: https:// www. bis. org/ bcbs/ publ/ d441. pdf. Table 8 (continued) Variable Definition Simplicity/Transparency Features The total number of simplicity (transparency) features that a specific deal adheres to. Takes a maximum value of 17 (5) if a deal adheres to all 17 (5) simplicity (transparency) features Panel D: Controls (tranche level) Subordination The fraction of the deal volume that can default before the tranche suffers first losses Disagree Rating Binary variable that is equal to one if at least two credit rating agencies issued diverging ratings at the time of tranche issuance Rating The best rating that one of the three rating agencies (S&P, Fitch, and Moody’s) assigned at the time of tranche issuance Panel E: Controls (deal level) Deal volume Volume of the deal measured at the time of deal issuance in billions of Euros Excess spread The excess coupon that the loans underlying the deal pay minus the value-weighted first average coupon that the securities within the deal pay Year of issuance Year in which the deal was issued Country of collateral Country in which the underlying mortgage loans were originated Originator Originator of the underlying mortgage loans M.Hibbeln et al. 9 Page 36 of 37 Berndt, A., & Gupta, A. (2009). Moral hazard and adverse selection in the originate-to-distribute model of bank credit. Journal of Monetary Economics, 56, 725–743. Billio, M., Dufour, A., Segato, S., & Varotto, S. (2023). Complexity and the default risk of mortgagebacked securities. Journal of Banking and Finance, 155, 1–19. Bonsall, S. B., & Miller, B. P. (2017). The impact of narrative disclosure readability on bond ratings and the cost of debt. Review of Accounting Studies, 22, 608–643. Bozanic, Z., Kraft, P., & Tillet, A. (2023). Qualitative disclosure and credit analysts’ soft rating adjustments. European Accounting Review, 32, 779–807. Bushee, B. J., Gow, I. D., & Taylor, D. J. (2018). Linguistic complexity in firm disclosures: Obfuscation or information? Journal of Accounting Research, 56, 85–121. Cheng, J. (2010). The fine print is enlarged, but will investors read it? The Wall Street Journal, May 31. Available online: https:// www. wsj. com/ artic les/ SB100 01424 05274 87039 57604 57527 25102 63503 010. Coval, J., Jurek, J., & Stafford, E. (2009). The economics of structured finance. Journal of Economic Perspectives, 23, 3–25. Debener, J., Fenner, A., Klein, P., Ongena, S. (2021). Linguistic complexity in ABS prospectuses—Evidence from european securitization data. Working Paper. Dyer, T., Lang, M., & Stice-Lawrence, L. (2017). The evolution of 10-K textual disclosure: Evidence from latent dirichlet allocation. Journal of Accounting and Economics, 64, 221–245. EBA. (2014). EBA Discussion Paper on Simple, Standard and Transparent Securitisations. European Banking Authority. Available online: https:// www. eba. europa. eu/ regul ationandpolicy/ secur itisa tionandcover edbonds/ techn icaladviceonsimplestand ardandtrans parentsecur itisa tions. Ellison, G., & Ellison, S. F. (2009). Search, obfuscation, and price elasticities on the internet. Econometrica, 77, 427–452. Ertan, A., Loumioti, M., & Wittenberg-Moerman, R. (2017). Enhancing loan quality through transparency: Evidence from the European central bank loan level reporting initiative. Journal of Accounting Research, 55, 877–918. EU. (2017). Regulation (EU) 2017/2402 Laying Down a General Framework for Securitisation and Creating a Specific Framework for Simple, Transparent and Standardised Securitisation. European Union. Available online: https:// eurlex. europa. eu/ legalconte nt/ EN/ TXT/? uri= CELEX% 3A320 17R24 02. Furfine, C. H. (2014). Complexity and loan performance: Evidence from the securitization of commercial mortgages. Review of Corporate Finance Studies, 2, 154–187. Gabaix, X., & Laibson, D. (2006). Shrouded attributes, consumer myopia, and information suppression in competitive markets. Quarterly Journal of Economics, 121, 505–540. Ghent, A. C., Torous, W. N., & Valkanov, R. I. (2019). Complexity in structured finance. Review of Economic Studies, 86, 694–722. Griffin, J., Lowery, R., & Saretto, A. (2014). Complex securities and underwriter reputation: Do reputable underwriters produce better securities? Review of Financial Studies, 27, 2872–2925. Gunning, R. (1952). The technique of clear writing. McGraw-Hill: 36–37. Hibbeln, M. T., & Osterkamp, W. (2024). Simple is simply not enough—features versus lables of complex financial securities. Review of Derivatives Research, 27, 113–150. Hwang, B.-H., & Kim, H. H. (2017). It pays to write well. Journal of Financial Economics, 124, 373–394. Jiang, W., Nelson, A. A., & Vytlacil, E. (2014). Securitization and loan performance: ex ante and ex post relations in the mortgage market. Review of Financial Studies, 27, 454–483. Keys, B. J., Mukherjee, T., Seru, A., & Vig, V. (2010). Did securitization lead to lax screening? Evidence from subprime loans. The Quarterly Journal of Economics, 125, 307–362. King, G., & Zeng, L. (2001). Logistic regression in rare events data. Political Analysis, 9, 137–163. Kruger, S. (2018). The effect of mortgage securitization on foreclosure and modification. Journal of Financial Economics, 129, 586–607. Levitin, A. J., & Wachter, S. M. (2012). Explaining the housing bubble. Georgetown Law Journal, 100, 1177–1258. Li, F. (2008). Annual report readability, current earnings, and earnings persistence. Journal of Accounting and Economics, 45, 221–247. Lo, K., Ramos, F., & Rogo, R. (2017). Earnings management and annual report readability. Journal of Accounting and Economics, 63, 1–25. Loughran, T., & McDonald, B. (2014). Measuring readability in financial disclosures. Journal of Finance, 69, 1643–1671. Not onthesame page: comprehensibility ofMBS investment… Page 37 of 37 9 Mian, A., & Sufi, A. (2009). The consequences of mortgage credit expansion: Evidence from the U.S. mortgage default crisis. The Quarterly Journal of Economics, 124, 1449–1496. Purnanandam, A. (2011). Originate-to-distribute model and the subprime mortgage crisis. Review of Financial Studies, 24, 1881–1915. SEC. (1998). A plain english handbook. Available online: https:// www. sec. gov/ pdf/ handb ook. pdf Timoneda, J. C. (2021). Estimating group fixed effects in panel data with a binary dependent variable: How the LPM outperforms logistic regression in rare events data. Social Science Research, 93, Article 102486. Wachter, S. M. (2015). The housing and credit bubbles in the United States and Europe: A comparison. Real Estate Economics, 47(S1), 37–42. Wang, Y., & Xia, H. (2014). Do lenders still monitor when they can securitize loans? The Review of Financial Studies, 27, 2354–2391. Zhang, H. H., Zhao, F., & Zhao, X. (2023). Neglected risks in the communication of residential mortgage-backed securities offerings. Journal of Finance, 79, 129–172. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.