We estimate the models progressively. First, we present the results for the full MBS sample. Subsequently, we provide estimations for the RMBS sample to test the robustness of our results with a uniform sample. We then split the sample into two groups according to risk categories—prime (AAA) tranches and non-prime (non-AAA) tranches, to examine whether issuer frequency effects differ depending on the level of risk taken by the investors.
We present the results for the broader MBS sample in Table 3, Panel A. Estimations for the baseline model are shown in column 1 and we include the interaction variables (Frequent issuer × CRA reported, Frequent issuer × Distance and Frequent issuer × Boom) separately in columns 2 to 4.
We find that the coefficients of Frequent issuer are negative and statistically significant at least at the 5% level in all models. MBS from frequent issuers carry lower spreads as investors evaluate these notes as relatively less risky. This result, supporting H1, shows that investors value frequent issuers and consider that the reputation concerns of these issuers should mitigate opportunistic behaviour. The coefficients for the number of CRA reported are not statistically significant, apart from 2 CRA reported in column 1, albeit only at 10% level. Results show that MBS tranches are not priced higher when only one or two credit ratings are reported in comparison to tranches where credit ratings from all three rating agencies are reported. We do not find a significant coefficient for 3 CRA reported in columns 3 and 4.Footnote 26 Overall, the results do not support H3.
Distance is statistically significant at the 5% level and has a positive sign (apart from column 4). In support of H5, we find that MBS issued by foreign banks carry a higher spread when compared to issuances by domestic banks. Therefore, investors consider MBS issued by foreign banks to be riskier. They value local issuer expertise, where it is expected that domestic banks would be more specialised due to their familiarity with the local market. Thus, investors perceive domestic banks to be more likely to detect borrower misrepresentation and, therefore, extend safer loans. MBS originated by foreign banks are deemed to be relatively less creditworthy possibly due to information asymmetries created by bank-borrower distance.
In columns 2–4, we interact Frequent issuer with Boom, 3 CRA reported and Distance, respectively. Frequent issuer \(\times \) Boom is significant at the 5% level and has a negative sign (column 2). MBS sold by frequent issuers during the credit boom period (2005-June 2007) in the run-up to the financial crisis were regarded to be relatively less risky compared to MBS they sold in normal periods. This result, supporting H2, indicates that investors perceived frequent issuers to be more reliable and trustworthy originators of high quality MBS during the progressive phase of the credit expansionary period when information asymmetries in the markets increased. Assuming securitization follows a repeated game structure, frequent issuers are more likely to be concerned about improving their reputation as competition for market share increases during the expansion phase of the credit cycle. Consequently, they are likely to be more diligent at the credit underwriting stage during these periods. Such issuers are also more likely to provide effective monitoring in an intensely competitive environment as smaller issuers would be more concerned with maintaining or increasing market share (Winton and Yerramilli 2020).
We also find a negative and significant (at the 10% level) coefficient for Frequent issuer \(\times \) 3 CRA reported. MBS tranches where a frequent issuer reports ratings from three credit rating agencies are regarded as less risky. This shows that the combination of frequent issuance with a clear indication of transparency by reporting three ratings is highly valued by investors, supporting H4. We do not find Frequent issuer \(\times \) Distance to be significant. It seems that, contrary to our expectations in H6, frequent issuance does not have a mitigating effect on the information asymmetries caused by distance to the origination market.
Retained is not significant in any of the specifications. Retention as an alignment device seems to have lost its importance since it does affect issuers’ borrowing costs. This result may also suggest that investors cannot rely on this indicator as retained tranches could be sold by the issuer. Ratings/Tranches is not significant in any of the models while Subordination is significant in all of the models. It seems that credit ratings do not completely capture the leverage effects within deals and higher subordination typically signals higher risk deals. Weighted Average Life is a key determinant of initial spreads as this variable is highly statistically significant and consistently positive in all specifications in Table 3, Panel A. This finding is consistent with Cuchra (2004) where initial launch spreads were persistently positively related to effective maturity. Liquidity, proxied by Size, is significant in all the models. In particular, we find that Size is now statistically significant and has a negative sign. This shows that investors require lower liquidity premiums for larger issues. With regards to collateral, spreads on RMBS notes were lower than initial funding costs associated with CMBS notes. This is because CMBSs are less regulated, less standardised and attract a higher risk weighting.Footnote 27
RMBS constitutes 81.27% of our sample. We run estimations on the RMBS subsample as it is more homogenous and can help to check the robustness of our reported results for the whole sample. The results are presented in Table 3, Panel B. We find that almost all the relationships established above for our main variables remain unchanged in the RMBS sample. We still find that frequent issuance leads to lower spreads. The possibility of rating shopping, shown by the positive coefficient of 2 CRA reported, is deemed risky by investors. The statistical significance of the Distance variable gets stronger. This is unsurprising as residential mortgage lending requires more local presence and expertise by the lenders and, as the literature argues, foreign banks may be at a disadvantage relative to local banks. The direction of the signs and significance of the interaction variables –Frequent issuer \(\times \) Boom, Frequent issuer \(\times \) 3 CRA reported and Frequent issuer \(\times \) Distance– do not change.
Prime versus non-prime tranches
We split the sample into two groups according to risk categories—prime (AAA) tranches and non-prime (non-AAA) tranches—to examine whether frequent issuer effects differ depending on investors’ risk preference. Results for the prime sample are presented in Table 4, Panel A. Broadly, we find similar results for the AAA tranches, which are deemed to be least risky. Our main variable Frequent issuer is still significant. Regarding the rating shopping hypothesis, we do not find any CRA reported variables to be significant. One different observation is the coefficient of the Distance variable, which is now not significant. We also do not find any significance for Frequent issuer \(\times \) Boom. However, we still find Frequent issuer \(\times \) 3 CRA reported to be significant and negatively related to spread. This confirms that MBS tranches, including prime ones with three reported ratings, from frequent issuers are regarded as the safest.
Estimations for the prime tranches of RMBS subsample are presented in Table 4, Panel B. We find that the coefficient of Frequent issuer is still significant. Similar to the findings above, none of the CRA reported variables are significant. It seems that the possibility of issuer rating shopping is not a concern for investors in AAA tranches. For the RMBS sample, we find Distance to be significant and still positively related to the spread. This supports our earlier interpretation that domestic banks are at an advantage in residential mortgage lending due to their local knowledge.
The results of the non-prime MBS sample are presented in Table 5, Panel A. We report some differences between the prime and non-prime tranches. Firstly, the coefficients of Frequent issuer are significant in columns 1 and 3 only, and their statistical significance is weaker. However, the results presented in Panel B for the non-prime RMBS sample, we still find large and statistically significant coefficients for this variable in all models. Overall, it seems that for non-prime tranches, which are more difficult for investors to evaluate due to higher information asymmetries, investors are more likely to rely on the certification effect of frequent issuers to mitigate MBS risks.Footnote 28 We find that 2 CRA reported is highly statistically significant. This result shows that the possibility of issuer rating shopping has a major effect on investors’ perceptions when evaluating riskier, non-prime, tranches. The coefficient of Distance is not statistically significant. We find the coefficients of the interaction variables Frequent issuer \(\times \) Boom and Frequent issuer \(\times \) 3 CRA reported to be negative and significant. These results show that, firstly, investors valued frequent issuers highly during the credit boom preceding the financial crisis and required lower spreads from frequent issuers during this period. Secondly, the combination of frequent issuance with three reported credit ratings seems to be perceived as an important transparency indicator and risk mitigation factor.
Further analysis and robustness checks
We conduct further analyses by testing the robustness of our findings. We interact Frequent issuer with tranche credit rating. The interaction variable should show us whether the importance of frequent issuers increases as the credit quality of a tranche deteriorates. To simplify the interpretation, we utilise the ordinal form of the credit rating variables (Tranche Credit Rating) in these estimations. Tranche Credit Rating takes values from 1 (AAA rated) to 21 (C rated) depending on the tranche’s composite credit rating. We predict a positive coefficient for Tranche Credit Rating, i.e. yield spreads should increase as credit rating deteriorates. The results are presented in Table 6. Consistent with our main results, the coefficient of Frequent issuer is still negative and significant. As expected, we find that yield spreads increase as the tranche credit rating declines (as the numeric rating value increases). We report a negative and statistically significant coefficient for the interaction variable Frequent issuer \(\times \) Tranche Credit Rating. These findings show that frequent issuers often issue securitizations at lower spreads and the value of frequent issuance increases for lowest quality securities.
We also use rating disagreements as a gauge to measure the level of information asymmetry. We hypothesize that dissimilar ratings by different agencies on a given tranche implies a higher degree of asymmetric information for investors. We utilise Rating Disagreement, a variable which equals to 1 if there is at least a one notch difference between the ratings and 0 otherwise. Additionally, we check whether the magnitude of rating differences influence our findings using Rating Gap (measured by the numeric difference between the highest and lowest rating). The results are presented in Table 7. We do not find significant coefficients for Rating Gap (column 1) and Rating Disagreement (column 2). It is worth noting that our main variable, Frequent issuer, is still highly significant in these specifications. In columns 3 and 4, we estimate our baseline model for subgroups categorised by Rating Disagreement. We find that Frequent issuer carries a negative sign in both specifications. This result shows that investors attach value to frequent issuers whether the rating agencies disagree or not. Overall, our findings presented in this section using alternative variables and subgroups are in line with our main results.
Pre- versus post-great financial crisis (GFC)
As discussed in Sect. 2, the dynamics of the securitization market and the regulations regarding ABS creation and issuance have changed after GFC. Hence, investors’ perceptions of frequent issuance as a mechanism to mitigate MBS risks may differ for the pre- and post-GFC periods as more transparency requirements have been introduced after the failure of this market. To capture the possible differences between the two periods, we re-run our baseline analysis separately for before and after the GFC.
In Table 8, we present the results for the pre- (columns 1–3) and post-crisis periods (columns 4–6), respectively.Footnote 29 We observe significant differences between the two periods. Firstly, we observe that Frequent Issuer is strongly significant (at 1% level) in both periods; however, the coefficient of this variable is much larger for the post-GFC period. The results indicate that MBS from frequent issuers carry even lower spreads in the post-crisis period; hence, investors seem to attach more value to Frequent Issuers after their negative experiences with MBS during the financial crisis, deeming bonds issued by them relatively less risky. This observation could be attributed to the implications of the increasing reputational concerns of the large players in the securitization market. As these issuers would certainly seek to prevent any further deterioration in their reputation, investors expect that such intensified reputational concerns in the post-GFC period should mitigate further opportunistic behaviour from issuers.
Secondly, we observe that the coefficients of all CRA reported variables are positive and highly significant in columns 1 and 2 (3 CRA reported being the base category) for the pre-crisis period, which was not the case in the regressions we estimated for the whole period. These results show that, in the pre-crisis period, MBS tranches were priced higher when only one or two credit ratings are reported in comparison to tranches where credit ratings from all three rating agencies are reported. The results, supporting H3, also confirm that tranches with only one credit rating reported are perceived to be riskier than tranches with two credit ratings. Our findings are in line with evidence provided by He et al. (2012) in support of the rating shopping hypothesis, widely observed in the pre-crisis period, where issuers that select and report only favourable credit ratings while suppressing unfavourable ratings are deemed to be more risky. This relationship between the number of reported ratings and initial yield spreads disappears for the post-GFC period, as we do not observe any significant coefficients for these variables. These findings show that investors’ do not price their suspicions of “rating shopping” in the post-GFC period as they did in the pre-crisis period. This could be the result of stringent new rules and regulations introduced in the post-GFC period regarding the assignment of credit ratings of MBS and limits in the closeness of the relationship between issuers and rating agencies. As of 2013, EU regulations required all structured finance securities to report at least 2 ratings thereby removing the information content of securing dual ratings. Investors seem to have faith in the new credit rating regulations as they do not seem to be adjusting their risk perception for possible “rating shopping”. We find another result that supports these arguments. We find that Subordination is not significant for the post-crisis period, even though it is highly significant for the pre-crisis period. This shows that in the pre-crisis period, credit ratings did not completely capture the leverage effects within deals and higher subordination typically signals higher risk deals. It seems that credit ratings assigned for the post-crisis MBS captures these effects.
Thirdly, we do not observe Distance to be significant in the post-GFC period estimations. It seems that the investors’ perception of MBS issued by foreign banks to be riskier in the pre-crisis period, has died down in the post-crisis period. This could be the result of a more uniform European market with stringent regulation reducing the level of information asymmetries in all MBS, reducing the emphasis attached to local expertise by investors.
Subsequently, we estimate the models for prime (AAA) and non-prime (non-AAA) MBS for the pre- and post-GFC periods to examine whether frequent issuer effects differed for these periods depending on investors’ risk preference. Results are presented in Tables 9 and 10 for prime and non-prime, respectively. For the prime sample, we find that after the GFC, investors started to attach more value to frequent issuers, although these tranches have the highest credit ratings and are least risky. This is perhaps a reflection of investors being more cautious after experiencing unprecedented losses from triple-A tranches during the 2007–2009 financial crisis. For the non-prime MBS, which are more risky and challenging to value, we find that (reported in Table 10) Frequent issuer is significant in all estimations pre- and post-crisis. However, we observe that coefficients of Frequent issuer for the post-crisis results are much larger, indicating that non-prime tranches carry a lower spread when they are issued by a frequent issuer. Investors seem to have intensified their reliance on the certification effect of frequent issuers when evaluating risky securities in the post-crisis period. Comparing results presented in Tables 9 and 10 to our results for the whole sample period in Tables 3, 4 and 5, we also find that investors have only been cautious about “rating shopping” (CRA reported variables) for the non-prime tranches and only for the pre-crisis period. Credit rating regulation introduced after the crisis seems to have decreased this investor scepticism about rating shopping, even for more risky tranches.