Hidden champions or black sheep? The role of underpricing in the German mini-bond market

Interestingly, Neuberger and Räthke-Doppner (2015) notice that loan prices of German SMEs decrease with their lending relationship and Deloof and La Rocca (2015) find that the amount of trade credits used by Italian SMEs decrease during the financial crisis highlighting the need for alternative corporate financing channels. Other sources for alternative financing recognized in the literature include crowdfunding (e.g., Mollick 2014; Ahlers et al. 2015; Günther et al. 2017; Vismara 2016a).

Colombo and Grilli (2007) find that young Italian technology firms suffer from credit market frictions (i.e., availability of bank financing), and Rostamkalaei and Freel (2016) show that high growth companies face higher credit costs.

bondm was the first market segment that specifically focused on SME bonds. Nine months after its introduction, the German Stock Exchange opened the Entry Standard segment for small and medium-sized bonds. Other stock exchanges in Düsseldorf (Mittelstandsmarkt), Munich (m:access), and Hamburg-Hannover (Mittelstandsbörse) also subsequently established specialized mini-bond market segments.

A total of fourteen companies issued more than €200 million in the specialized mini-bond segment (for example, Deutsche Börse AG, Air Berlin). However, none were SMEs.

For prominent examples, see, e.g., FFK Environment, Centrosolar, Windreich I & II, and Getgoods (see Table A3 for more details).

Vander Bauwhede et al. (2015) find a negative relation between SME financial reporting quality and financing costs. They argue that as high-quality financial statements reduce information asymmetry between SMEs and banks, they are rewarded by lower credit costs. Thus, financial statement quality can also be considered as a costly signal of firm quality.

Other models also focus on information asymmetry levels and place different weights on the objectives of the players involved. See, for example, Rock (1986), who explores information asymmetries among several types of investors, Benveniste and Spindt (1989), who focus on information asymmetries between underwriters and investors, and Baron (1982), who examines information asymmetries between issuing firms and underwriters.

Ritter (2003) provides an overview of the theoretical and empirical determinants of underpricing across different European markets. Khurshed et al. (2014) show that a transparent bookbuilding can take a certification role in Indian IPOs, thereby reducing information asymmetries, while Akyol et al. (2014) find that enhanced disclosure requirements reduce IPO underpricing that is associated with information uncertainty. See, for example, Vismara (2014) or Bonardo et al. (2011) for the effects of other signaling mechanisms on IPO underpricing.

Chan and Lo (2011) and An and Chan (2008), among others, document that equity offerings of firms with credit ratings are less underpriced. This generally supports the view that ratings can reduce asymmetric information. However, in this paper, we assume ratings are inflated, thus overriding their certification ability.

Similar results are reported for the Swiss bond market by Wasserfallen and Wydler (1988). They find that newly issued bonds tend to be underpriced by 0.46%, which is positively correlated with maturity and negatively correlated with coupon size. Analyzing the same market but over a different time period (1999–2001), Kovács and Zeder (2003) find that straight bonds of foreign issuers quoted in CHF are overpriced, which may be attributable to the negative development of the Swiss bond market.

We also replicated all multivariate regressions with the least common factor of 42 observations. We found that the signs as well as the magnitude of the coefficients on the main explanatory variables remained the same, but the overall significance was lower (see Tables A6 and A7 in the online appendix).

Note that we do not deduct a cumulative bond index return until the n’th trading day (such as the former Lehman Brothers index used in Cai et al. 2007). This is because we would need an index for the respective rating class given by the rating agencies, which is highly questionable, and we would need the same maturity, which is typically not provided.

The controlling variables based on the balance sheet are taken from the year prior to the mini-bond issuance.

For the different specifications in Table 8, we do not use Implied PD z-score and ∆ Implied PD jointly in the regression because both variables are highly correlated. If we consider them separately, we find no indication of multicollinearity as per the mean Variance Inflation Factor (VIF) of 1.21 and the maximum value of 1.50.

Studies based on the US bond market before the recent financial market crisis usually document higher yield to maturities for nonrated in comparison to rated bonds (see, for instance, Asquith et al. 2013). The surprisingly low average yield to maturities of nonrated mini-bonds in our sample are mainly explainable by 1) a low-yield environment in Germany with considerable interest (by retail investors) in nonrated (mini-)bonds during our sample period, 2) five mini-bonds listed in the Prime Standard segment requiring issuing companies to be larger and more transparent than their peers issuing mini-bonds (corresponding with lower default risk), 3) one mini-bond with a special collateralization against a proportion of graduates’ future salary from a well-known German private university (Universität Witten/Herdecke), and thus six out of eleven mini-bonds in the nonrated group having a considerable lower yield to maturity and thereby downward biasing the average yield to maturity in this group.

An alternative channel using underpricing as a signal would be to issue the mini-bond below par. However, all considered mini-bonds were issued at par value. Therefore, we are unable to derive any conclusions about a signaling mechanism using this channel.

Cai et al. (2007) show that bond issuances (unlike equity offerings) may suffer from insufficient or even non-existent trading at the first trading day. Therefore, they recommend calculating underpricing only if liquid bond prices are available within the first seven calendar days after issuance. In our sample, the vast majority were traded on the first trading day (92). Only four were traded for the first time on the second day and only two on the fifth trading day (see Table 7). This clearly indicates that illiquidity is unlikely to be affecting our results.

We decided to present the results that can be easily reproduced by using commercial databases. The bond prices we used as a robustness check were obtained directly from the exchanges and are not readily available.

The main differences between the scores are 1) “slightly” different factors and 2) that the o-score uses nine instead of five factors. Thus, less weight is allocated to a single factor, which makes the o-score less sensitive.

In unreported results, which are available upon request from the authors, we address country differences. We derive scores and mapping using US firms, which probably deviate from German SME companies with respect to characteristics such as financing structure (public listings), principal bank lending versus the use of capital markets, different sample industries, and different accounting standards. To address the issue of different country sensitivities, we use coefficients from Beinert et al. (2008) to calculate Altman’s (1968) z-score based on a German sample rated by Moody’s (z _{Germany Moody ’ s} =  – 0.354 ∙ X1 – 12.222 ∙ X2 – 11.456 ∙ X3 + 4.274 ∙ X4 + 8.906), Standard & Poor’s (z _{Germany S & P}  = 0.185 ∙ X1 + 0.238 ∙ X2 − 4.472 ∙ X3 − 0.027 ∙ X4 + 0.545), and HGB as the accounting standard. The results were again highly similar to those obtained from the other approaches. This result is not surprising in the light of Altman et al.’s (2014) study of the performance of our main metric (Altman z-score as calculated in Equation (3)) in an international context. For their study, they use the ORBIS database of Bureau van Dijk (BvD), which is the same data we use in our study, and underline the fact that the accounting data is “harmonized” for the different international contexts, thereby reducing the problem of an international comparability (Dafne is the German and Amadeus the European subset of companies from Orbis.). They show compelling evidence that the z-score we used for our analyses performs very satisfactorily in an international context.