Analysis of IPO pricing efficiency under the registration system

The empirical research of this paper mainly consists of two parts: one is to quantitatively estimate the IPO pricing efficiency of the pilot registration system listed companies and the other is to compare the changes of IPO pricing efficiency before and after the registration system reform.

In quantitative estimation of the IPO pricing efficiency of listed companies, we have observed the Chinese stock market and found that China does not deliberately lower the IPO price. The reasons are as follows: first, investors can obtain excess returns by purchasing IPO stocks in most cases. Offline investors may bid up to increase the allocation of shares. Second, issuers and underwriters have little risk of breakouts. This is because issuers are hungry for financing and underwriters may be linked to the total amount of financing. Not only will the underwriters not lower the IPO price but also there is a possibility of raising the IPO price. Therefore, when we study the IPO pricing efficiency, we added the production function into the stochastic frontier model and take the deviation between the actual IPO pricing and the unbiased estimate as an indicator to examine the IPO pricing efficiency.

In comparing the efficiency of IPO pricing before and after the registration system reform, in this study, we selected listed companies before the registration system reformed on the GEM as a comparative sample and compared with the pricing efficiency of GEM companies after the reformation of the registration system. This research can more reasonably express the conclusion of whether the registration system improves the efficiency of IPO pricing.

Over time, the overheated market sentiment in the early stage of the registration system reform is gradually returning to rationality. Also, the market’s self-correcting function gradually comes into play. In this case, institutional investors and underwriters will also be more cautious about IPO pricing. Based on the above analysis, my research hypothesis 1 is as follows:

H1: The implementation of the registration system has gradually improved the efficiency of IPO pricing in the Chinese stock market.

Because the Science and Technology Innovation Board is a newly opened market, there is no reasonable comparable company as a reference for IPO valuation. However, the GEM has both stock and increment. If the registration system is implemented on the GEM, there will be more companies that can be referred to in pricing and valuation; so, investor sentiment will be more stable. Based on the above analysis, my research hypothesis 2 is as follows:

H2: The pricing efficiency of the GEM IPO under the registration system is higher than that of the Science and Technology Innovation Board.

Underwriters are an important intermediary between issuers and investors and play an important role in the IPO pricing process. It is generally believed that underwriters with a higher reputation can provide higher quality prospectuses to reduce information asymmetry between issuers and investors. In this way, the IPO pricing will be closer to the intrinsic value of the company. Based on the above analysis, my research hypothesis 3 is as follows:

H3: Underwriter reputation has a positive effect on IPO pricing efficiency. The higher the underwriter reputation, the higher the pricing efficiency.

At present, the commonly used methods for efficiency measurement include the nonparametric data envelope model (DEA) and the parametric stochastic frontier model (SFA). They all measure the technical efficiency by constructing the frontier production function and judging the distance between the output and the frontier. DEA is a linear programming model used to evaluate a set of homogeneous decision-making units with multiple inputs and multiple outputs. It is a method of nonparametric frontier efficiency analysis. However, nonparametric models have certain limitations. It mainly uses the linear programming method for computational efficiency, rather than the parametric method, which has the statistical test number as a reference for the sample fit and statistical properties. In addition, the nonparametric method has certain restrictions on the observed values, and sometimes, some sample values must be discarded, which affects the stability of the observed value results. Moreover, DEA does not consider the influence of random factors. Therefore, we choose the parametric method to calculate the efficiency. The parametric method - the stochastic frontier model is introduced below.

The stochastic frontier model (SFA) believes that the “real output” of any firm cannot exceed the “output boundary”, and the degree of deviation between the two is regarded as an efficiency loss. Hunt (1996) found that the measure of IPO pricing efficiency is similar to that of production efficiency; so, the stochastic frontier model was introduced into the study of IPO pricing efficiency. In this study, Hunt regards the price maker as a producer and regards the influencing factors of pricing as an input; then, the issue price is an output. The research is based on a stochastic frontier model in the form of a production function. Because in China’s stock market, there is no situation where the issue price of shares is deliberately lowered. Therefore, we added the cost function into the stochastic frontier model and constructed a stochastic frontier cost model, which is specifically expressed as follows: 1 lnPi=βXi+εi=βXi+vi+ui $$ln{P_i} = \beta {X_i} + {\varepsilon_i} = \beta {X_i} + {v_i} + {u_i}$$

In this model, P_i is the column vector composed of the IPO price of the sample company; X_i represents the matrix composed of various factors that affect the IPO pricing; β is the parameter to be estimated and ε_i is the total error of the model, which is an asymmetric error term. ε_i includes two parts: v_i and u_i, and the two are independent of each other. v_i is a symmetric random error with a normal distribution. u_i is the systematic deviation between the actual issue price and the potential effective price. If u_i is >0, there is a systematic positive bias in IPO pricing, and IPO pricing is overvalued.

We define the IPO pricing efficiency as the ratio eff between the actual IPO pricing and the potential effective price, which is specifically expressed as follows: 2 effi=E(PEi|ui,xi)E(PEi|ui=0,xi)=exp⁡(ui) $$ef{f_i} = {{E(P{E_i}|{u_i},{x_i})} \over {E(P{E_i}|{u_i} = 0,{x_i})}} = \exp ({u_i})$$

The stochastic frontier model divides the error between the actual price of IPO pricing and the estimated effective price into random errors and unilateral systematic errors; so, the premise of using the stochastic frontier model is the existence of an inefficiency term. Coelli (1996) used the parameter γ as a parameter to characterise the suitability of the stochastic frontier model. If γ is not significantly zero, it means that there is an inefficiency factor, and the stochastic frontier model is applicable; if γ is significantly close to zero, it means that the inefficiency factor does not exist. At this time, the difference between the actual value and the estimated effective value is all caused by the random error term, and the stochastic frontier model is equivalent to the least squares (OLS) model.

The stochastic frontier cost model is used to compare the degree of deviation between the actual IPO pricing and the intrinsic value of the company to analyse the IPO pricing efficiency before and after the registration system reform. The constructed model is as follows: 3 $$\displaylines{ ln{p_i} = {\beta _0} + {\beta _1}ln(ep{s_i}) + {\beta _2}ln(growt{h_i}) + {\beta _3}ln(nap{s_i}) + {\beta _4}ln(asse{t_i}) + {\beta _5}ln(al{r_i}) + {\beta _6}ln(r{d_i}) \cr\quad\ \ \ {\rm{}} + {\beta _7}ln(siz{e_i}) + {\beta _8}ln(fe{e_i}) + {v_i} + {\mu _i} \hfill\cr} $$

In this model, the explained variable p_i is the actual price of the IPO; naps_i is the earnings per share after deducting non-recurring gains and losses in the year before the listing, which measures the profitability of the company; growth_i is the growth rate of the company’s operating income, which measures the growth of the company; eps_i is the net assets per share of the company in the year before the listing, which measures the value of the company’s per share; asset_i is the total assets of the company in the year before the listing, which measures the scale of the company; rd_i is the proportion of the company’s intangible assets to total assets, which measures the company’s scientific and technological attributes; size_i is the number of new shares issued, which measures the scale of new share issuance; and fee_i is the share issuance cost per share, which measures the underwriter’s reputation.

To test whether the reform of China’s registration system has achieved the purpose of improving the efficiency of IPO pricing, we select the companies after the reform of the registration system on the Sci-Tech Innovation Board and the GEM as a sample and select the companies before the registration system reform on the GEM as a comparison sample. There are two reasons for this consideration: first, the listed companies on the Science and Technology Innovation Board and the GEM are both high-tech enterprises; so, the two boards are comparable. Second, the Science and Technology Innovation Board is an incremental market, and the GEM is a market in which stock and increment are synchronised. At the same time, the data of two sections are selected, which is more complete and time sensitive in the selection of samples. The data can make the research more reasonable to reflect whether the registration system has improved the efficiency of Chinese IPO pricing.

After excluding samples with missing important data, 271 listed companies since the opening of the Science and Technology Innovation Board on July 22, 2019 and 127 companies since the GEM pilot registration system on August 24, 2020 were screened. Then, 761 listed companies under the non-registration system were considered since the GEM launched as a comparative sample. All data come from the Cathay Pacific Database (CSMAR). Tables 2–4 shows the descriptive statistics for each sample variable.

It can be seen from the descriptive statistics table that the average IPO pricing of the Sci-Tech Innovation Board is the highest at 35.75 yuan, followed by the GEM after the reform of the registration system, which is 28.46 yuan, and the GEM IPO before the reform of the registration system, with the lowest average price of 22.83 yuan. This shows that the registration system has opened the price-earnings ratio limit, making IPOs tend to set high prices. The same rules are followed in terms of issuance costs per share. The Science and Technology Innovation Board is the highest, followed by the GEM after the reform of the registration system. The GEM before the reform of the registration system has the lowest issuance cost per share. This shows that the reform of the registration system has increased the cost of due diligence for underwriters and increased the cost of issuance. From the perspective of company size and total assets, the asset size of companies listed on the Science and Technology Innovation Board is higher than that of the Growth Enterprise Market. This may be due to the fact that the Growth Enterprise Market encourages and supports SMEs that do not meet the scale requirements for listing on the main board to conduct equity financing. In addition, the assets of GEM listed companies after the reform of the registration system have declined, indicating that the registration system has lower requirements for enterprise scale. From the perspective of asset-liability ratio, the average asset-liability ratio of STAR Market companies is lower than that of GEM companies, and among GEM companies, the asset-liability ratio after the registration system has declined. This shows that the implementation of the registration system makes enterprises more inclined to raise funds directly through issuance and listing.

Using Frontier 4.1, a piece of software suitable for stochastic frontier analysis, the samples of the Science and Technology Innovation Board and the GEM are analysed. The estimated results are shown in Table 5.

In Table 5, the gamma value of the Science and Technology Innovation Board is 0.7208, which is significantly not zero. There is a systematic one-sided error. After the registration system reform, the gamma value of the ChiNext board is 0.0065, which tends to zero significantly. Therefore, there is no systematic one-sided error, and the MLE estimation degenerates into an OLS estimation. Before the registration system reform, the GEM gamma value was 0.7910, which was significantly different from zero. There is a systematic one-sided error, suitable for stochastic frontier models.

Using the stochastic frontier cost model, the average pricing efficiency of the Sci-Tech Innovation Board of 1.2387 was estimated, and the degree of overestimation is 23.87%. The average pricing efficiency value of GEM after the reform of the registration system is 1.0089; so, the pricing efficiency is very high. Before the reform of the registration system, the average pricing efficiency of the GEM was 1.3760, and the degree of pricing overestimation was 37.60%. It can be seen that the registration system reform has improved the efficiency of IPO pricing. Compared with the Science and Technology Innovation Board, the GEM with a certain stock base has higher pricing efficiency. This conclusion proves my research hypotheses 1 and 2 and also shows that it is feasible to implement the registration system in the A-share and other incremental markets in the future. Regarding the phenomenon that the pricing efficiency of the Science and Technology Innovation Board is lower than that of the ChiNext, we think the main reasons are as follows: first, the Science and Technology Innovation Board is a newly opened market. When IPO pricing, the lack of reasonable comparable companies as a reference limits the pricing power of lead underwriters to a certain extent. This makes the IPO pricing efficiency temporarily lower than that of the ChiNext under the registration system. Second, compared with the ChiNext, the Science and Technology Innovation Board gathers industries such as biomedicine, software information and computer communications. These industries are strongly supported by the state and contain high policy dividends. Finally, since the outbreak of the new crown epidemic in 2020, the industries on the Science and Technology Innovation Board have been less affected or even profitable, which has made the estimates of related companies hit new highs. Therefore, it naturally led to an increase in IPO pricing.

In addition, observing the impact of various variables on the IPO pricing, it is found that the issuance fee per share (fee) has a significant positive correlation with the pricing of new shares. The higher the issue fee per share, the higher the price of the new issue. Regarding the issue fee per share, we use it to express the reputation of the underwriter. Generally, the higher the reputation of the underwriter, the higher the success rate of the IPO and the higher the fees charged. That is, there is consistency between issuance fees and the reputation of the underwriters. Therefore, the higher the reputation of the underwriters, the more inclined they are to set high prices and the lower the efficiency of IPO pricing. This is inconsistent with hypothesis 3 of my research. The possible reasons are as follows: first, the underwriter’s sponsorship fee is linked to the total amount of funds raised. Reputable underwriters are more likely to use their industry position and reputation to intentionally set IPO prices higher to capture more underwriting fees. Second, the underwriting fees are too high, which can offset or even cover the risk of loss caused by co-investment. Regulators have set up a sponsor co-investment mechanism for underwriters to restrain the moral hazard of underwriters. However, the proportion of co-investment is basically at the level of 2%-3%, while the cost of sponsor underwriting is about 5%. Therefore, the sponsorship fee can far cover the risk of loss caused by co-investment.

The impact of net assets per share (eps) on the pricing of new shares is significant and positively correlated, that is, the higher the intrinsic value per share, the higher the stock price, which is in line with common sense. For the size of IPOs, there is a significant positive impact in all three sectors. It shows that the larger the issuance scale, the higher the IPO pricing and the lower the pricing efficiency. In terms of total enterprise assets, the impact on the Sci-tech Innovation Board under the registration system and the GEM is not significant, but it has a significant impact on the GEM under the non-registration system. This reflects the reduced focus on company size in IPO pricing under the registration-based system. The asset-liability ratio (alr) has an insignificant impact on the Sci-tech Innovation Board, while its impact on the GEM has changed from positive to negative. This shows that the registration system makes enterprises tend to direct financing and pay less attention to the asset-liability ratio when pricing. The impact of enterprise science and technology innovation attributes (rd) on the science and technology innovation board is not significant. It reflects that on the Sci-Tech Innovation Board, investors are mostly speculating on concepts but not paying much attention to the actual technological innovation attributes of companies. In terms of the company’s earnings per share (naps) and the company’s growth ability (growth), the impact on the Sci-tech Innovation Board is significantly positive, but it has no significant impact on the ChiNext after the reform of the registration system.