The influence of platform firms on stock analyst recommendations
Abstract:
This quantitative research combining financial data from Thomson Financial Datastream and analyst data from I/B/E/S from 2009 up to 2017 investigates the difference in analyst recommendations for platform firms compared to traditional firms while also accounting for the role of goodwill. Using Resource Based View, I propose that platform firms receive higher recommendations, motivated by stronger intangible assets. Using a robust regression, this research finds no significant relationship between platform firms and stock analyst recommendations. In addition, goodwill has both a negative moderating effect and a (debatable) mediating effect on this insignificant relationship.
Key words: Platform firms, Resource Based View (RBV), Goodwill, Intangible assets, network effects
Name: Nino Hendrik Bernard Heijnen Student number: 2968452 Email: n.h.b.heijnen@student.rug.nl
Track: MSc Business Administration – Strategic Innovation Management (SIM)
Supervisor: prof. dr. J.D.R. Oehmichen Co-Assessor: dr. T.L.J. Broekhuizen
2
Introduction
In 2020, ArBnB is planning to submit its initial public offering (IPO) (Klebnikov, 2019) as the latest in a series of several large-scale offerings of platform companies after Alibaba, Pinduoduo, Uber, and Lyft (Winck, 2019). This in itself is not surprising: many companies plan to go public after growing in both status and market value (see figure 1). Furthermore, already well-established platform companies and companies that shifted part of their business towards the platform business model make up some of the largest public companies in the world (e.g. Microsoft, Google, Facebook, and Apple), while Alibaba and Facebook submitted two out the eleven largest IPOs in history (Rusli et al., 2012; Demos, 2014). However, going public is not a guarantee for success for these companies, as can be seen by the disappointing IPOs of both Uber and Lyft. There has also been discussion on whether platform firms actually offer any substantial value to the market (Fijneman et al. 2018) or if they are simply rent leeches, made possible by their control of the markets through their monopolistic positions (Oh et al., 2015; Jacobides et al., 2006). Nevertheless, it is remarkable how a relatively new type of business model already has such a prominent presence in the market. Which begs the question: how do stock analysts, who are considered a proxy for the market (Davies and Canes, 1978; Black, 1973, Groth et al., 1979; Givoly and Lakonishok, 1979; Copeland and Mayers, 1981; Dimson and Marsh, 1984; Liu et al., 1990), rate platform firms compared to traditional businesses? With a new type of business model, there are not as many examples of similar businesses in the past to look at to make accurate predictions for the future. Furthermore, the financial situation for platform firms is often very different compared to those of traditional firms. Most platform firms are not making a profit and are reporting heavy losses such as Lyft, Uber, WeWork, Snapchat and Pinterest (Vincent, 2019). In addition, internally generated intangible assets do not have to be reported on the balance sheet. For tech companies, this means that their most valuable assets often go unreported (Lev, 2004). Yet, finances and annual reports are primary sources of analysis for stock analysts (FASB, 2019). Currently there is no research that makes the distinction between platform firms and other firms with regards to recommendations by stock analysts. This research aims to fill a gap in literature on the role of platform firms in the stock market and contribute to a clearer understanding of analyst behaviour in regards to evaluating emerging business models and unreported intangible assets. The main research question is as follows:
Do platform firms receive different recommendations from analysts compared to traditional firms by stock analysts?
3
firms receive higher recommendations by stock analysts compared to traditional firms. In addition, I expect platform firms with high reported goodwill to receive higher recommendation compared to platform firms with low reported goodwill.
Figure 1: Share of private/public platform companies, by size (Fijneman et al., 2018).
Theory and Literature
Platform firms
4
such as video game console manufacturers selling their systems at cost price, or supply to the platform itself (e.g. Amazon and Apple’s iOS). This can be challenging, as a successful platform requires a large user base on both sides of the market (Rochet and Tirole, 2006).
Resource Based view
Resource-based view (RBV) is a managerial framework that places focus and importance on the resources of a firm. RBV puts the resources at the centre of attention and proposes that different firms will pursue different strategies based on the internal resources that they have. Through the lens of RBV, a firm’s competitive advantage are obtained through obtaining and developing internal resources. Barney (1991) noted that resources only have the potential to become a competitive advantage for the firm if they are valuable, rare, inimitable, and not substitutable (VRIN). More recent literature on RBV expended on the original framework by further emphasizing the need for dynamic capabilities, that is the ability to integrate internal and external resources fast and efficiently to match technological progress and changing consumer needs (Helfat and Peteraf, 2009; Lichtenthaler, 2009). This concept of dynamic capabilities seems especially applicable to platform firms, who are seemingly better suited at meeting consumer needs through technological advancements (Fijnemans et al., 2018).
Resource Based View and analyst recommendations of platform firms
As mentioned previously, while there are numerous examples in recent years of IPO’s from platform firms, the research on how stock analysts rate these firms is still lacking. Before any extensive research can be conducted on whether valuations of platform firms are justified or biased, it is important to investigate if there are significant differences in how stock analysts evaluate platform firms compared to traditional firms.
5
platform that they perceive as going to be the biggest network (Eisenmann et al., 2006). This makes it very difficult for new platform firms to enter the market as both sides of the market have little incentive to switch to a platform with a smaller network. This makes the network effects that platform firms create not only valuable, but also rare, inimitable and not substitutable.
While traditional firms are certainly able to create network effects, it can often not be done to the same extent as online platform firms. Rather than benefitting from network effects by connecting both sides of the market, many traditional firms that do not facilitate a platform only create a network one side of the market (e.g. customers), or on both sides of the market but separated (e.g. a network with customers, and a network with its vertical supply chain).
Stock analysts are a proxy to the market (Davies and Canes, 1978; Black, 1973, Groth et al., 1979; Givoly and Lakonishok, 1979; Copeland and Mayers, 1981; Dimson and Marsh, 1984; Liu et al., 1990). Investors often do not have the time or interest to analyse many firms intensively to make an informed investment decision. Therefore, investors pay attention to the recommendations of stock analysts. Stock analysts will analyse a company’s economic health and try to make a calculated estimation on whether a company is likely to grow and be profitable in the future. To determine a company’s value and project their future earnings, financial statements serve as one of the primary sources for analysis. However, one issue with only analysing reported financials is that you do not account for intangible assets that are generated internally, as they do not have to be reported in financial statements (Buesser, 2019). This also includes all resources that are generated by platform firms through their network, such as personal information, information on customer needs, and innovations. Because of this, stock analyst already incorporate different methods of evaluating the value of intangible assets (Rodov and Leliaert, 2002; Joia, 2000). Therefore, I posit that stock analysts also consider unreported intangible assets in order to more accurately project a company’s future earnings. As a result, I argue that platform firms are rated higher than traditional firms are by stock analysts, with all other factors being equal.
Hypothesis 1: There is a positive linear relationship between platform firms and the average recommendations by analysts.
Goodwill
6
accounting for attributing the premium paid towards goodwill is that, when acquiring a business, the value that the acquiring company sees in its acquisition is based on potential future cash flows. Existing customer relationships, brand recognition, patents, and a plethora of other resources are attributed as goodwill as well. However, for this research, I am particularly interested in the intangible asset that cannot be recognized that compromise goodwill (Bloom, 2009; WGARIA, 2015). This distinction can be seen in the figure 2.
Figure 2: distinction of intangibles versus goodwill (WGARIA, 2015).
Bloom (2009) also made a different distinction of two classes of goodwill: (1) goodwill caused by internally generated value, and (2) goodwill that was acquired. While goodwill can only be reported after an acquisition, it can already in part be seen in firm through the difference between the firms total market value and the total book value, also known as market value added (MVA) (Ellis, 2001). Goodwill is only seldom mentioned within the context of RBV. Adding on to Hofer and Schendel’s (1978) five different types of resources, Grant (1991) specifically mentions goodwill as part of a sixth type of resource in intangible resources. Later work by Kristandl and Bontis (2007) aims at clearly defining and constructing intangible resource within the context of RBV. Kristandl and Bontis (2007) also make the distinction between unrecognized intangibles and other recognizable components of goodwill.
7
empirically. Nonetheless, the combination of literature regarding goodwill in the context of RBV, and the fact that goodwill is in part an unrecognized intangible asset, similar to the proposed network effects of platform firms, leads me to believe that goodwill is worth analysing in relation to our proposed main effect.
The role of goodwill in analyst recommendations
With seemingly high IPO’s for platform firms, the estimation of value of intangible assets in stock analysis is likely to be driven by goodwill. Because a portion of goodwill cannot be fully attributed to recognizable intangible assets and is therefore difficult and time consuming to properly analyse, stock analysts use the reported goodwill of a firm as an important measuring instrument in their evaluations. Therefore, platform firms that are also reporting high goodwill will receive higher recommendation compared to platform firms with low reported levels of goodwill.
Hypothesis 2: The relationship between platform firms and average recommendations by analysts is positively moderated by Goodwill.
Platform firms that are either acquired are acquire additional businesses themselves are likely to report relatively higher levels of goodwill due to the relative large portion of purchased value being intangible. Accordingly, goodwill can also potentially serve as a validation measurement that gives stock analysts confidence to predict positive earnings for platform firms. There is a strong possibility that stock analyst require certain financial requirements to be met, in order to confidently estimate positive future earnings for platform firms, caused by the uncertainties that accompany a relatively new and uncommon business model (Nielsen and Bukh, 2011). Therefore, platform firms that have high reported goodwill will receive higher recommendation by stock analysts compared to platform firms with low levels of reported goodwill.
Hypothesis 3: A firm’s goodwill will mediate the positive linear relationship between platform firms and average recommendations by analysts.
8
Figure 3: Conceptual model
Methodology
The goal of this research is to identify whether stock analysts rate platform firms than traditional firms. In order to do this, I will be conducting a quantitative analysis of available financial data and stock recommendation, using available data on the largest public companies globally. Using a quantitative approach, I hope to identify significant patterns in stock recommendation that help us to answer the hypotheses. In this section, I will explain how I created the sample, what the variables for analysis are, and which analyses are going to be conducted. The sample was created using Excel and all analyses were done through Stata.
Sample
9
matches, (3) WRDS’ compustat matches, and (4) manual matches using WRDS’s manual code lookup. Using these methods, I managed to identify 96% of all the I/B/E/S tickers of firms in our initial sample. With these tickers, I obtained all data related to stock recommendation from I/B/E/S for our time period. This resulted in a total of 21277 firm years from 3679 individual firms, using 4946 unique company names (Caused by companies using unique company names for different parts of their organization while operating under the same ticker). Lastly, this data was merged with the financial dataset from Thomson Financial Datastream.
Based on the research of Wiersema and Zhang (2011), I have made adjustments to the dataset. The Study of Wiersema and Zhang also involved using data from I/B/E/S and as I deem their methodology exemplary, I resolved to follow their procedures where it would also be relevant for this study.
Dependent variable
Average analyst recommendation
The stock recommendations from I/B/E/S are on a 5-point scale ranging from strong buy to sell. In order to use these recommendations for our quantitative analysis, I encoded the recommendations, ranging from 1 up till 5. In addition, the recommendations were reverse coded to facilitate better interpretation of our analyses so that a high number would imply a high recommendation. This was done by subtracting the numerical value from six. Subsequently, I created our independent variable called “Average analyst recommendation” using the mean recommendation for all stock analysts that cover a particular firm in a single year. In addition to the approach of Wiersema and Zhang (2011), I collapsed our data. This is done because (1) to avoid a false positives caused by numerous observations with the same mean value, and (2) to make the data more manageable, and (3) to operate on the firm year level rather than on the individual analyst level.
Independent variables
Platform firms10
A modern platform business can be defined as a multi-sided business model focused on creating value by facilitating interactions between two or more interdependent groups through technology. These platform businesses are positioned to benefit from the “network effect” which is the positive effect on a good or service as the number of customers or participants increases.
The index contains 75 firms in total of which most were present in our sample. In addition, I had a manual look through the data in order to identify well-known companies which are known to conduct in platform activity. This resulted an additional platform firms, bringing the total to 62. I have created a dummy variable called platdummy, where platform firms were given a value of 1 and traditional firms a value of 0.
Goodwill
While the dataset already contains a clear variable for goodwill named goodwill gross, this variable does not take into consideration relative size of goodwill for a particular firms finances. To adjust for this, I have divided goodwill by total assets.
Control Variables
Analyst coverageSimilar to Wiersema and Zhang (2011), I controlled for the number of analysts that are following a firm, with the argument being that once more analysts follow a firm, investors can make a more informed investment decision. While the previous researched measured analyst coverage using the average number analysts that provide recommendations over a six-month period, we have opted to do the same but over a one-year period to have this measurement in line with my dependent variable. Cash & short-term investments
We have controlled for liquidity by using cash & short-term investments. As many platform firms only provide digital services, they have very different levels of liquidity compared to traditional firms.
Firm size
Wiersema & Zhang (2011) controlled for firm size using additional data on the firm’s market capitalization. However, I was unable to replicate this variable with the limited information provided. Instead, I controlled for size using a combination of different financial variables, specifically employees, total assets, total liabilities, and net sales
11
I controlled for the country in which a firm’s headquarter is located using the country variable. The main reason for this control variable is to account for social, economic, and political factors that might affect a country and thus indirectly a company located in that country. Using the “encode” command in Stata, I assigned a numerical value to each unique country based on alphabetical order. In total, our firm contains firms from 49 different countries. The three countries with the most firms are the USA, Japan, and China with 5266, 3172, and 1002 firm year observation respectively. The country with the least observations is Argentina. With 21,277 observations from 49 different countries, the country with the amount of companies closest to the mean of 434 is South Africa, with 435 observations. Year
To control for potential economic factors I also use a control variable for the date of recommendation. Wiersema and Zhang (2011) grouped all recommendations into 6-month categories. Therefore, I opted to follow their approach. However, I have grouped the recommendations into categories spanning one year, as I was unable replicate Wiersema and Zhang’s approach in the data and was forced to use this alternative with the limited time span of this research.
Industry
Lastly, I also controlled for the industries in which the firms operate. This is to control for potential biases by stock analysts for particular industries. Data from I/B/E/S already contains a standard industry code (SIC) variable that only uses the first character of the SIC. This could only potentially cause conflict with firms in industries with SIC’s in the 1000 to 1999 range as this contains two industries, specifically mining and construction. However, as these industries are likely not to contain any platform firms, I felt safe to use this variable directly. This variable was renamed to industry.
Analyses
12 Tabel 1: Descriptive Statistics
Variable Obs Mean Std.Dev. Min Max
average recommendation 21277 3.441 .441 1 5
platdummy 21277 .019 .135 0 1
relative goodwill 6465 .125 .16 0 1.448
analyst coverage 21277 17.013 11.13 1 106
cash & short term investments
18507 2260000 8370000 0 2.68e+08
employees 18951 38522.04 82065.27 0 2300000
total assets 21200 4.59e+07 1.75e+08 2090 3.34e+09 total liabilities 21201 3.81e+07 1.64e+08 -105000 3.07e+09 net sales 21194 1.15e+07 2.35e+07 -1.30e+07 4.74e+08
year 21277 2012.556 2.887 2008 2017
country 21277 30.706 15.781 1 49
13 Table 2: Correlation statistics of all variables
Matrix of correlations Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (1) average recommendations 1.000 (2) platdummy 0.021 1.000 (3) relative goodwill 0.018 -0.001 1.000 (4) analyst coverage -0.090 0.022 -0.032 1.000
(5) cash & short term investments 0.013 0.244 -0.097 0.111 1.000
14
To test our main effect (hypothesis 1), we will first be performing a linear regression model. Stata offers many options for regression but I have opted for the regress regression variant. Thereafter, we will check for robustness. For testing my hypotheses, I will be running five different analyses. Initially, I will run a base regression with only my dependent variable and my control variables, in order to provide a base level of explained variance R2, which can used for comparison after introducing the main independent variable and the possible moderating or mediating variable. Then, I can run the most important analysis: the linear regression, testing my main effect (H1). Subsequently, I will test for robustness. The reason for this is that potential outliers might negatively influence the reliability of our results. This potential threat is also reflected in my initial Cronbach’s alpha (=0.5951) which does not meet the minimum level of 0.6 (DeVellis, 2012; Geoge and Mallery, 2003; Kline, 2008). This is done by calculating Cook’s distance (Cook, 1977), and, if outliers are to be found, then our predictor variables will be winsorized. These winsorized variables will be used for a new robust regression for hypothesis 1, and will also be used in any further analysis. To test for a possible moderation effect for goodwill, I will run a fourth regression, including the moderator variable for goodwill, and its interaction with direct effect. For the last analysis I conducted a mediation test. There are different options for this in Stata, but I choose the ml_mediation test, as is the most appropriate analysis to use, given that we have a binary independent variable. This approach was adapted from Krull & Mackinnon (2001).
15
Results
The initial regression, using only my control variables, can found below in table 3. As you can see, without including my main independent variable, the model explains just over ten percent of the explained variance for stock analyst recommendations (R²=0.103). This initial regression contains 16517 out of 21277 (77.7%) of all observations in the sample.
Table 3: Base regression with only control variables for R-Squared comparison
avgrec Coef. St.Err. t-value p-value [95% Conf Interval] Sig ancov -0.004 0.000 -11.47 0.000 -0.005 -0.003 *** cash_st_inv 0.000 0.000 -0.09 0.929 0.000 0.000 employees 0.000 0.000 2.48 0.013 0.000 0.000 ** tot_ass 0.000 0.000 8.74 0.000 0.000 0.000 *** tot_lia 0.000 0.000 -8.56 0.000 0.000 0.000 *** net_sales 0.000 0.000 -2.38 0.018 0.000 0.000 ** Constant 4.079 0.297 13.74 0.000 3.498 4.661 Mean dependent var 3.447 SD dependent var 0.434
R-squared 0.103 Number of obs 16517
Adjusted R-squared F-test
0.099 26.994
Prob > F 0.000
*** p<0.01, ** p<0.05, * p<0.1 Note: Year, Industry, and Country are included
As shown in table 4, firms that are identified as platform firms receive higher recommendation by stock analysts. The coefficient for platform firms is positive and significant at a 5% level (b = 0.050, p = 0.033). This would suggest that Hypothesis 1 is supported. However, the level of explained variance is still the same after including my predictor variable (R² = 0.103, Adjusted R²=0.099).
16
Table 4: Linear regression for main effect
avgrec Coef. St.Err. t-value p-value [95% Conf Interval] Sig
platdummy 0.050 0.023 2.14 0.033 0.004 0.096 ** ancov -0.004 0.000 -11.61 0.000 -0.005 -0.004 *** cash_st_inv 0.000 0.000 -0.41 0.682 0.000 0.000 employees 0.000 0.000 2.59 0.010 0.000 0.000 ** tot_ass 0.000 0.000 8.75 0.000 0.000 0.000 *** tot_lia 0.000 0.000 -8.54 0.000 0.000 0.000 *** net_sales 0.000 0.000 -2.44 0.015 0.000 0.000 ** Constant 4.081 0.297 13.74 0.000 3.499 4.662
Mean dependent var 3.447 SD dependent var 0.434
R-squared 0.103 Number of obs 16517
Adjusted R-Squared F-test
0.099 26.684
Prob > F 0.000
*** p<0.01, ** p<0.05, * p<0.1 Note: Year, Industry, and Country are included
Using the conventional cut-off point of cook’s distance values greater than 4/n (Algur and Biradar, 2017), thus 4/21277, I concluded that my dataset did contain outliers. Therefore, hypothesis 1 could not yet be supported. For robustness, I winsorized my independent variables when possible at 95%, excluding goodwill (since it is a percentage), country, industry, year, and analyst coverage. These winsorized variables in my dataset are indicated with a capital W (e.g. tot_ass becomes Wtot_ass). This step also improved reliability considerably, with an increase in Cronbach’s alpha (=0.6371). With these new regressions, I conducted a robust regression to test my main hypothesis (H1).
Table 5: Robust regression (H1)
avgrec Coef. St.Err. t-value
p-value [95% Conf Interval] Sig
platdummy 0.033 0.023 1.40 0.162 -0.013 0.078 ancov -0.005 0.000 -12.77 0.000 -0.005 -0.004 *** Wcash_st_inv 0.000 0.000 4.34 0.000 0.000 0.000 *** Wemployees 0.000 0.000 1.82 0.069 0.000 0.000 * Wtot_ass 0.000 0.000 7.05 0.000 0.000 0.000 *** Wtot_lia 0.000 0.000 -7.83 0.000 0.000 0.000 *** Wnet_sales 0.000 0.000 1.09 0.274 0.000 0.000 Constant 4.072 0.296 13.74 0.000 3.491 4. 653 ***
Mean dependent var 3.447 SD dependent var 0.434
R-squared 0.107 Number of obs 16517
Adjusted R-Squared
F-test 27.631 0.103 Prob > F 0.000
*** p<0.01, ** p<0.05, * p<0.1 Note: Year, Industry, and Country are included
17
After adjusting for outliers by winsorizing the control variables, the coefficient for platform firms has become lower and no longer significant (b = 0.033, p = 0.162). Therefore, Hypothesis 1 is not supported. Consequently, platform firms do not receive higher recommendations by stock analysts compared to traditional firms. To test the moderating hypothesis (H2) that the positive effect of platform firms on stock analysts recommendation is stronger when a firm’s reported goodwill is high, I added an interaction term using the relative size of a firm’s goodwill compared to their total assets. Because of missing values for goodwill, only 4911 observation are used for this analysis. As shown in table 6, the interaction term of relative goodwill and platform firms is negative and significant (b = -0.625, p = 0.003), thus not supporting hypothesis 2. However, it is important to note that there is a significant negative moderating effect. Therefore, platform firms with high relative goodwill will receive lower stock analyst recommendation compared to platform firms with low relative goodwill, all other factors being equal. Another interesting finding is the significant positive relationship between goodwill and stock analyst recommendations (b = 0.230, p = 0.000). Furthermore, both R² (=0.154) and adjusted R² (=0.141) have improved.
Table 6: Robust regression with interaction for goodwill (H2)
avgrec Coef. St.Err. t-value p-value [95% Conf Interval] Sig
1.platdummy 0.093 0.058 1.60 0.111 -0.021 0.208 goodwill 0.230 0.040 5.73 0.000 0.152 0.309 *** 1.platdummy#c.goodwill -0.625 0.211 -2.97 0.003 -1.039 -0.212 *** ancov -0.005 0.001 -8.57 0.000 -0.006 -0.004 *** Wcash_st_inv 0.000 0.000 3.41 0.001 0.000 0.000 *** Wemployees 0.000 0.000 -1.26 0.208 0.000 0.000 Wtot_ass 0.000 0.000 5.17 0.000 0.000 0.000 *** Wtot_lia 0.000 0.000 -5.80 0.000 0.000 0.000 *** Wnet_sales 0.000 0.000 1.59 0.111 0.000 0.000 Constant 3.434 0.121 28.42 0.000 3.197 3.671
Mean dependent var 3.400 SD dependent var 0.422
R-squared 0.154 Number of obs 4911
Adjusted R-Squared
F-test 12.553 0.141 Prob > F 0.000
18
19
Table 7: Mediation analysis (H3)
Coef. Std. Err. Z p>z [95%Conf. Interval}
20
Discussion
This study was motivated by the visible increase in the number of platform firms on the global stage. While prior research and literature has covered the business model of platform firms, there is still a lack of research on the performance of platform firms on the stock market. This study is one of the first studies to investigate how stock analysts evaluate platform firms compared to traditional firms. Using arguments from RBV, there are indications that stock analysts would rate platform firms higher than traditional firms. However, after empirically testing this notion, the results indicate that platform firms are not rated significantly different compared to traditional firms. Nevertheless, the results show that goodwill does have a significant influence on this insignificant relationship, as well as a significant direct effect on analyst recommendations. While this study does not try to answer whether a higher recommendations can be attributed to justified future performance estimates or simply a bias, it could potentially serve as part of a foundation for future research.
Theoretical implication
While resource based view would dictate that platform firms should outperform firms using a traditional business model with all other factors being equal, this is not reflected in the recommendations by stock analysts. One possible explanation for this could be the fact that resource based view is a theoretical lens in the area of business rather than finance. While resource based view could serve to explain how platform firms have taken a more prominent role in the current business landscape, it might by itself be inadequate to fully explain performances in the world of finance and the stock market.
21
Under the assumption that RBV is correct in valuing intangible resources as potential source for competitive advantages, another possible explanation could be that stock analyst are not accurately assessing the value of intangible assets, thereby negatively effecting platform firms more than traditional firms. On the other hand, given the role of stock analysts as proxies for the market and predictors of future success, this research could also serve to challenge the notion that intangible assets such as network effects create competitive advantages for the firm (Gulati, 1999). However, that notion will likely answer itself in the future through the successes or failures of the platform firms mentioned in this article.
Managerial implication
Within the theoretical lens of RBV, it is difficult to give meaningful managerial recommendation, because of the fact that a common critique of RBV is that it does not produce tangible managerial advice (Kraaijenbrink et al., 2009). However, since no significant relationship between platform firms and higher stock recommendation was found, I would advise managers of platform firms to not be overly optimistic in their decision to take their company public. While there are examples of platform firms with successful IPO’s, it cannot be said that stock analysts rate platform firms planning an IPO higher than their traditional counterparts. While future research might provide better indicators for higher recommendations for platform firms, as of yet this is not the case. Thus, the main advice that can be derived from this research for upper management of platform firms is to treat their firms similar to traditional firms in planning an IPO.
Limitations
22
that the sample of platform firms used was relatively small, with only 395 out of 21277 firm years pertaining to platform firms (1,86%). This problem could not be solved by gathering additional data, partially because of the scope of this research, but predominantly but there is no established research at this point to identify platform firms through financial or other quantitative data. Furthermore, the small number of firm years is can also be attributed to the fact that many of the platform firms in our sample only went public very recently. As these firms are public for a longer period, and as more platform firms go public, the number of firm years available will naturally increase.
Another limitation of this study is our decision to categorize analyst data by year. This was done to keep the scope of this research manageable. This could have resulted in positive recommendations in the beginning of the year cancelling out negative recommendation during the end of the year. A better solution would have been to categorize by quarters. This would also account for changes in perception and recommendations caused by financial reporting of quarterly earnings reports.
In additional, this research did not account for one potentially influential factors: the firm fixed effects. To account for this, I have run a Hausman test (1978), which proved the need to run a fixed regression model (Wooldrigde, 2016; Sayrs, 1989). However, I was unable to do so successfully. Stata was simply unable to run the regressions with all the different types of categorical data that would be a required for this analysis. That is: our independent variable for platform firms (platdummy), as well as industry, year, country, and firm. Strangely enough, Stata would omit all but three countries (China, Germany, and South Africa) from analysis for collinearity. But, more importantly, even our main independent variable for platform firms (platdummy) got omitted as well for the same reason.
While multicollinearity could indicate that it is possible to identify platform firms through the variables used for this analysis, this seems highly unlikely given the limitations in creating this variable as mentioned before. Hence, I opted to keep using my regression as is, as it incorporates the most categorical data. A potential solutions for this issue was beyond the scope of this research, as well as my current knowledge.
Future research
23
However, if other methods are discovered in the future to identify platform firms, then additional data has to be collected potentially. If future researchers also place importance on making a distinction between different levels of platform activity as previously mentioned, then it is necessary to create a continuous variable for platform activity. Since platform firms going public seems to be a recent development, a replication of this study is best delayed in order to allow more firms to go public and a larger quantity of firm years being available for research.
In the conceptualization phase of this study, I also developed three other potential moderating factors. These factors did not originate from financial data and would require additional data collection in order to be researched, hence why these ideas ware abandoned because of the scope of this research. However, I still would like to offer them as suggestions for potential future research.
The first, and most researched, potentially moderating factor is the presence of institutional investors for a given stock. Research by Lungqvist et al. (2007) concluded that when stocks are owned by institutional investors, stock analysts give less optimistic recommendation. The argued cause of this were the periodic evaluations by institutional investors of stock analysts which determine the choice of brokerage firms. Potentially negative evaluations of stock analysts are considered harmful for both an analyst’ reputation and career perspectives. Analyst that acquire a reputation of being unbiased and accurate in their recommendations generate additional business for their brokerage firms (Irvine, 2003; Cowen, Groysberg, and Healy, 2003; Jackson, 2003) as well as job offers from prestigious institutions (Hong and Kubik, 2003; Hong, Kubik, and Solomon, 2000) Thus, I argue that the presence of institutional investors of platform firms would negatively moderate the positive relationship.
24
analysts, positively moderating the positive relationship between platform firms and stock recommendations.
25
References
Ackerberg, D.A. and Gowrisankaran G. “Quantifying Equilibrium Network Externalities in the ACH Banking Industry.” forthcoming, Rand Journal of Economics, 2004.
Algur, S. P., & Biradar, J. G. (2017). Cooks Distance and Mahanabolis Distance Outlier Detection Methods to identify Review Spam. International Journal of Engineering and Computer Science, 6(6), 21638-21649.
Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173-1182.
Black, F. (1973), Yes Virginia, There is Still Hope: Test of the Value Line Ranking System. Financial Analyst Journal, 29, 10-14.
Bloom, M. (2009). Accounting for goodwill. Abacus, 45(3), 379-389.
Buesser, G. (2019). For the investor: internally generated intangible assets. Financial Accoutning Standards Board. Retrieved 04 January 2020 from https://www.fasb.org/jsp/FASB/Page/SectionPage&cid=1176173166185
Cennamo, C., & Santalo, J. (2013). Platform competition: Strategic trade‐offs in platform markets. Strategic management journal, 34(11), 1331-1350.
Copeland, T., Mayers, D. (1982), The Value Line Enigma: A Case Study of Performance Evaluation Issues. Journal of Financial and Quantitative Analysis, 35, 485-498.
Cook, R. D. (1977). Detection of influential observation in linear regression. Technometrics, 19(1), 15-18.
Cowen, Amanda, Boris Groysberg, and Paul M. Healy. (2003). Which types of analyst firms make more optimistic forecasts? Unpublished working paper, Harvard Business School. Cusumano, M. (2010). Technology strategy and management The evolution of platform
thinking. Communications of the ACM, 53(1), 32-34.
Cusumano, M. A. (2011). Platform wars come to social media. Communications of the ACM, 54(4), 31-33.
Davies, P., Canes, M. (1978), Stock Prices and the Publication of Second-Hand Information. Journal of Business, 51 (1), 43-56.
Demos, T. (2020). Alibaba IPO Biggest in History as Bankers Exercise 'Green Shoe' Option. WSJ. Retrieved 20 January 2020, from https://www.wsj.com/articles/alibaba-ipo-biggest-in-history-as-bankers-exercise-green-shoe-option-1411334271
26
Dimson, E., Marsh, P. (1984), An Analysis of Brokers’ and Analysts’ Unpublished Forecasts of UK Stock Returns. Journal of Finance, 39, (5) 1257-1292.
Eavis, E. (2012). Facebook Raises $16 Billion in I.P.O.. Dealbook. Retrieved 20 January 2020, from https://dealbook.nytimes.com/2012/05/17/facebook-raises-16-billion-in-i-p-o/
Eisenmann, T., G. Parker, and M. Van Alstyne. (2006). Strategies for two-sided markets. Harvard Business Review 84: 92–101.
Ellis, M. (2001). Goodwill Accounting: Everything has changed and nothing has changed. Journal of Applied Corporate Finance, 14(3), 103-112.
Fijneman, Kuperus, and Pasman, (2018). Unlocking the value of the platform economy: mastering the good, the bad and the ugly. KPMG N.V.
George, D., & Mallery, P. (2003). SPSS for Windows step by step: A simple guide and reference. 11.0 update (4th ed.). Boston: Allyn & Bacon
Giuliani, M., & Brännström, D. (2011). Defining goodwill: a practice perspective. Journal of Financial Reporting and Accounting, 9(2), 161-175.
Givoly, D., Lakonishok, J. (1979), The Information Content of Financial Analysts’ Forecast of Earnings. Journal of Accounting and Economics, 1, (3), 165-185.
Grant, R. M. (1988). 'On dominant logic' and the link between diversity and performance', Strategic Management Journal, 9, pp. 639-642.
Groth, J., Lease, R., Lewellen, W., Schlarbaum, G. (1979), An Analysis of Brokerage House Securities Recommendations. Financial Analysts Journal, 35, (1), 32-40.
Gulati, R. (1999). Network location and learning: The influence of network resources and firm capabilities on alliance formation. Strategic Management Journal, 20: 397-420.
Hausman, J. A. (1978). Specification tests in econometrics. Econometrica 46: 1251–1271. Hayes, A. F. (2009). Beyond Baron and Kenny: Statistical mediation analysis in the new
millennium. Communication Monographs, 76 408-420.
Helfat, C. E., & Peteraf, M. A. (2009). Understanding dynamic capabilities: Progress along a developmental path. Strategic Organization, 7(1): 91-102.
Hofer, C. W. and D. Schendel. (1978). Strategy Formulation: Analytical Concepts, West Publishing, St. Paul, MN.
Hong, Harrison, Jeffrey D. Kubik, and Amit Solomon. (2000). Security analysts’ career concerns and the herding of earnings forecasts, RAND Journal of Economics 31, 121-144. Hong, Harrison, and Jeffrey D. Kubik. (2003). Analyzing the analysts: Career concerns and
biased earnings forecasts, Journal of Finance 58, 313-351.
27
Jackson, A. (2003) Trade generation, reputation, and sell-side analysts, unpublished working paper, London Business School.
Jacobides, M. G., Knudsen, T., & Augier, M. (2006). Benefiting from innovation: Value creation, value appropriation and the role of industry architectures. Research policy, 35(8), 1200-1221.
Joia, L. A. (2000). Measuring intangible corporate assets. Linking business strategy with intellectual capital. Journal of Intellectual Capital, 1(1), 68-84.
Judge, G. G., Hill, R. C., Griffiths, W. E., Lütkepohl, H., & Lee, T. C. (1988). Introduction to the Theory and Practice of Econometrics.
Klebnikov, S. Everything To Know About Airbnb’s 2020 IPO—Which Could Be Nontraditional. (2019). Forbes. Retrieved 02 January 2020, from https://www.forbes.com/sites/sergeiklebnikov/2019/11/19/heres-everything-to-know-about-airbnbs-2020-ipowhich-could-be-nontraditional/#19e54d5ecff6
Kline, P. (2000). The handbook of psychological testing (2nd ed.). London: Routledge, page 13 Kraaijenbrink, J., Spender, J. C., & Groen, A. J. (2010). The resource-based view: a review and
assessment of its critiques. Journal of management, 36(1), 349-372.
Kristandl, G., & Bontis, N. (2007). Constructing a definition for intangibles using the resource based view of the firm. Management decision, 45(9), 1510-1524.
Krull,J.L. & MacKinnon,D.P. (2001) Multilevel modeling of individual and group level mediated effects. Multivariate Behavioral Research, 36(2), 249-277.
Lev, B. (2004). Sharpening the intangibles edge. Harvard business review, 6, 109-116.
Lhaopadchan, S. (2010). Fair value accounting and intangible assets: Goodwill impairment and managerial choice. Journal of Financial Regulation and Compliance, 18(2), 120-130.
Lichtenthaler, U. (2009). Absorptive capacity, environmental turbulence, and the complementarity of organizational learning processes. Academy of Management Journal, 52(4): 822-846.
Liu, P., Smith, S., Syed, A., (1990), Stock Price Reactions to The Wall Street Journal’s securities Recommendations. Journal of Financial and Quantitative Analysis 25, 399-410.
Ljungqvist, A., Marston, F., & Wilhelm Jr, W. J. (2006). Competing for securities underwriting mandates: Banking relationships and analyst recommendations. The Journal of Finance, 61(1), 301-340.
28
Nielsen, C., & Bukh, P. N. (2011). What constitutes a Business Model: The perception of financial analysts. International Journal of Learning and Intellectual Capital, 8(3), 256-271. Nobes, C. & Norton, J. (1996). International variations in the accounting and tax treatments of
goodwill and the implications for research. Journal of International Accounting, Auditing & Taxation, 5(2), 179-196.
Oh, J., Koh, B., & Raghunathan, S. (2015). Value appropriation between the platform provider and app developers in mobile platform mediated networks. Journal of Information Technology, 30(3), 245-259.
Park, C. W., & Stice, E. K. (2000). Analyst forecasting ability and the stock price reaction to forecast revisions. Review of Accounting Studies, 5(3), 259-272.
Petersen T, Koput KW. (1991). Density dependence in organizational mortality: legitimacy or unobserved heterogeneity? American Sociological Review 56: 399–409.
Rodov, I., & Leliaert, PH. (2002). FiMIAM: financial method of intangible assets measurement. Journal of Intellectual Capital, 3(2), 323-336.
Rochet, J. C., & Tirole, J. (2003). Platform competition in two-sided markets. Journal of the european economic association, 1(4), 990-1029.
Rochet, J. C., & Tirole, J. (2004). Two-sided markets: an overview. Institut d’Economie Industrielle working paper.
Rochet, J. C., & Tirole, J. (2006). Two‐sided markets: a progress report. The RAND journal of economics, 37(3), 645-667.
Sayrs, L. W. (1989). Pooled time series analysis (No. 70). Sage.
Shrout, P. E., & Bolger, N. (2002). Mediation in experimental and nonexperimental studies: New procedures and recommendations. Psychological Methods, 7, 422-445.
Vincent, M. (2019). Loss-making tech companies are floating like it’s 1999. Financial Times. Retrieved 03 January 2020 from https://www.ft.com/content/c5705ca2-605c-11e9-9300-0becfc937c37
Wagenhofer, A. (2005), Internationale Rechnungslegungsstandards – IAS/IFRS, IAS, Frankfurt.
Whitwell, G. J., Lukas, B. A., & Hill, P. (2007). Stock analysts' assessments of the shareholder value of intangible assets. Journal of Business Research, 60(1), 84-90.
Working Group Accounting and Reporting of Intangible Assets (WGARIA). (2005). “Corporate reporting on intangibles”, Schmalenbach Business Review, Vol. 2, pp. 65-100. Wiersema, M. F., & Zhang, Y. (2011). CEO dismissal: The role of investment
analysts. Strategic Management Journal, 32(11), 1161-1182.
29
