Is The Effect Of Conservative Accounting More Pronounced In Tech Stocks?
Business
Tech stocks have been game changers in the global equities market since the dawn of the internet.
Furthermore, over the last thirty years, tech stocks have become known for their ambitious research and development spending. At the same time, the FASB adopted a strict R&D rule in 1974 to limit almost all R&D expenses, citing the absence of a relation between R&D expenditures and subsequent benefits. Similarly, marketing expenses, in their view, demonstrate an even weaker connection between investment and subsequent returns in the future. Therefore, as illustrated by Penman and Zhang (2021), most countries have adopted conservative accounting principles to capture this risk of weak connections.
As also noted by Penman and Zhang (2021), conservative accounting has impacts on ROE (numerator and denominator effects).
In particular, conservative accounting would deflate ROE during the early stages of investments and inflate ROE during the late stages of investments. Therefore, contrary to previous research that treats ROE as a measure of profitability and therefore a proxy of the required return on equity (notably Fama and French (2006), Fama and French (2015), Voulteenaho (2002), Chen and Zhang (2010)), a deeper look into conservative accounting may offer an explanation for whether the conservative accounting effect played a role in contributing to the returns of stocks, and specifically tech stocks.
The research questions are therefore as follows:
- Is the conservative accounting effect more rampant in tech stocks?
Even though tech companies that operate in industries with higher R&D levels tend to spend more on R&D in the years, traditional industries could also spend more money on marketing and brand building (like Coca Cola) that would also build a large conservative accounting reserve. However, the paper finds that despite the lower levels of LIFR and marketing expenses, tech stocks do have higher conservative accounting reserves.
In fact, further analysis shows that some industries do have higher c-scores than others with statistically significant differences. This further expands Lev (1996)’s finding that some industries are more R&D intensive than others. In fact, they also
- Is the relationship between conservative accounting and stock prices pronounced in tech equities?
Do investors care about conservative accounting when doing analysis on tech stocks? Will a lower ROE as a result of conservative accounting lead to higher stock premiums for these companies? The answer is no – a higher level of conservative accounting reserve is not correlated with a higher stock price premium or subsequent returns. This could mean that investors overlook previous investments in technology or brand while valuing tech stocks. This coincides with Lev (1996)’s finding that there seems to be a systematic mispricing of R&D expenses in the stock market, even though the relationship between investment type expenses and revenue growths (as well as returns) is very pronounced.
To be safest certain web properties of 7 figures or greater in revenue should consider a dedicated ecommerce accountant to help with financials.
- When does the conservative accounting effect contribute to the stock price?
Investors have to care about this kind of accounting distortions eventually, so I wanted to find what they actually care about. In fact, the answer to this question is the change in conservative accounting – there is a significant relationship between delta in conservative accounting reserve and stock price premium as well as 1-year-forward stock returns, even though the R squared is not significant.
However, the relationship between delta and both premium and stock returns is positive, meaning that investors favor companies increasing their conservative accounting reserve, or, in another way, the current expenses instead of historical expenses.
However, in reality, historical investments in research and brands are more important than current expenses to investors – historical investments would come into fruition first and the risk associated with them would decrease as they are monetized. The current expense carries more risk and would not return benefits until afterwards.
Therefore, there may be two explanations: 1. This may show evidence that investors tend to overlook historical investments, failing to capitalize the historical R&D, and overfocus on the current R&D/marketing expenses. This would happen until the historical investment finally starts to monetize, leading to higher ROEs. Or 2. Investors feel confident that historical R&D has already been priced in when it was incurred in the past. They choose not to capitalize R&D to avoid double counting the impacts of this expense.
Previous Studies:
Sharpe (1964): CAPM – the most basic explanation of market returns
Fama & French Five Factor Model (2015): ROE, B/P, and size? These may be factors that could influence emerging markets.
Guo, Zhang, Zhang, and Zhang (2017): Size, value and profitability returns strong significance, yet investment only has average returns.
Penman and Zhang (2021): Conservative accounting is rampant and captures the risk of investments and unrealized returns through numerator and denominator effects, thus compensating investors with higher returns when earnings are realized.
Lev (1996): R&D is significantly different for different companies.
Gharbi, Sahut, and Tuelon (2014): R&D investments lead to a higher level of volatility of stock returns for tech stocks.
Contribution:
While there have been a lot of papers trying to explain the high returns in tech stocks and the risks, none of them looked at the conservative accounting angle to explain the outsized returns for these growing companies.
This paper also offers an alternative angle on the conservative accounting problem – this paper favors short-term (1 year) returns on the stock price. I showed that investors, in the short-term, are overlooking capitalized expenses in the past and over focusing on current expenses while valuing companies, as a result of FASB’s rule on expensing R&D.
In real life, we do often see analysts and investors doing analysis on stocks with common features on the income statement and connecting current R&D expenses with higher future growths, which would further contribute to buy recommendations and higher stock prices. Very rarely do we see them capitalize R&D expenses and pay premiums on stocks because of how much money they have spent on R&D and marketing in the last 10 years.
Model:
C-Score: Following Penman and Zhang (2005), I will construct the C-Score based on capitalizing R&D, marketing, and inventory reserves. The reason why I do this instead of Penman and Zhang (2021)’s approach is that most tech companies do not have significant depreciation reserves, bad debt reserves, etc. Further, I want to focus on the “investment” side of the conservative accounting concept. The C-Score should show a risk/reward probability of investments in brands or technology instead of just being conservative on risky accounts receivables and other expenses.
To normalize the conservative accounting reserve, the C-Score would be the percentage of Net Operating Asset (NOA).
Cit = (INVit + RD + ADV ) / NOA
INV: Inventory hidden reserve INV equals the LIFO reserve reported in footnotes. Its value is zero for non-LIFO firms.
RD: Capitalized R&D expense, using the straight-line approach. I followed Lev (1996)’s industry coefficient but used a more simplified approach because of the lack of difference. For tech stocks, I used 7 years, while the rest of the industry would be 5 years.
ADV: Capitalized over two years as suggested by previous literatures. (Bublitz and Ettredge (1989) and Hall (1993))
C-Score is a level, capturing the historical accumulated accounting effects. Delta C will be the yoy changes on C scores. This would be calculated based on the change of C-Score from the last year.
Tech stocks: I followed Walcott (2000)’s approach to determine which industries are more R&D intensive. These are the SIC codes that I have included in the tech subsets:
‘2833’, ‘2834’, ‘2835’, ‘2836’, ‘3571’, ‘3572’, ‘3575’, ‘3577’, ‘3578’, ‘3579’, ‘361’, ‘3651’, ‘3652’, ‘3661’, ‘3663’, ‘3669’, ‘3671’, ‘3672’, ‘3674’, ‘3675’, ‘3676’, ‘3677’, ‘3678’, ‘3679’, ‘3721’, ‘3724’, ‘3728’, ‘3761’, ‘3764’, ‘3769’, ‘3812’, ‘3821’, ‘3822’, ‘3823’, ‘3824’, ‘3825’, ‘3826’, ‘3827’, ‘3829’, ‘3841’, ‘3842’, ‘3843’, ‘3844’, ‘3845’, ‘3861’, ‘4812’, ‘4813’, ‘4822’, ‘4841’, ‘4899’, ‘7371’, ‘7372’, ‘7373’, ‘7374’, ‘7375’, ‘7376’, ‘7377’, ‘7378’, ‘7379’, ‘8711’, ‘8712’, ‘8713’, ‘8731’, ‘8732’, ‘8733’, ‘8734’
Data:
I used data from COMPUSTAT and CRSP. I used annual data from 1990 to 2022 because prior to 1990, the tech stock subset does not have adequate data because of the lack of listings.
First, I dropped all financial firms because their ROE calculation and other accounting metrics follow different rules than the rest of the market. (SIC 6000 – 6099). This leaves 171,820 firm year observations.
Then, I dropped data points with NAN or blank stockholders’ equities and net income figures. I do this because net income/stockholders’ equities data is the most basic type of data of a company, so the void of this means that this company is either delisted or has erroneous data in the database.
Then, I dropped data points with no R&D expense. This is a major manipulation on the dataset as I have dropped almost 30% of the data. However, since most tech companies, my research target, engage in R&D and without appropriate R&D expenses, it would not be an apple-to-apple comparison. Further, R&D expense is a material component of the C-Score calculation, so without this datapoint it would not be significant enough. I have decided to drop data points with NAN or blank research expenses.
Then, I dropped the top and bottom 2% of C-Scores. They are mostly results from extremely high expenses or negative net operating assets. An abnormally high NOA is not representative of the sample as they are mostly pharmaceutical companies or even companies with accounting issues. A negative C-Score is not representative either, since they are mostly companies that are overleveraged with financial liabilities more than equity + cash.
Finally, I dropped some outliers on RNOAs. Some companies have net operating assets close to zero, leading to an insanely high RNOA in that year. This skews the data a lot, so I took the top 0.01% from the dataset.
As a result, here is a description of the data: I have got 79967 firm year data points for the total dataset and 35667 data points in the tech subset.
Total Dataset:
Tech:
As illustrated, the average C-Score in the total dataset is 0.22, while the median is 0, as a result of some skewed effects of the data. This shows that while some companies have a lot of conservative accounting reserves hidden from their balance sheet, average firms do not generally engage in this practice.
However, for companies in the tech subset, the cscore has a median of 0.067794, or 6.779% of the net operating assets. This shows that the median tech companies generally practice conservative accounting because of their ambitious expansion plans.
Then, a correlation table has been performed on key variables.
Total Dataset:
Tech:
To be noted that no significant correlation has been found among these variables of both datasets, except for mcap and NOA, which is very intuitive given a larger balance sheet is typically correlated with a larger market capitalization.
However, it is interesting to see that in both datasets, C-Score exhibited a negative correlation with ROE, even if it is not very significant. This is consistent with what previous literature has shown on the relationship between conservative accounting practices and book returns on equity.
Empirical Analysis:
- C-Score in the tech stocks v. C-Score in the total dataset.
First, I grouped the entire dataset into 99 subsets based on their industries (first two letters of the SIC code). I then took the means of these 99 groups and performed t-tests on all the pairs. Here is a list of industries which exhibited at least one significant difference from another industry (P value < 0.05) :
{’36’: 60, ’38’: 55, ’28’: 59, ’34’: 33, ’50’: 37, ’35’: 54, ’73’: 55, ’26’: 33, ’51’: 26, ’39’: 49, ’87’: 34, ’45’: 20, ’49’: 19, ’22’: 26, ’54’: 41, ’33’: 35, ’67’: 50, ’99’: 43, ’37’: 39, ’20’: 23, ’30’: 26, ’25’: 36, ’59’: 29, ’29’: 34, ’65’: 31, ’48’: 30, ’32’: 34, ’53’: 30, ’13’: 35, ’21’: 17, ’79’: 47, ’24’: 28, ’23’: 32, ’27’: 28, ’31’: 36, ’58’: 44, ’57’: 29, ’80’: 43, ’10’: 25, ’56’: 35, ’63’: 50, ’60’: 16, ’78’: 14, ’70’: 47, ’62’: 32, ’16’: 18, ’52’: 28, ’75’: 43, ’61’: 29, ’17’: 11, ’55’: 39, ’15’: 23, ’64’: 30, ’82’: 19, ’72’: 12, ’14’: 33, ’76’: 58, ’42’: 5, ’46’: 12, ’44’: 11, ’83’: 16, ’12’: 47, ’47’: 20, ’90’: 4, ’40’: 1}
Notably, here are the top ten industries with the most significant pairs:
- 36 (Electronic And Other Electrical Equipment And Components, Except Computer Equipment): 60
- 28 (Chemicals And Allied Products): 59
- 76(Miscellaneous Repair Services): 58
- 38 (Measuring, Analyzing, And Controlling Instruments; Photographic, Medical And Optical Goods; Watches And Clocks) : 55
- 73 (Business Services): 55
- 35 (Industrial And Commercial Machinery And Computer Equipment): 54
- 67 (Holding And Other Investment Offices): 50
- 63 (Insurance carriers): 50
- 39 (Miscellaneous Manufacturing Industries): 49
- 70 (Commercial establishments): 47
While some of these groups have significantly higher C-Scores (like Chemicals, Electronics, Computers, etc.), some groups have significantly lower C-Scores (like Investment holdings, commercial establishments, etc.), it is clear that the practice of conservative accounting differs on industries. Different industries do exhibit different conservative accounting reserves.
Then, I performed another t-test on the tech subset and the others dataset. The p-score was 0, meaning that there is statistically significant evidence to show that the groups selected for the tech subset exhibit significant differences in means of C-Scores from the rest of the dataset.
This shows that despite the fact that traditional companies may also engage in R&D activities, marketing expenses, and conservative inventory practices, their conservative accounting reserve is still lower than the tech companies. There is a significant difference for these tech companies, so it is reasonable to proceed with more analysis on the relationship between C-Score and book rate of returns.
- C-Score and ROE? Is the effect present?
Then, I performed OLS regressions on both datasets to explore the relationship between C-Score and book rate of returns.
First, I performed a regression on the whole dataset:
As a result, the p-score of this regression is lower than 0.05, meaning that there is a significant relationship between C-Score and the book rate of returns for the total dataset. A lower C-Score is correlated with a higher ROE, which is consistent with Penman and Zhang (2002)’s findings on the denominator and the numerator effect on ROE.
Then I performed a regression on the tech subset of the data.
To my surprise, while close, the p score for this regression is slightly higher than 0.05, so I cannot reject the null hypothesis that there is a significant relationship between C-Score and ROE for the tech subset. However, given how close it is, I contemplate that it is a result of the outliers and more negative average ROE for the tech subsets. Because of how risky these companies are, their earnings are often volatile due to reasons unexplained by conservative accounting. Further, the shareholders’ equity is also affected by the volatile earnings in the past due to numerous reasons.
Finally, another explanation is that the C-Score is a level instead of the current year’s changes, so there is a mismatch linking the C-Score to ROE, which is a current ratio. Therefore, I decided to explore the relationship between delta-C and ROE. I calculated this as the difference between this year’s C-Score and last year’s C-Score. As a result, I got 16609 valid delta-c data points in the tech subset, and 34400 valid delta-c data points in the total dataset. Then, I performed an OLS regression on delta-c and ROE in the tech subset.
The result shows that there is significant statistical evidence to show a decrease in C-Score, or a negative delta-c, is correlated with an increase in ROE. This is, again, consistent with Penman and Zhang (2002)’s findings. Therefore, the relationship between conservative accounting and book rate of returns is still present for these tech companies.
- Connecting C-Score and Market Rate of Returns
I want to connect C-Score and market rate of return for tech companies. Since there is a negative relationship between C-Score and ROE, then the relationship between C-Score and E/P can be boiled down to the C-Score and P/B, which is the premium investors are paying for this company, because P/E = P/(ROE * B) = P/B * (1/ROE)
However, there is no significant evidence to show such a relationship. C-Score seems to be uncorrelated with E/P. This means that, even though the C-Score has a significant relationship with ROE, the book rate of return, it has little effect on the market pricing the stocks for investors.
Similarly, I have found no significant relationship between C-Score and 1-year forward returns. The current level of C-Score seems to have no effect on the near-term market returns of a company. Investors tend to overlook the cumulative effects of conservative accounting while doing valuations of a company. In another word, they do not seem to care about how much a company has spent in the past on R&D and other conservative accounting practices.
An analysis on delta-c shows another story:
There is a significantly negative relationship between the change in C-Score level and the premium investors are paying for this company as proxied as E/P. When the C-Score increases, indicating that the company are engaging in riskier investments which are often expensed on the income statement, investors price that in while doing valuations on that stock. Similarly, there is a positive relationship between delta-c and 1-year forward returns.
As shown above, another regression between delta c and P/B confirms that the change in C-Score does have a significant relationship with the premium investors are paying for this company. The higher it increases, the more people are paying for this stock and the more returns this stock would generate in the future. This could be argued that investors are pricing in the potential R&D/marketing returns when these investments are announced, and some of the 1-year forward return comes from the successful fruition of investments.
This contradicts the very reason why conservative accounting exists – investors seem to care a lot about current spendings on R&D and marketing instead of the cumulative spendings on R&D.
There could be two explanations: 1. Investors do not capitalize conservative accounting, but they understand the implications of these “investment” style expenses. In this scenario, tech companies announce ambitious investments that have to be expensed, but investors are pricing in potential upsides. This is different from other industries where riskier investments carry risks that would require higher returns and lower current premiums. This growth potential drives up the price and normally will also bring in higher returns. 2. Investors believe that the historical investments have been perfectly priced in. If every move in conservative accounting reserves has been correlated with premium and stock returns, then the cumulative impact is not relevant to the current price. What really matters is the new information – the current expenses that would contribute to the changes in the C-Score.
Regardless, this has two implications: 1. Conservative accounting does have a relationship with the stock performance and is of importance to investors. 2. Investors care about new information, believing that all the historical expenses, even though they were expensed and no longer shown on the financial statements and their impacts have come into effect, have been priced in. This is incorrect as conservative accounting conveys the risk/reward of new investments and the stock performance should depend on how much “risk” this company still carries and how much “risk” has been either monetized or turned into losses. The level of conservative accounting should play a role, but it is the change of the level that actually played a role in the market rate of returns.
Therefore, when tech companies convince the investors that there are still opportunities by investing heavily in marketing or R&D and driving up the C-Score, they convey growth potential that is the main driver behind stock returns instead of actual profitability. In fact, tech stocks have much lower median ROE throughout the firm-years. When they stop investing, investors would view this unfavorably and, after the growth potentials have been stripped from the valuation, the profitability of the business normally cannot sustain the current valuation and will lead to lower stock returns in the future.
Conclusion:
Tech stocks are by nature riskier and more ambitious in expansion, and conservative accounting captures this nature – the tech subset had a higher C-Score than the rest of the data with statistical significance. Further, the denominator and numerator effects are still pronounced in this industry. Finally, investors seem to care more about current expenses than the level of capitalized expenses. The market appears to have priced in the change in C-Score rather than the C-Score level already, even though the C-Score level should be more relevant to stock performance as it is more relevant to near-term growth.
This leads me to believe that investors care about growth potential when it comes to tech stocks, which generally have lower ROE and RONA compared to the rest of the dataset. Every time management announces ambitious marketing and research plans, investors view them favorably but fail to recognize how much the company has already spent on such ventures. Further, when management decides to cut back on these expenses, the premium drops and the hype stops for these names as their current growth and profitability may not be able to support their high valuation.
We see this all the time in the market – companies’ prices go up after the announcement of ambitious plans or when there is hype, such as with EVs, lithium, semi, or even Chinese e-commerce. However, when the company starts making money by cutting its marketing or research expenses to focus on profitability, the price or valuation either takes a long time or never returns to its peak during the hype, even though its ROE is much higher than before, and profitability makes it a safer bet for investors. In some cases, growth does not slow down for a couple more years due to past investments (already high C-Score level), but investors’ willingness to pay a premium decrease drastically.
Further studies could be done to create a systematic trading algorithm to determine the peak and valley of C-Scores over a life cycle to monetize this trading opportunity.
Written by Tianyi Steven Li
Reference
Gharbi, Sami & Sahut, Jean-Michel & Teulon, Frederic. (2014). R&D investments and high-tech firms’ stock return volatility. Technological Forecasting and Social Change. 88. 306–312.
Hall, B. 1993. Industrial research during the 1980s: Did the rate of return fall? Brookings Papers
on Economic Activity, Microeconomics, 289-344
Lev, B., and T. Sougiannis. 1996. The capitalization, amortization and value-relevance of R&D.
Journal of Accounting and Economics 21, 107-138
Penman, S. and X. Zhang. 2002. Accounting conservatism, the quality of earnings, and stock
returns. The Accounting Review 77, 237-264
Sharpe, W.F. 1964. Capital Asset Prices: A Theory of Market Equilibrium under Conditions of Risk. Journal of Finance, 19, 425-442.
Vuolteenaho, T. 2002. What drives firm-level stock returns? Journal of Finance 57, 233-264.
Walcott, Susan. 2000. Defining and Measuring High Technology in Georgia.
Zhang, X. 2000. Conservative accounting and equity valuation. Journal of Accounting and
Economics 29, 125-149.
