# «Elizabeth Blankespoor University of Michigan Stephen M. Ross School of Business blankbe January 2012 Abstract This paper examines the ...»

TagsNotes, Post, XBRL, and the control variables are as defined earlier. ComplicatedFirms represents firms with a complicated information environment, based on three measures: MultipleIndustries, EarningsVolatility, and AnalystDispersion. MultipleIndustries is an indicator variable equal to one if the firm has operating segments in different industries, using Compustat’s Segment data and defining industries using 3-digit SICs, and zero otherwise. EarningsVolatility is an indicator variable equal to one if the firm has above median earnings volatility, where earnings volatility is defined as the firm disclosure of quantitative footnote disclosures, I examine the relation between changes in analysts and changes in quantitative footnote disclosures, lag changes in analysts and changes in disclosure, changes in institutional holding and changes in disclosure, and lag changes in institutional holding and changes in disclosure. In all cases, I find a positive relation, consistent with findings in prior literature.

standard deviation of the change in split-adjusted earnings per share over the previous five years (including the current year) (Waymire 1985).23 AnalystDispersion is an indicator variable equal to one for firms that have an above-median level of analyst dispersion and zero otherwise, where analyst dispersion is the median monthly standard deviation of analyst earnings forecasts for the fiscal period during the 12 months prior to the fiscal period end, scaled by the mean analyst estimate (Roulstone 2003).

For all three measures, if the decrease in investor processing costs affects complicated firms relatively more and thus has a larger impact on the disclosure choice of firms, the coefficient for Post*XBRL*ComplicatedFirms (β2) should be positive. Table 5 reports the results of Model 3 for all measures of complicated firms, showing that although the average non-complicated XBRL firm increases their disclosure, the increase is even greater for complicated XBRL firms, significant at the 5% level or better. Specifically, the disclosure increase is twice as large for complicated firms, or 160 versus 80 footnote numbers. The larger impact of XBRL adoption on complicated firms’ disclosure is consistent with the reduction in investor processing costs driving the increase in firm disclosure.

Sophisticated Investors To test whether firms with more sophisticated investors or stakeholders (i.e. those with inherently lower processing costs) have a smaller increase in disclosure as investor processing costs decrease (P3), I estimate the following regression with year and firm fixed effects and firm-clustered standard

**errors:**

TagsNotesi,t = β0 + β1Post*XBRLi,t + β2Post*XBRL*SophisticatedInvi,t + β3 SophisticatedInvi,t +

To be clear, I calculate the median value for the XBRL and non-XBRL groups separately and assign the indicator variable value using the median value for the group which the observation belongs to. I choose this method for all cross-sectional indicator variables to ensure each group has a comparable number of high versus low observations.

TagsNotes, Post, XBRL, and the control variables are as defined earlier. SophisticatedInv is an indicator variable equal to one if the firm has investors with better processing ability and zero otherwise, based on two definitions: Analysts and Institutions.

Analysts is an indicator variable equal to one if the number of analysts covering the firm is above the median and zero otherwise, and Institutions is an indicator variable equal to one if the percent of shares owned by institutions is above the median and zero otherwise.24 If the decrease in investor processing costs helps sophisticated investors relatively less and thus has a smaller impact on the disclosure pressure for the firms they follow, the coefficient for Post*XBRL*SophisticatedInv (β2) should be negative.

Table 6 provides the results of the sophisticated investor model for both Analysts and Institutions.

The coefficient for Post*XBRL*SophisticatedInv is negative and significant at the 10% level or better in both regressions. In addition, the coefficient for Post*XBRL is positive and significant at the 1% level, and the sum of these two coefficients is positive and significant at the 1% level for both regressions. These results combined indicate that firms with more sophisticated investors still increase quantitative footnote disclosure upon adoption of XBRL detailed tagging, but the increase is significantly less than that for firms with less sophisticated investors. Specifically, firms with sophisticated investors increase quantitative footnote disclosures approximately half as much as firms with less sophisticated investors (approximately 90 versus 180 numbers in the footnotes). The impact of XBRL on disclosure varies based on the inherent processing costs of firms’ investors, providing additional support for the increase in firm disclosure being driven by an anticipated reduction in investor processing costs.

For all cross-sectional indicator variables (EarningsVolatility, AnalystDispersion, Analysts, and Institutions), I choose to separate observations into two groups (High and Low) because it simplifies interpretation. However, inferences are unchanged if the cross-sectional variables are separated into quartiles or deciles.

5. Additional Tests and Robustness Analyses

5.1 Additional Tests I perform two additional tests to better understand the quantitative disclosure measure and to confirm that the results are not bring driven by non-discretionary or non-informative disclosures.

Financial Instruments and Derivatives To control for changes in disclosure requirements during my period, I include year fixed effects.

However, this assumes that the altered disclosure requirement affects all firms equally. If a change in disclosure requirements systematically affects XBRL firms differently from non-XBRL firms, it could affect the difference-in-difference coefficient. To provide some comfort that my results are not being driven by disclosure requirement changes, I focus on two of the largest disclosure changes during my period: financial instruments and derivatives. SFAS 157 increases the amount of disclosure required related to fair values of financial instruments, and it went into effect for fiscal 2008 (periods beginning after November 15, 2007). SFAS 161 increases the amount of disclosure required related to derivatives and hedges, starting in fiscal 2009 (periods beginning after November 15, 2008). I measure the extent of each firm-year’s disclosure related to financial instruments and to derivatives by counting the number of times the words financial instrument and derivative appear in the footnotes, respectively (NumWords_FinInstr and NumWords_DerivHedge).25 To confirm that these count variables capture the impact of changes in financial instrument and derivative disclosure requirements, I examine their movement over time. For firms with financial instruments (i.e. firm-filings with at least one mention of financial instruments in their footnotes), the number of financial instrument-related words increases over the period, with the largest increase happening the year SFAS 157 went into effect (fiscal 2008). For firms with derivatives, the number of derivative-related words also increases over the period, with the largest increase happening the Specifically, I count the number of derivative-related words (i.e. derivative, derivatives, hedge, hedges, hedging, hedged) and financial instrument-related words (i.e. financial instrument, financial instruments).

year SFAS 161 went into effect (fiscal 2009). Given these movement patterns, the two count variables appear to capture the impact of mandatory disclosure changes. Note, however, that firms could also have chosen to voluntarily provide more quantitative information related to financial instruments and derivatives. By controlling for these count variables, I remove both mandatory and voluntary disclosure related to financial instruments and derivatives, thus biasing against finding an XBRL-related disclosure increase.

Table 7 Panels A, B, and C provide the main results including NumWords_FinInstr and NumWords_DerivHedge. As shown, the impact of XBRL is smaller but still significant. In the main difference-in-difference test, XBRL firms increase their quantitative footnote disclosure by 111.9 more numbers than non-XBRL firms. Complicated XBRL firms increase their disclosure more than simple XBRL firms, and XBRL firms followed by sophisticated investors increase their disclosure by a smaller amount. All coefficients are significant at the 10% level or better, except for the Analysts cross-sectional cut, which has a p-value of 0.112.

Non-Zero Quantitative Filings Firms adopting XBRL may choose to restructure the formatting of their footnotes to reduce the cost of tagging going forward. Specifically, organizing quantitative disclosures into tables and ensuring that every year has a value – even if that value is zero – makes it easier to automatically roll forward tags in subsequent years. This paper’s primary measure of disclosure – Notes_tags – counts all numbers provided in the footnotes, including zeroes. I choose to include zeroes in the count because disclosures of the absence of a financial item can be informative for investors. However, if the zero is simply a result of firms adjusting their formatting and “filling in” empty blanks, it could be less informative. To ensure that the increase in zeroes is not driving my results, I rerun my main analyses using non-zero quantitative footnote disclosures as the dependent variable. As Table 8 Panels A, B, and C show, the impact of XBRL is slightly smaller but still significant. In the main difference-in-difference test, XBRL firms increase their quantitative footnote disclosure by 124.8 more numbers than non-XBRL firms. Complicated XBRL firms increase their disclosure more than simple XBRL firms, and XBRL firms followed by sophisticated investors increase their disclosure by a smaller amount. All coefficients are significant at the 10% level or better.

5.2 Robustness Analyses I also conduct several robustness tests. First, I repeat my main analyses using a subset of the observations to attempt to compare XBRL and non-XBRL firms more similar in size. Since the XBRL adoption requirements are based on firms’ market float, XBRL firms are by definition much larger than non-XBRL firms. I address this in my main tests by including numerous control variables – including the log of market value – that have historically been related to disclosure choice, as well as firm fixed effects. With these controls in place, differences between the XBRL and non-XBRL firms should be accounted for. However, as an alternate approach, I restrict my sample to the smallest 50% of XBRL observations and largest 50% of non-XBRL observations, in terms of market float. The advantage of using this subset of observations is that the comparability of firms increases, with the mean market float of XBRL firms being $6.6 billion and non-XBRL $1.9 billion, rather than the $18.9 billion and $1 billion for XBRL and non-XBRL firms in the full sample. However, the disadvantage is that I lose half the observations and am now estimating the effect of XBRL for just those firms in the range of $260 million to $11.4 billion, rather than all firms, which changes the generalizability of the inferences. Using this subsample, the results are similar but weaker, with one coefficient becoming marginally insignificant (Post*XBRL*Analysts with a p-value of 0.127) and Post*XBRL*Institutions becoming insignificantly different from zero. Overall, the results are still consistent with firms increasing disclosure upon adoption of XBRL in anticipation of reduced investor processing costs.

Second, I separately identify firms that voluntarily adopted XBRL’s detailed tagging requirements. Of the 323 XBRL firms that filed detailed requirements for fiscal 2010, 25 of them could be considered voluntary detail filers, either because they began filing (quarterly) XBRL statements before the adoption date or because they were below the $5 billion market float requirement for fiscal year end 2010. Since they voluntarily chose to provide XBRL statements, their disclosure incentives could be different from mandatory adopters. If they started filing before the mandatory date, they could have altered their disclosure in the pre-adoption period as part of their overall early approach, biasing against finding results of increased disclosure from the pre- to postadoption period. If they provided the detailed XBRL filings even though they were below the $5 billion cutoff, these firms could be more responsive to investor needs and thus be driving the overall results. Therefore, I rerun my main analyses and include Vol and Post*Vol variables in the main regression, as well as an interaction term Post*Vol*Cross-SectionalVariable for each of the crosssectional tests. For all tests, the inferences on the main variables remain the same, reducing any concerns that the voluntary filers are driving the results. Consistent with these findings, when I examine more closely the firms identified as voluntary, 20 of the 25 are “threshold crossers,” or firms that dropped below the $5 billion threshold after initial adoption of basic XBRL requirements but continued to file XBRL statements as Tier 1 filers. Because these firms were effectively mandatory adopters, I would expect their behavior to be similar to mandatory XBRL firms and thus not inappropriately biasing the results.

Third, I rerun my main analyses in a non-XBRL adoption year as a pseudo-adoption year. If the change in disclosure is related to the adoption of XBRL detailed tagging requirements and not other factors, I should not find a change in disclosure for XBRL firms in a non-adoption year. I drop fiscal years after the SEC’s announcement of XBRL requirements (fiscal years 2009 and 2010), and I set fiscal year 2008 as the pseudo post-adoption year. Using fiscal years 2006 and 2007 as the pseudo pre-adoption years, I do not find a difference in quantitative footnote disclosures for XBRL firms in the pseudo post-adoption year relative to non-XBRL firms, providing additional evidence that the change in quantitative footnote disclosure for XBRL firms is related to the regulatory change.