DO LOSS FIRMS MANAGE EARNINGS AROUND SEASONED EQUITY OFFERINGS?

DO LOSS FIRMS MANAGE EARNINGS AROUND SEASONED EQUITY OFFERINGS? Fernando Comran, Unversty of San Francsco, School of Management, 2130 Fulton Street, CA 94117, Unted States, fcomran@usfca.edu Tatana Fedyk, Unversty of San Francsco, School of Management, 2130 Fulton Street, CA 94117, Unted States, tfedyk@usfca.edu Joohyung Ha, Unversty of San Francsco, School of Management, 2130 Fulton Street, CA 94117, Unted States, jha6@usfca.edu ABSTRACT The unqueness of ths paper s n lnkng frms valuaton to ther dscretonary choces and by demonstratng that loss frms do not manage earnngs durng as earnngs are not nformatve for ther valuaton. Instead, we fnd that loss frms manage upward research and development expendtures (R&D),.e., overnvest n R&D, consstent wth hgh R&D valuaton for loss frms. We further show that overnvestment n R&D s negatvely assocated wth future stock return for loss frms, and, therefore, market does not recognze loss frms dscreton. BODY In ths paper, we examne earnngs management among loss seasoned equty offerngs (). Whle pror lterature on earnngs management around generally concludes that SEO frms manage earnngs, we demonstrate that only proftable frms nflate ther earnngs, whereas loss frms do not manage earnngs. Instead, they nflate research and development expendtures (R&D), a strategy that s actually decreasng bottom lne earnngs. Our fndng s consstent wth dfferental valuaton for proft and loss frms, and hghlghts the mportance of more accurate approach n earnngs management research to study accountng choces for proft and loss frms. Pror lterature on managers reportng behavor around manly concentrates on two types of earnngs management, accrual earnngs management (Teoh et al., 1998; Rangan, 1998; Shvakumar, 2000; DuCharme et al., 2004) and real earnngs management (e.g., Cohen and Zarown, 2010). The general concluson s that SEO frms overstate ther earnngs by usng accrual-based or transacton-based earnngs management. By manly focusng on earnngs, t s mplctly assumed that earnngs are the man focus of manageral dscreton durng. Whle ths assumpton sounds reasonable for proftable frms as earnngs are value-relevant for ther valuaton (e.g., Ohlson, 1995; Graham et al., 2005), earnngs management among loss frms s reasonably questonable manly because earnngs are not that nformatve for them. For example, Hayn (1995) demonstrates that loss frms have lower earnngs response coeffcent than proftable frms, explanng such dfferental percepton of earnngs for loss vs. proft frms by the exstence of a lqudaton opton. Callen et al. (2008) suggest that for loss frms, tradtonal earnngs-based valuaton models do not allow for relable market value estmates, and demonstrate that market partcpants value loss frms based on the level and growth n revenues rather than earnngs. Usng the resdual-ncome valuaton model (Ohlson, 1995), Franzen and Radhakrshnan (2009) show that the valuaton multpler on R&D expendtures s dfferent for proft and loss frms: t s negatve for proft frms, but postve for loss frms. Based on pror lterature fndngs that earnngs are not that value-relevant for loss frms (e.g., Hayn, 1995; Demers and Lev, 2001; Callen et al., 2008), we post that managers of loss SEO frms do not engage n earnngs management, but rather manage other accountng tems that

are mportant for ther frms valuaton: R&D expendtures. To test our predctons, we frst separate all SEO frms nto two groups: proft and loss frms, and then consder whether earnngs and R&D expendtures are managed dfferently by these two groups. We expect that only proftable frms manage earnngs upward; whereas for loss frms we do not expect to fnd evdence of earnngs management, but rather R&D management. Our sample perod begns n 1989, as statements of cash flow become avalable, and ends n 2011. The lst of IPO frms, data on underwrter names, venture captalst (VC) backng, offer prce, and market captalzaton are obtaned from the Securtes Data Corporaton Global New Issues Database by Thompson Fnancal. Foundng dates and underwrters reputaton rankngs are obtaned from Jay Rtter s webste (http://bear.warrngton.ufl.edu/rtter/podata.htm). Fnancal accountng data s from Compustat. Patent data s obtaned from the Natonal Bureau of Economc Research (NBER) webste (http://www.nber.org/patents/) that contans nformaton for all patents granted n the U.S. from 1975 to 2006. For the perod 1996 2000, we calculate ownershp retenton and market captalzaton usng data manually collected by Alexander Ljungqvst. We follow the IPO lterature and exclude from the sample unt offerngs, Real Estate Investment Trust offerngs, Amercan Depostory Recepts, closed-end funds, and frms n regulated ndustres (SIC 4910 4939) and fnancal nsttutons (SIC 6000 6999). We further exclude IPOs wth an offer prce of less than $1 per share or market value of less than $20M mmedately after ssue. Fnally, as many of our varables are scaled by average total assets, to avod a small denomnator problem, we remove frms wth average total assets of less than $100,000. Our fnal sample ncludes 3,763 IPO frms. We start our analyss by examnng the weght SEO nvestors put on dfferent tems n the fnancal statements (sales growth, R&D expendtures and earnngs) for all SEO frms, and separately for proft and loss frms. Ths analyss provdes a benchmark to see what drves the market value for two groups of frms, and where n turn, managers mght be tempted to focus ther efforts of dscreton (assumng that frm s valuaton s a major ncentve for managers durng ). The results reported n Table 1 below show that proft and loss SEO frms have dfferent accountng value drvers, consstent wth our expectatons and fndngs n pror lterature. Thus, proft frms are manly prced on earnngs whereas loss frms are not prced on earnngs at all. On the other hand, loss frms are postvely prced on R&D expendtures whereas proft frms are not prced on R&D. TABLE 1 Prcng of IPO Frms Coeffcent estmates from an ordnary least squares regresson (OLS) of the market value of equty, MV, at the end of the month after the frst annual fnancal statements of the SEO frms are reported, on postve and negatve earnngs, PEBXI_noRD and NEBXI_noRD, adjusted for R&D expendtures, R&D expendtures, RD, the natural logarthm of sales growth, LOGSGR, the natural logarthm of the frm s age, LOGAGE, the rato of total labltes to average total assets, LEV, and ndustry prce-to-earnngs rato, IND_PE. MVE, PEBXI_noRD, NEBXI_RD and RD are scaled by average total assets, TA. MV = β + β PEBXI _ nord + β NEBXI _ nord + β RD + β LOGSGR + β LOGAGE + β LEV + β IND _ PE + ε 5 0 1 6 7 2 Varable Proft Intercept 4.18 3.1 5.34 (16.98) (10.54) (10.66) PEBXI_noRD 5.97 10.65-3 4 2

(9.58) (11.20) - NEBXI_noRD -0.134-0.666 (-0.31) - (1.27) RD 6.8-1.00 6.12 (11.73) (-0.63) (7.65) LOGSGR 1.83 2.66 1.13 (12.27) (12.12) (4.97) LOGAGE -0.62-0.48-0.874 (-8.18) (-5.97) (-4.64) LEV -2.055-1.958-1.903 (-9.81) (-7.81) (-4.87) IND_PE -0.053-0.049-0.035 (-5.39) (-4.41) (-1.61) Year dummes Yes Yes Yes Adjusted R 2 22.52% 26.32% 17.92% N (obs.) 3,130 2,369 761 Next, we estmate dscreton over earnngs and R&D for all SEO frms, and separately for proft and loss frms. Consstent wth pror lterature results that accruals are abnormally hgh around (Rangan, 1998; Teoh et al., 1998; Shvakumar, 2000; DuCharme et al., 2004), we also document sgnfcantly postve abnormal accruals for a full SEO sample. However, when we splt all SEO frms nto two groups, we fnd that dscretonary accruals are abnormally hgh only for proft frms, whle for loss frms dscretonary accruals are actually sgnfcantly negatve. Ths result demonstrates n Table 2 below that loss SEO frms do not manage earnngs upward durng seasoned equty offerngs. Moreover, we document that loss SEO frms overnvest n R&D (a strategy that actually reduces earnngs) as R&D expendtures are postvely valued by nvestors for these frms. Ths new fndng hghlghts the mportance of lnkng frms valuaton to ther dscretonary accountng choces nstead of usng a default assumpton of prevalng ncentves for earnngs management. TABLE 2 Tme-seres Profle of Dscretonary Accruals and Dscretonary R&D Panel A: Mean and medan asset scaled dscretonary accruals, n percent, for all SEO frms and by group, from year -1 to +3 relatve to the seasonal equty offerng (year 0). Dscretonary accruals are defned as the dfference between the scaled current accruals and ftted values from the estmaton regresson. Proft Year Mean Medan Mean Medan Mean Medan -1-0.00 0.99 *** 1.19 *** 1.36 *** -3.64 *** -0.68 *** 0 1.80 *** 2.13 *** 4.64 *** 3.05 *** -6.63 *** -1.57 *** +1 0.57 * 1.47 *** 1.58 *** 1.85 *** -2.63 *** -0.32 +2 0.25 1.09 *** 0.69 *** 1.13 *** -1.29 * 0.75 +3 0.23 0.89 *** 0.27 0.88 *** 0.13 0.92 Panel B: Mean and medan asset scaled dscretonary R&D, n percent, for all SEO frms and by group, from year -1 to +3 relatve to the seasonal equty offerng (year 0). Dscretonary 3

R&D expendtures are defned as the dfference between the scaled R&D and ftted values from the estmaton regresson. Proft Year Mean Medan Mean Medan Mean Medan -1 1.15 *** 0.00 *** 0.89 *** 0.00 ** 1.81 *** 0.09 *** 0 0.35 ** 0.00-0.35 *** 0.00 *** 2.17 *** 0.05 *** +1 0.19 0.00 ** 0.25 * 0.00 0.04-0.12 *** +2-0.14-0.00 *** 0.05 0.00 * -0.72 * -0.30 *** +3-0.18-0.00 *** -0.16 0.00 * -0.25-0.19 * ***, **, * Denotes sgnfcance at the <.01, <.05, and <.10 levels, respectvely. Fnally, we examne whether dscretonary tems durng are assocated wth post- SEO operatng underperformance, what would be consstent wth opportunstc or myopc motves drvng manageral reportng. Alternatvely, f managers use dscreton over accountng tems to sgnal superor future performance (.e., sgnalng hypothess), dscretonary tems wll be postvely assocated wth future performance. Consstent wth pror lterature (Teoh et al., 1998, Cohen and Zarown, 2010), we fnd that dscretonary accruals are negatvely related to future operatng performance, but we document that ths assocaton holds only for proft frms. Next, n Table 3, we provde new evdence that dscretonary R&D expendtures are also negatvely related to future operatng performance; but ths relaton s completely drven by loss SEO frms. Therefore, we do not fnd evdence n support of the sgnalng hypothess. Our fndngs of negatve assocaton between dscretonary tems (accruals for proft frms, and R&D expendtures for loss frms) are consstent wth opportunstc or myopc manageral behavor. To sum, we present robust evdence that durng the, frms manly manage accountng tems that are value-relevant for nvestors, and that loss frms do not manage earnngs, but manage R&D nstead. TABLE 3 Future Operatng Performance and Dscretonary Items for SEO Frms by Group Regresson analyss of the assocaton between a change n return on assets, ROA from pre- SEO year (year -1) to one year after the SEO (year +1) and dscretonary accruals, DACC, and dscretonary R&D, DRD. Panel A: Regresson analyss of the assocaton between a change n return on assets, ROA, from pre-seo year (year -1) to one year after the SEO (year +1) and dscretonary accruals, DACC, for all SEO frms and by group: ROA = β + β DACC + β SGR + β CAPEX GR + ε Δ t 1, t+ 1 0 1, t 2, t 4 _, t Model I raw ROA Model II performance-adjusted ROA Varable Proft Proft Intercept 0.0009-0.0062 * 0.0623 *** 0.0023 0.0139 0.0129 (0.15) (-1.66) (4.07) (0.31) (1.83) (0.64) DACC -0.0927 *** -0.1652 *** 0.0551-0.0656 ** -0.1709 ** 0.0702 (-2.35) (-4.12) (0.81) (-2.29) (-2.18) (0.54) SGR 0.0185 * 0.0077 ** 0.0229 0.0127 ** 0.0226 0.0159 (1.83) (2.28) (1.33) (2.29) (1.50) (1.34) CAPEX_GR 0.0000 *** -0.0026-0.000 *** 0.0000 0.0050 0.0000 4

(2.64) (-0.97) (-3.61) (0.46) (0.78) (1.61) Adjusted R 2 2.24% 1.88% 2.94% 0.51% 1.25% 0.94% N (obs.) 3,113 2,364 609 2,553 1,989 564 ***, **, * Denotes sgnfcance at the <.01, <.05, and <.10 levels, respectvely. Panel B: Regresson analyss of the assocaton between a change n return on assets, ROA from pre-seo year (year -1) to one year after the SEO (year +1) and dscretonary R&D, DRD, for all SEO frms and by group: ROA = β + β DRD + β SGR + β CAPEX GR + ε Δ t 1, t+ 1 0 1, t 2, t 4 _, t Model I raw ROA Model II performance-adjusted ROA Varable Proft Proft Intercept 0.0061 0.0308 *** -0.1356 *** 0.0069 0.0093-0.0051 (0.58) (3.62) (-4.75) (0.61) (0.87) (-0.21) DRD -0.6866 ** -0.0759-0.7369 * -0.6706 *** 0.0527-0.7971 *** (-2.07) (-0.21) (-1.71) (-2.68) (0.15) (-3.35) SGR -0.0064 0.0155-0.0140 0.0228 0.0231 0.1434 *** (-0.61) (0.82) (-1.05) (1.40) (1.27) (6.59) CAPEX_GR -0.0217 *** 0.0095-0.0319 *** -0.0094 0.0046-0.0086 * (-3.91) (1.22) (-3.45) (-1.60) (0.59) (-1.65) Adjusted R 2 3.58% 0.60% 7.21% 2.07% 0.83% 11.94% N (obs.) 2,051 1,475 576 1,594 1,188 406 ***, **, * Denotes sgnfcance at the <.01, <.05, and <.10 levels, respectvely. Varables are defned as follows: DACC = Dscretonary accruals; Dscretonary accruals are defned as the dfference between the scaled current accruals and the ftted values from the estmaton regresson. DRD = Dscretonary R&D expendtures. Dscretonary R&D are defned as the dfference between the scaled R&D and the ftted values from the estmaton regresson. CAPEX_GR = Growth n captal expendtures, defned as percentage change n captal expendtures: ((CAPEX t - CAPEX t-1 )/CAPEX t-1 ). Our paper makes mportant contrbutons to the SEO earnngs management lterature. Frst and foremost, we lnk SEO frms valuaton wth the use of dscreton over varous accountng tems. Our study s unque n that t dentfes a settng where dfferent frm types have the ncentves to apply dscreton over dfferent accountng tems, and fnd results consstent wth the ncentves. Next, we are the frst to document that loss frms do not manage earnngs durng, but they do manage R&D,.e., they overnvest n R&D as R&D expendtures are mportant value-drver n ther valuaton. Fnally, we demonstrate that dscretonary tems are assocated wth future underperformance: dscretonary accruals are negatvely related to future performance for proft frms, whereas dscretonary R&D expendtures are negatvely related to future performance for loss frms. Whle pror lterature fnd negatve assocaton between dscretonary accruals and future performance for SEO frms n general, we show that ths assocaton s manly drven by proft frms. REFERENCES Callen, J., S. Robb, and D. Segal. 2008. Revenue manpulatons and restatements by loss frms. Audtng: A Journal of Practce and Theory 27(2): 1-29. 5

Cohen, D., and P. Zarown. 2010. Accrual-based and real earnngs management actvtes around seasoned equty offerngs. Journal of Accountng and Economcs 50: 1-33. Demers, E., and B. Lev. 2001. A rude awakenng: Internet shakeout n 2000. Revew of Accountng Studes 6(2-3): 331 359. DuCharme, L., P. Malatesta, and S. Sefck. 2004. Earnngs management, stock ssues, and shareholder lawsuts. Journal of Fnancal Economcs 71: 27 49. Franzen, L., and S. Radhakrshnan. 2009. The value relevance of R&D across proft and loss frms. Journal of Accountng and Publc Polcy 28(1): 16-32. Graham J., C. Harvey, and S. Rajgopal. 2005. The economc mplcatons of corporate fnancal reportng. Journal of Accountng and Economcs 40: 3-73. Ohlson, J. 1995. Earnngs, book values, and dvdends n equty valuaton. Contemporary Accountng research 11(2): 661-687. Hayn, C. 1995. The nformaton content of losses. Journal of Accountng and Economcs 20: 125 153. Rangan, S. 1998. Earnngs before seasoned equty offerngs: Are they overstated? Journal of Fnancal Economcs 50: 101 122. Shvakumar, L. 2000. Do frms mslead nvestors by overstatng earnngs before seasoned equty offerngs? Journal of Accountng and Economcs 29: 339-371. Teoh, S., I. Welch, and T. Wong. 1998. Earnngs management and the underperformance of seasoned equty offerngs. Journal of Fnancal Economcs 50: 63-99. 6