A Time-Series Analysis of Nonseasonal Quarterly Earnings Data

Statistically based quarterly earnings expectation models for nonseasonal firms

“…3 We agree with the findings of previous works that have shown that a majority of firms exhibit seasonal quarterly earnings properties (e.g., Brown and Rozeff 1979;Foster 1977; Lorek and Willinger 2007, among others), however, our results indicate that a sizable and growing percentage of firms exhibit quarterly earnings patterns that are clearly nonseasonal in nature. The percentage and number of nonseasonal firms identified in extant work have been relatively small but increasing [i.e., 12.1% (n = 29) in Lorek and Bathke (1984); 16-18% (n = 155) in Brown and Han (2000), and 28.2% (n = 167) in Bathke et al (2006)]. In contrast, we identify 35.6% (n = 296) of our sample as nonseasonal firms using Lorek and Bathke's (1984) nonseasonality screening filter.…”

“…Lorek and Bathke (1984) developed a filtering mechanism that has been used to screen nonseasonal from seasonal firms to facilitate empirical analysis. They identified a parsimonious autoregressiveintegrated-moving-average (ARIMA) quarterly earnings expectation model for nonseasonal firms, a first-order stationary autoregressive process (i.e., AR1).…”

“…ARIMA models for nonseasonal firms are parsimonious, not requiring seasonal differencing and/or seasonal autoregressive or moving-average parameters, unlike the ARIMA models identified for seasonal firms. Lorek and Bathke (1984) demonstrated that the use of such complex seasonal ARIMA models for nonseasonal firms results in overdifferencing of the stationary time series, parameter redundancy, and is a violation of the principle of parsimony. Simpler quarterly earnings expectation models have proved to be more robust than more complex ones and are consistent with Occam's razor.…”

“…Both Lorek and Bathke (1984) and Brown and Han (2000) used quarterly earnings databases beginning with the early 1960s while Bathke et al (2006) employed data from the 1978 to 1996 time period. We observe that Thomas (1993) has speculated that quarterly earnings time-series properties have gradually changed over time.…”

“…Bernard and Thomas (1990) employed a seasonal RW model as their proxy for the unobservable market expectation of quarterly earnings. Brown and Han (2000), on the other hand, restricted their analysis to firms that exhibited nonseasonal quarterly earnings characteristics and employed the AR1 nonseasonal model identified originally by Lorek and Bathke (1984). Williams (1995) has argued that the choice of expectation model in this setting is crucial since the likelihood of drawing erroneous inferences is influenced by the extent to which the expectation model measures the market's expectation of earnings with error.…”