This effect of the base-ten revenue threshold exists independently of the reporting currency; our results apply to companies reporting in the US. A pervasive numerical cognitive bias can lead to an increase in the visibility of companies that report accounting numbers that are just above the ten threshold. Companies that report financial numbers that meet or just exceed the base-ten threshold may be more likely to attract attention from investors and the financial press.

H1: Companies are more likely to report revenue just above the base ten threshold than just below it, regardless of reporting currency. We also explore some of the motives that drive managers to engage in these behaviors that reach the base-ten threshold. H2a: Analysts are more likely to issue revenue forecasts that are just above the base ten threshold than just below it.

Consequently, the managers of these firms may be more likely to make efforts to report income above a base-ten threshold. H3b: The base-ten revenue threshold effect is greater for those firms with higher past revenue growth. H4: Firms reporting earnings just above a base-ten threshold will experience a greater increase in news coverage compared to a set of control firms.

H5: Firms that report revenues just above the baseline threshold of ten will report lower revenue growth in the following year compared to a control group of firms.

## Sample and Research Design Sample Selection

In summary, our method involves calculating the ratio of the number of observations in an interval (or “bin”) just above the base-ten thresholds to the number of observations in the bin just below it, and then comparing this above/below ratio to the value of the ratios for a set of pseudothresholds generated from empirical distributions using random sampling techniques. If managers believe that reporting earnings greater than or equal to the base-ten threshold is beneficial, then we expect a greater number of observations in bins just above the base-ten threshold than in bins just below these same thresholds. As shown later, considering which bin widths show significant results can give some indication of the amount of flexibility that firms have on average to reach the base ten threshold.

We calculate the test statistic, Sj, by adding the number of observations in the bins just above all thresholds based on ten, the ABOVE bins, and dividing by the sum of the number of observations in the bins just below all thresholds based on ten, the containers BELOW. If managers are trying to increase sales above the base ten thresholds, we would expect the Sj ratio to be greater than 1.6. There is a maximum value of 53 base ten thresholds, but many of the tests do not include all potential base ten thresholds because of the sample size limitations as described in Appendix 2.

Thus, this test allows us to estimate the significance of the base-ten threshold effect by using randomly selected points in the empirical distribution as the benchmark group. This variation in the number of thresholds examined occurs because narrow bins are more likely to include few or no observations at the extreme tails of the distribution of observations.

## Empirical Results

For all base-ten thresholds, the value of S0.0010, the ratio of the number of observations ABOVE to observations BELOW, is 1.55. For example, we see in Panel A that there are 32,475 observations in the ABOVE bins within 2.33% of a base-ten threshold. Data comparing the size of the effect of the base ten threshold on reported income numbers with the corresponding effect on cost of goods sold, total assets, and market value of equity are presented in Table 4.

The ratio-of-ratio analysis shown in Table 4, Panel A demonstrates that for each of the bin widths reported, there is a significantly greater proportion of revenue observations relative to cost of goods sold observations, just above a base-ten threshold; the approximate random p-value reported in the last column is less than 0.01 in each case. Accordingly, we repeat the ratio-of-ratio analysis on international data to demonstrate the robustness of the base-ten income threshold effect and to show that this effect is not unique to the US. We show that the effect is significant for income numbers within one or two percent of a base-ten threshold.

We examine the existence of a base-ten threshold effect in analysts' earnings forecasts in Table 6. The results in Table 6, panels A and B confirm the existence of a base-ten threshold effect in analysts. For both the initial revenue forecast and the final revenue forecast, for each of the four reported bin widths, the approximate randomization test reveals the existence of a highly significant base-ten threshold effect.

For the three narrowest bin widths, the estimated randomization test confirms that the base-ten threshold effect is significantly stronger in the initial revenue forecast compared to the final forecast. The results from Panel D show that the base-ten threshold effect is significantly stronger among the early analysts. Accordingly, H3a and H3b predict that the turnover threshold effect based on ten is stronger for firms with a high price-to-sales ratio and firms with high past sales growth.

In Table 7 , Panels A and B, we report results examining the relationship of the base-ten threshold effect with past revenue growth and price-to-sales ratios, respectively. The matched sample includes firm-years in the same two-digit SIC code and year with revenues within 50% of the base-ten threshold and total assets within 50% of the total assets of the matched firm. Overall, the results reported in Table 8, Panels A and B provide support for the prediction in H4 that reaching a base-ten income threshold results in a relative increase in press coverage.

These results show that some companies are struggling to meet the basic revenue threshold and that this high revenue growth is not sustainable. Overall, the results in Table 9 show that, on average, the revenue growth required to reach the base ten representative threshold of $1 billion is not sustainable.

## Summary and Conclusions

*Starbucks: “Poised To Join $1 Billion-a-Year Club” (September 1997)**Callaway Golf: “Exceeded the billion-dollar mark” (April 2007)**Heinz: Achieves Record Sales of Over $10 Billion (May 2008)**Facebook: First Reported Earnings of $1.000 billion (February 2012)**U.S. Federal Government 2016 Budget Proposal (February 2015)*

For all threshold values of the form T=10K, for integers K (5 to 10). conditions of the common logbook of . gain). This table presents the results of estimated randomization tests that estimate the significance of the threshold effect in reported earnings. Each panel provides statistics for a variety of recorded bin widths, including the number of observations in the bins just above and below the threshold points, the associated binomial test p-value (assuming an equal probability of observations falling into both bins), and the ratio of the observations in the bins above and below (the test statistic used to generate the estimated p-value of the randomization).

Panel B: Incremental observations - All base ten thresholds, all bandwidths For all thresholds of the form T = N x 10K, for integers N (1 through 9) and K (5 through 10). conditions of the common logbook of . gain). This table presents the results of estimated randomization tests that estimate the significance of the threshold effect for reported income relative to cost of goods sold, total assets, and market value of equity. Each panel provides statistics for a variety of recorded bin widths, including the ratio of the number of observations in the bins just above and below the threshold points for sales, cost of goods sold, total assets, and market value of equity, as well as the ratio of the income ratios and the proportions of the three comparison items (the test statistic used to generate the estimated p-value for randomization).

Panel B: Comparison of all base-ten thresholds, income and total assets for all thresholds of the form T = 10K, for integers K (5 to 10). the terms of the joint register of . Income). This table presents the results of random approximation tests that assess the significance of the Threshold Effect for reported revenues relative to that of cost of goods sold for a set of non-US countries. Each panel provides statistics for a variety of recorded bin widths, including the ratio of the number of observations in bins just above and below the threshold points for both revenue and cost of goods sold, and the ratio of these ratios ( the test statistic used to generate the approximate randomness value p-).

Panel A: Analysts' initial revenue forecast compared to actual cost of goods sold For all thresholds of the form T = N x 10K, for integers N (1 through 9) and K (5 through 10). Panel B: Analyst's latest revenue forecast compared to actual cost of goods sold For all thresholds of the form T = N x 10K, for integers N (1 through 9) and K (5 through 10). Panel C: Analyst's first revenue forecast compared to analyst's last revenue forecast For all thresholds of the form T = N x 10K, for integers N (1 through 9) and K (5 through 10).

Panel D: First Analyst revenue forecast compared to actual revenue For all thresholds of the form T = N x 10K, for integers N (1 to 9) and K (5 to 10). This table presents the results of approximate random tests that estimate the significance of the Threshold effect for analyst earnings forecasts. Each panel presents results for the full set of all base-ten thresholds and provides statistics for four recorded bin widths, including the ratio of the number of observations in the bins just above and below the threshold points as well as the ratio of these ratios ( the test statistic which is used to generate the approximate random p-value).

Panel E: Recent analyst forecast earnings compared to actual earnings for all thresholds of the form T = N x 10K, for integers N (1 to 9) and K (5 to 10). This table presents the results of approximate randomness tests that assess the significance of the difference in the Threshold Effect for high and low past revenue growth firms and price-to-sales ratios, where high (low) growth (ratio P/S) is defined as the top (bottom) percentage income growth decile (P/S ratio).