Skip to main content

Income splitting and anti-avoidance legislation: evidence from the Canadian “kiddie tax”


We examine whether “kiddie tax” legislation in Canada, effective as of 2000, deters income splitting between parents and minor children by taxing at the top marginal rate certain types of non-labor income received by children. OLS estimates based on cross-province and time-series data reveal that the share of dividend income reported by children aged 19 and under declines by 86 % after the introduction of this anti-avoidance rule. The estimates also reveal that the share of capital gains (income not covered by the legislation) reported by minor children increases by 70 % in the post-legislation period, suggesting that parents are switching to an alternative income splitting technique. However, the latter percentage effect is on a small base, and thus, the decrease in dividend income is much larger than the increase in capital gains income. Hence, our analysis suggests that the “kiddie tax” is an effective method to deter income splitting.

This is a preview of subscription content, access via your institution.

Fig. 1
Fig. 2


  1. As Brunson (2011) suggests, little research has been conducted to assess the effectiveness of these rules in any country. Furthermore, the empirical literature on tax avoidance through intra-household income shifting is relatively thin. Maki (1996), Dunsky and Follain (2000), and Gervais and Pandey (2008) focus on the impacts of mortgage deductibility on the portfolio of non-residential assets and liabilities held within a household. Stephens Jr and Ward-Batts (2004) and Schuetze (2006) evaluate the shifting of unearned investment income and employment income, respectively, between spouses.

  2. Examples of relevant province-specific policies include differences in marginal personal income tax rates across provinces that might have an indirect effect on the age distribution of reported taxable income. This is especially relevant for small family owned businesses, which have to pay corporate taxes, personal income taxes, and dividends to family owners, and thus have to calculate the optimal mix in order to minimize overall tax liability.

  3. For example, please see Individual Income Tax Returns Publication 1304 (Complete Report) available through the IRS website.

  4. Although it would be preferable to have available age groups that are completely consistent with the effective age of “kiddie tax” legislation (i.e., 17 and under), we believe that our empirical age groups are acceptable. The percentage of the under-20 population aged 18 and 19 is between 10.0 and 11.9 % from 1997 to 2009 (Statistics Canada, CANSIM tables), and thus, any miscategorization in our provincial analysis is relatively minor and would bias against potential findings. If more income is split with children 17 and under, the inclusion of 18- and 19-year-olds lowers the average income share for the under-20 group and reduces the difference between minor and non-minor children. Conversely, if more income is split with children 18 and over, then the inclusion of 18- and 19-year-olds in the under-20 age group raises the average share. Again, the result is a smaller difference between minor and non-minor children.

  5. For example, a child could invest $1 in a company, provide no labor effort, and yet receive thousands of dollars in dividends. Such strategies were of low cost and could be easily adjusted according to changing financial circumstances of family members. This activity was at issue in both the McClurg and Neuman cases, two of three cases that had primary influence on the enactment of the “kiddie tax” legislation in Canada (i.e., McClurg v. Canada, 1990 CanLII 28 (SCC); Neuman v. M.N.R., 1998 CanLII 826 (SCC); and Canada v. Ferrel, 1999 CanLII 7418 (FCA)—see

  6. The lack of exceptions can alternatively be viewed as over-inclusiveness; undoubtedly, some children have earned these types of income through their own efforts, and not as a result of a transfer of assets from a parent. Thus, in policy terms, a trade-off may exist between effectiveness and equity in anti-avoidance legislation.

  7. For potential strategies to avoid “kiddie tax,” see Weinstein (2008) and Corley (2010) for the USA, and Donnelly et al. (2000) and Ideias (2014) for Canada.

  8. Capital gains are taxed when realized and no legislation exists to prevent a minor child from holding transferred property until they reach the age of 18, when attribution in Canada normally ceases to apply (Donnelly et al. 2000). As a caveat, however, a gift of shares typically results in immediate taxation to the transferor on any unrealized capital gains. Furthermore, capital gains can be volatile and difficult to predict, and could result in losses, likely making it a less preferred activity prior to “kiddie tax” legislation.

  9. Provincial data are not available prior to 1997. We exclude 1997 from the provincial sample because, conceptually, it may not be comparable to the other years in the pre-legislation period. The Supreme Court decision in Neuman , released early in 1998, overturned an earlier lower-court decision (to the surprise of tax practitioners) and thus effectively created the primary market for dividend splitting with minors beginning with that year; supra footnote 5. However, our analysis is robust to the inclusion of 1997 in our provincial sample.

  10. Investment income for purposes of this study refers to the total of bond, bank and mortgage interest, income from trusts, and foreign interest and dividends.

  11. With aggregate data, one cannot determine whether or not the underlying individuals reside with their parents and no links exist between parents and children in the data. Thus, it is impossible to study explicitly income shifting at the parent (i.e., transferor) level.

  12. It is also important to reiterate that our tax data provide information on income reported, not assets held. Furthermore, the “kiddie tax” is triggered based on when income is reported, not when assets are acquired or transferred. Since we report average effects based on shares of income reported, it is irrelevant that the composition of our three age groups implicitly changes throughout the sample period (e.g., some minor children will become non-minor children). What is relevant is the total share of income each respective age group implicitly chooses to report in each year or period of the sample. The income shares are not cumulative, so the amounts reported each year, where applicable, represent a decision in that year about whether or not to shift income.

  13. For example, Slemrod (1996) and Saez (2004) also employ a simple multivariate regression model where the dependent variable is the total taxable income of a group as a share of total taxable income with respect to a given province/state in a specific year.

  14. Population figures are from Statistics Canada, CANSIM database, table 510001, matrix 3604.

  15. We solve a system of two equations and two unknowns. Let \(E\) be actual eligible dividend, and \(N\) be actual non-eligible dividend. Then, Total Taxable Dividend = \(E*1.45 + N*1.25\) and Total Dividend Tax Credit = \(E*1.45*0.1897 + N*1.25*0.1333\). As total taxable dividend and tax credit are known, we solve for \(E\) and \(N\). The resulting estimate represents a lower bound because private corporations can pay eligible dividends (when taxable income exceeds a threshold used to define income eligible for small-business incentives), biasing our estimate of private dividends downward. We thank an anonymous referee for pointing out the need to try to infer the division between public and private dividends.

  16. The amounts of income being split are relatively small for any particular child—averaging less than $11,000 for all years for dividend income, and less than $6,000 for all years for business income (for returns having that type of income)—and thus, administration costs bear heavily on planning choices. It appears that the inclusion of net business income in the “kiddie tax” legislation was a preventative step by the government which merely preserved the status quo of little use of these structures.

  17. In 1988, the capital gains inclusion rate increased from one-half to two-thirds, and in 1990, it increased again to three-fourths. In 2000, it was reduced twice, from three-fourths to two-thirds, and from two-thirds back to one-half.

  18. This specification is validated by likelihood ratio tests based on Box-Cox regressions that do not reject the use of a log specification. Although a power transformation is most suitable, the empirical results and model fit are highly similar with either a log or power transformation. For ease of interpretation, we report the semi-log specification.

  19. Our results might be skewed, given that most observations occur post-legislation. However, this result holds with the addition or subtraction of a few years of data also. We thank an anonymous referee for suggesting this sensitivity test.

  20. Within the Report, age categories 25–34, 35–44, 45–54, and 55–64 are tabulated separately. The results and inferences of our subsequent analyses are unchanged if we estimate individual coefficients for each of these categories rather than a single DAGE36 coefficient.

  21. The use of such difference-in-difference specifications that estimate the effects of a tax amendment on a treatment group relative to a control group that was not directly affected by such policy changes is standard in the literature. For example, as noted above, LaLumia (2008) estimates the effects of 1948 legislation that shifted the USA to a regime of joint taxation of household income, by investigating the effects of such laws on a treatment group (married individuals) relative to a control group (single individuals). Furthermore, our regression coefficients reflect changes in income shares only, since population, a component of the dependent variable, is unaffected by the legislation.

  22. Saskatchewan is the omitted province in our model.

  23. Data on dividends in the Report are recorded as taxable amounts, which reflect a dividend gross-up. We have removed the gross-up in our data to reflect actual dividends paid. Different gross-up rates have been available in Canada over our sample period: 1/3 for 1975–1977 and 1987; 1/2 for 1978–1986, 1/4 for 1998–2005, and both 1/4 and 9/20 for 2006–2009. We use our previously noted decomposition technique for eligible and non-eligible dividends to calculate the actual dividends paid for 2006–2009.

  24. Refer to for more details.

  25. Given the small number of provinces (10), we cluster by province-year and use an accompanying bootstrapping process recommended by Cameron et al. (2008); clustering only by year or province yields consistent results. Results are consistent also when using generalized least squares in order to correct for first-order autocorrelation.

  26. See Kennedy (1981), with respect to interpretation of indicator variables in semi-logarithmic equations. With coefficient \(c\) and variance \(v\), the marginal/percentage impact of any indicator variable coefficient is expressed as 100[\(e^\wedge (c-0.5*v)-1\)]. Term \(v\) is given by the square of the standard error of the coefficient. These adjusted marginal effects are reported in straight brackets for each indicator variable coefficient in Tables 2, 3, and  4.

  27. By comparison, applying a similar adjustment to the sum of coefficients \(\textit{DLEG}_\textit{it}\) and \(\textit{DAGE1}_\textit{ijt}*\textit{DLEG}_\textit{it}\), which is negative and statistically significant at the 1 % level, yields an overall decrease of 73.5 % in minors’ share of dividend income from pre-legislation to post-legislation. Thus, the conclusion that the legislation has a large and statistically significant negative effect on minors’ share of dividend income persists where non-minor children are not used as a control group.

  28. The sum of \(\textit{DLEG}_\textit{it}\) and \(\textit{DAGE36}_\textit{ijt}*\textit{DLEG}_\textit{it}\) (with Kennedy’s adjustment) implies that the adult age group has a 13.5 % increase in its share of dividend income from the pre-legislation period to the post-legislation period.

  29. The sum of coefficients \(\textit{DLEG}_\textit{it}\) and \(\textit{DAGE1}_\textit{ijt}*\textit{DLEG}_\textit{it}\) for capital gains, while positive, is not statistically significant at conventional levels. Thus, an increase in minors’ overall share of capital gains from pre-legislation to post-legislation is not statistically evident.

  30. To the extent that the share of employment income is jointly determined along with our four income shares of interest, its inclusion as a control variable could bias our estimates. If share of employment income is omitted from the model, untabulated results are consistent with the evidence in this study.

  31. We obtain these dollar estimates using the following procedure (taking dividends as an example). First, we estimate the percentage share of income for minor children without legislation by dividing actual average share (from 2000 to 2009) of taxable dividends reported by minor children by one plus the estimated amount that the percentage share of that type of income has changed from pre- to post-legislation, i.e., divide by 1 + (– 0.861) = 0.139. Second, we calculate the difference between minors’ actual average share (from 2000 to 2009) and the previous as if estimate and multiply by the respective average aggregate taxable dividends reported by all age groups. A similar procedure is used for the capital gains estimate.

  32. If we measure these revenue estimates using the total change for minor children (i.e., \(\beta _{1}\) + \(\beta _{3}\), with Kennedy’s adjustment), dividend income decreases by $660 million and capital gains increase by $23 million. Thus, the offset of increased revenue from capital gains splitting would be 3 %.

  33. As an alternative, untabulated test, we re-estimate Eq. (1) without observations from 1999. The coefficient estimates of \(\textit{DLEG}_\textit{it}\), \(\textit{DAGE1}_\textit{ijt}*\textit{DLEG}_\textit{it}\), and \(\textit{DAGE36}_\textit{ijt}*\textit{DLEG}_\textit{it}\) are nearly identical to the respective coefficients reported in Table 4.

  34. The sign and significance of these coefficients remain the same or improve if 2002 and/or 2001 are also excluded.


  • Brunson, S. D. (2011). Grown-up income shifting: Yesterday’s kiddie tax is not enough. University of Kansas Law Review, 59(3), 457.

    Google Scholar 

  • Cameron, C., Gelbach, J. B., & Miller, D. (2008). Bootstrap-based improvements for inference with clustered errors. Review of Economics and Statistics, 90(3), 414.

    Article  Google Scholar 

  • Canada Revenue Agency. (1997–2009). Income statistics-table 4. Ottawa, ON: Canada Revenue Agency.

  • Corley, T. (2010). Kiddie Tax rules and planning strategies. Kiddie Tax Rules and Planning Strategies January 9, 2010. Accessed 2 Nov 2014.

  • Donnelly, M., Magee, J., & Young, A. (2000). Income splitting and the new kiddie tax: major changes for minor children. Canadian Tax Journal, 48(4), 979.

    Google Scholar 

  • Dunsky, R. M., & Follain, J. R. (2000). Tax-induced portfolio reshuffling: The case of the mortgage interest deduction. Real Estate Economics, 28(4), 683.

    Article  Google Scholar 

  • Gervais, M., & Pandey, M. (2008). Who cares about mortgage interest deductibility? Canadian Public Policy, 34(1), 1.

    Article  Google Scholar 

  • Ideias, P. (2014). What a tax planner needs to know about the tax on split income (the “Kiddie Tax”) and the attribution rules. Taxnet Pro. Toronto, ON: Carswell.

  • Kennedy, P. E. (1981). Estimation with correctly interpreted dummy variables in semilogarithmic equations. American Economic Review, 71, 802.

    Google Scholar 

  • Kesselman, J. (2008). Income splitting and joint taxation of couples: What’s fair? IRPP Choices 14, February, Institute for Research on Public Policy, Available at

  • LaLumia, S. (2008). The effects of joint taxation of married couples on labor supply and non-wage income. Journal of Public Economics, 92(7), 1698.

    Article  Google Scholar 

  • Macnaughton, A., & Matthews, T. (1999). Is the income-splitting tax needed? Some empirical evidence. Canadian Tax Journal, 47(5), 1164.

    Google Scholar 

  • Maki, D. M. (1996). Portfolio shuffling and tax reform. National Tax Journal, 49(3), 317.

    Google Scholar 

  • Saez, E. (2004). Reported incomes and marginal tax rates, 1960–2000: Evidence and policy implications. In J. Poterba (Ed.), Tax policy and the economy, Vol. 18, pp. 117–174.

  • Schuetze, H. J. (2006). Income splitting among the self-employed. Canadian Journal of Economics, 39(4), 1195.

  • Slemrod, J. (1996). High-income families and the tax changes of the 1980s: The anatomy of behavioral response. In M. Feldstein & J. Poterba (Eds.), Empirical foundations of household taxation. Chicago: University of Chicago Press and the National Bureau of Economic Research.

    Google Scholar 

  • Stephens, M, Jr, & Ward-Batts, J. (2004). The impact of separate taxation on the intra-household allocation of assets: Evidence from the UK. Journal of Public Economics, 88(9–10), 1989.

  • Weinstein, G. W. (2008). How kiddie tax can still be countered. Financial Times February 16, 2008.

Download references


Bauer acknowledges the Canada Revenue Agency for providing early Income Statistics data. We thank workshop participants at the School of Accounting and Finance, University of Waterloo, and participants at the Graduate Student Research Conference, University of Waterloo for providing helpful comments and suggestions.

Author information

Authors and Affiliations


Corresponding author

Correspondence to Andrew M. Bauer.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Bauer, A.M., Macnaughton, A. & Sen, A. Income splitting and anti-avoidance legislation: evidence from the Canadian “kiddie tax”. Int Tax Public Finance 22, 909–931 (2015).

Download citation

  • Published:

  • Issue Date:

  • DOI:


  • “Kiddie tax”
  • Income shifting
  • Income splitting
  • Anti-avoidance
  • Canada

JEL Classification

  • D13
  • H24
  • H31
  • K34