Archived - Tax Expenditures and Evaluations 2014: part 3

Table of Contents

In Canada, all provinces and territories levy corporate income taxes, in addition to the federal government.^[1] The definition of the tax base is broadly consistent across provinces, but provincial corporate income tax (CIT) rates differ from one province to another.^[2]

Inter-jurisdictional differences in taxation levels, such as interprovincial differences in CIT rates, can be viewed as reflecting different preferences across jurisdictions for different levels of public services and public goods. To that extent, fiscal decentralization—the ability of jurisdictions to set their own levels of spending and taxation—can be welfare-enhancing when jurisdictions are tailoring their “menus” of public policies to address their citizens’ needs and preferences, and competition across jurisdictions induces governments to be more efficient in designing and implementing these policies.

On the other hand, differences in tax policies across jurisdictions could also result in inefficiencies where the tax policies implemented in one jurisdiction have negative spillover effects for other jurisdictions. Such effects could be real effects—for instance, differences in taxation levels may distort the location of investment across jurisdictions, with more capital being invested in lower-tax jurisdictions than would be justified on the basis of pre-tax rates of return alone. Tax rate differentials may also give rise to tax planning intended to move income to lower-tax jurisdictions, without a concurrent relocation of income generating activities. These two categories of effects may have negative fiscal implications for jurisdictions, in particular by encouraging excessive tax competition to attract mobile tax bases.

This study examines the extent to which differences in provincial CIT rates may induce corporations to seek tax planning opportunities to shift taxable income from jurisdictions with higher tax rates to those with lower tax rates, without a concurrent shift in real business activities. Corporate-level taxation data are used to estimate the elasticity of taxable income to provincial CIT rates, that is, the extent to which the level of taxable income reported by a corporation in a given province is affected by that province’s CIT rate, taking into account other possible determinants of taxable income. This empirical analysis is performed to provide insights on the importance of tax planning transactions that are motivated by CIT rate differentials across Canadian provinces. The analysis focuses on groups of commonly-controlled corporations, and compares the tax planning behaviour of corporate groups that have the flexibility to engage in interprovincial tax planning to those that do not (for example, because these corporations only have operations in one province).

This paper is organized as follows. The next section reviews provincial CIT rates over time and provides information on the ways in which corporations may undertake tax planning activities to take advantage of tax rate differentials. The paper then looks for general evidence of tax planning behaviour by examining changes in the distribution of taxable income across provinces over time. This discussion is followed by the presentation of an econometric analysis of the sensitivity of taxable income to provincial CIT rates for groups of commonly-controlled corporations. The final section of the paper summarizes the results and presents the conclusions of the analysis. A technical annex provides further information about the econometric model and discusses the results in more detail.

Chart 1 shows trends in provincial statutory CIT rates from 2000 to 2014. It shows separately the CIT rates for the four largest provinces (Quebec, Ontario, Alberta and British Columbia, which together account for about 90% of taxable income in Canada), in addition to the unweighted average of the CIT rates in the other provinces.^[3]

In 2000, the CIT rate differentials among provinces were substantial, with CIT rates ranging from 9% in Quebec—the lowest-tax-rate jurisdiction in Canada that year—to 17% in Manitoba, New Brunswick and Saskatchewan.

Over the last decade, the Quebec government increased its CIT rate, while most other jurisdictions decreased their rates. This led to a convergence of CIT rates, in particular among the four largest provinces. The gap between the highest and lowest CIT rates for these provinces declined substantially, from 7.5 percentage points in 2000 to 1.9 percentage points since 2011. On average, the four largest provinces implemented larger declines in their CIT rates than other provinces over the 2000 to 2014 period.

Chart 2 compares the trends in corporate taxable income in each of the four largest provinces from 2000 to 2012, as well as that of other provinces. It shows that taxable income increased in all provinces over the last decade, although at varying paces. Overall, growth in taxable income in Alberta was relatively strong, while growth in Quebec and Ontario was below average.

The impact of economic cycles on taxable income is to some extent smoothed by the ability of corporations to carry over unused losses to past and future taxation years.^[4] For instance, one could expect that a marked reduction in taxable income would be observed in 2009 because of the recent recession. However, losses incurred in 2009 were used in part to reduce taxable income reported in years 2006 to 2008, and in part to reduce taxable income reported after 2009, therefore diminishing the observable impact of the recession on taxable income.

A business conducting income-earning activities in more than one province may do so under two broad corporate structures: it may operate as a single corporation with permanent establishments in more than one province (“multi-jurisdictional corporation”), or it may operate via a group of commonly-controlled corporations, each having a permanent establishment in one or more province (“corporate groups”).

Multi-jurisdictional corporations in Canada must allocate taxable income to each province where the corporation has a permanent establishment using a general allocation formula or, for corporations in certain specific sectors, using a formula that applies exclusively to those sectors. The general formula allocates income based on the proportions of salaries and wages paid and gross revenue earned in each province where the corporation has a permanent establishment.^[5] Under this regime, a multi-jurisdictional corporation would allocate a greater share of its taxable income to a province if either the share of salaries and wages paid to employees that report to permanent establishments located in that province increases, the share of gross revenue attributable to that province increases, or a corporation that previously had no permanent establishment in that province decides to establish one. Therefore, corporations subject to the allocation formula have limited flexibility to undertake interprovincial tax planning without a concurrent shift in business activity.

Corporations that are members of a corporate group are taxed on a stand-alone basis. Although there is no formal rule for transferring income, losses or other tax attributes among members of a corporate group, corporations could use related-party transactions and various provisions of the Income Tax Act to shift income among group members. For example, a loan from one group member to another would reduce the taxable income of the borrowing corporation, as interest payments on this loan would represent a tax-deductible expense, and would increase the taxable income of the lender, as the interest revenue would be included in the taxable income of that corporation. Similarly, transfers of property with an accrued gain between group members on a tax-deferred basis may result in the gain being realized on the final disposition of the property by another group member. These transactions may be undertaken for genuine business reasons and the end result on income allocation may well be reflective of where the substantive activities take place, but when such transactions coincide with an increase in taxable income allocated to low-tax rate jurisdictions and a decrease in taxable income allocated to high-tax rate jurisdictions, it remains that the corporate group as a whole would pay a lower amount of provincial taxes than it would in the absence of these transactions.

Interprovincial CIT rate differentials are not the only reason that corporate groups may engage in tax-motivated
intra-group transactions. As noted, under the Income Tax Act, a corporation that incurs a net loss in a given taxation year can apply the loss against income of a previous or future taxation year; however, the loss cannot directly be transferred to another corporation and applied against the income of that other corporation (except in some limited circumstances, for instance when the corporation is being acquired and the acquirer continues the same business). This implies that the loss would be of limited value to a corporate group if the corporation that incurred the loss has no or little taxable income in past and future taxation years and cannot apply the loss during the carry-over period. As such, corporate groups have an incentive to enter into related-party transactions that will effectively enable the use of losses accumulated by one member of the group to offset income earned by other members of the group. Corporate groups are often able to use the existing flexibility in the tax system to transfer income or losses between related corporations through financing arrangements, reorganizations (such as amalgamations or wind-ups), and transfers of property on a tax-deferred basis, in order to access these losses. Such transactions affect the distribution of taxable income across provinces if the corporations that incurred the losses and the corporations that apply the losses to reduce their taxable income face different provincial income allocations.

Direct measurements of the magnitude and directions of interprovincial tax planning transactions are not currently available. In principle, tax planning could be measured in one of two ways: either by identifying and measuring the impact of the types of transactions used to shift income across provinces, or by comparing the allocation of taxable income among provinces to a theoretical benchmark that would represent how income should be allocated between provinces in the absence of tax planning.

Both of these approaches face significant challenges. Measuring the impact of related-party transactions is often not possible due to a lack of data, and it can be difficult to distinguish between transactions that are motivated by tax planning considerations from those related to genuine business motivations. As for the second approach, determining how income would be allocated in the absence of tax planning would also raise conceptual issues, and would require making assumptions about the appropriate way to allocate profits, revenues and expenses that are common to members of a group.

Nevertheless, in the absence of a satisfactory benchmark or direct measurements of related-party transactions, it would be expected that if corporations were engaging significantly in interprovincial tax planning activities, a disconnect would be observed between the relative amount of taxable income reported in a given jurisdiction and the relative amount of economic activity in that jurisdiction. For example, interprovincial tax planning could result in the share of taxable income attributed to a low-tax rate province being higher than its share of overall economic activity.

For each of the four largest provinces, Chart 3 compares the gap between the province’s share of total taxable income and its share of nominal gross domestic product (GDP, a standard measure of economic activity) against the gap between the province’s CIT rate and the unweighted average CIT rate for all provinces.^[6] Interprovincial tax planning would be suspected if a reduction in a province’s CIT rate compared to the average provincial CIT rate coincides with an increase in its share of taxable income, without a concurrent increase in its share of GDP.

Chart 3 shows that overall, discrepancies between taxable income shares and GDP shares seem to be negatively correlated with provincial tax rate differentials:

• Quebec’s CIT rate increased after 2005, which led to a marked narrowing in the tax rate gap between Quebec and other provinces. Quebec’s share of taxable income was higher than its share of nominal GDP in the early 2000s, when Quebec still had the lowest CIT rate among provinces. Since 2006, however, its taxable income share has moved below its GDP share.
• In the early 2000s, when Ontario’s CIT rate was lower than the average for all other provinces, its taxable income share was generally higher than its share of GDP. Once Ontario’s CIT rate moved above the national average, its share of taxable income moved below that of GDP.
• Coinciding with the widening of the gap between Alberta’s CIT rate and the average for all other provinces, Alberta’s taxable income share increased rapidly from 2003 to 2009 and has remained above its share of GDP.
• British Columbia’s CIT rate was higher than the average for all other provinces in 2000 and 2001, and subsequently has been lower than the average with a gradually increasing gap. Coinciding with the widening of the CIT rate gap, British Columbia’s taxable income share increased relative to that of GDP up until 2007, and has since remained slightly below the GDP share.

While the trends observed in Chart 3 provide general indications, it is difficult to draw firm conclusions from this analysis, as there are a number of other factors that may be at play and that may explain the apparent correlation between CIT rates and taxable income shares. As a measure of overall economic activity, GDP reflects not only trends in corporate activity, but also the economic activity of the unincorporated sector, individuals and governments. As such, GDP trends may differ significantly from trends in corporate taxable income for reasons that are not attributable to tax planning. A gap between a province’s shares of taxable income and GDP may also be attributable to factors other than interprovincial tax planning—for instance, a gap could arise if taxable income and GDP are affected differently by provincial economic shocks (which would be expected since GDP trends do not account for loss carry-overs).

Given the difficulties associated with measuring interprovincial tax planning directly, this study relies on econometric analysis to produce indirect measures of the prevalence of interprovincial tax planning. While numerous studies have used econometric methods to measure tax planning at the international level, very few have focused on interprovincial tax planning in Canada (often referred to as “income shifting” in the literature). Two examples of studies on interprovincial tax planning in Canada include those by Lachance and Plante (1994), and Mintz and Smart (2004). These studies both found statistical evidence of interprovincial tax planning in Canada.

Lachance and Plante (1994) used provincial-level data to estimate the impact of CIT rate reductions in Quebec during the early 1980s on corporate tax revenues from the manufacturing sector in that province. Similar to the concept introduced in the previous section, their empirical analysis was based on the notion that, without income shifting activities, the share of taxable income reported in the manufacturing sector in Quebec should reflect the share of manufacturing production in the province.^[7] Their results confirm that there is a positive relationship between taxable income and production activity, and that reductions in Quebec’s CIT rate had a positive impact on Quebec’s share of taxable income. Overall, they concluded that the reductions of Quebec’s CIT rate resulted in a transfer of taxable income into the province without an equivalent increase in production.

Mintz and Smart (2004) studied income shifting across all Canadian provinces. Using corporate income tax records that were aggregated into sub-categories based on province, industry, year, and type of firm defined based on the ability of a firm to engage in income shifting (e.g., whether a firm paid taxes in only one province or not, whether a firm was a subsidiary), they compared the elasticity of taxable income with respect to CIT rates for firms with income shifting opportunities to that of firms without shifting opportunities. Their estimate of the elasticity of taxable income is higher for categories of firms with shifting opportunities compared to categories without shifting opportunities.^[8] As such, the authors concluded that “evidence suggests that income shifting has pronounced effects on provincial tax bases”.

These two studies faced limitations that are related to the use of aggregated data. In the case of Lachance and Plante (1994), their measure of taxable income reflected the total for all corporations in Quebec, regardless of differences such as the extent to which each corporation could engage in interprovincial tax planning activities. With more detailed data on taxable income, Mintz and Smart (2004) made progress in narrowing in on those corporations with the potential to shift income. However, their data restrictions limited their ability to clearly distinguish between corporations that have the opportunity to engage in interprovincial tax planning and those without. In particular, they could not identify members of corporate groups and how they are related.

In contrast, this study uses corporation-level tax data to examine the determinants of taxable income within groups of commonly-controlled corporations. Similar to Mintz and Smart (2004), the approach compares the elasticity of taxable income to provincial CIT rates for corporations that are expected to have interprovincial tax planning opportunities (referred to as “potential shifters”) to that of corporations without such opportunities (“non-shifters”). Using corporation-level data for groups of commonly-controlled corporations as opposed to aggregated data has two significant advantages. First, by observing the provincial CIT rates faced by each member of a group, it is possible to clearly identify corporations with an incentive to conduct interprovincial tax planning. Second, access to financial and taxation information for each corporation makes it easier to control for heterogeneity across corporations. Corporations differ not only in their ability to use tax planning strategies (interprovincial or other) to reduce taxable income, but also in their ability to generate taxable income in the first place. The greater the ability to control for this heterogeneity, the greater the ability to isolate the impact of provincial CIT rates on reported taxable income from that of other factors related to differences between corporations.

The elasticity of taxable income with respect to provincial CIT rates for corporations with and without interprovincial tax planning opportunities is estimated using regression analysis. This approach isolates the impact of provincial CIT rates from that of other factors that influence taxable income. In particular, the specification tests a dimension that has not been previously explored, which is whether the availability of unused losses within a same corporate group has an impact on the elasticity of taxable income. The presumption is that when corporate groups can reduce taxable income through a more tax-efficient use of losses, provincial CIT rates may play a secondary role in tax planning decisions, because groups would transfer income from profitable corporations to loss-making corporations, and these corporations may not necessarily be located in high-tax provinces and low-tax provinces respectively. As such, it is assumed that the elasticity of taxable income with respect to CIT rates would be lower for corporations that are part of groups that have accumulated unused losses.^[9]

The rest of this section explains how corporations with and without interprovincial tax planning opportunities were identified, summarizes the key differences between these two groups of corporations, provides details on the econometric approach adopted in this study, and presents the results.

All of the corporations included in the regression analysis belong to groups of commonly-controlled corporations. These groups are identified based on shareholder information reported for tax purposes, as well as inter-corporate ownership information collected by Statistics Canada under the Corporations Returns Act.^[10] This information is used to link corporations together according to the degree of joint ownership between corporations. To be included in the regression analysis, at least 50% of a corporation’s share capital must be held together by other members of a corporate group. Corporate groups are identified for the years 2005 to 2012, but the data does not allow individual groups to be analyzed over time. As such, this analysis makes use of pooled cross-section data for the years 2005 to 2012.

As noted, the dataset is divided into “potential shifters” and “non-shifters”. These two groups are defined as follows:

• Corporations are identified as potential shifters when corporate group members are present in at least two provinces. Corporations are also classified as potential shifters when all members of a group operate in the same province but face different CIT rates, for instance a province’s general CIT rate and manufacturing CIT rate.^[11] Multi-jurisdictional corporations, which are single corporate entities with permanent establishments in more than one province, are generally included in the analysis when they are part of a corporate group.^[12]
• When all the members of a corporate group operate in the same province and face the same provincial CIT rate, they are classified as non-shifters, meaning that they have no flexibility to transfer income to a member in another province that has a different CIT rate.

Groups of commonly-controlled corporations identified using tax and administrative data account for nearly 68% of corporate taxable income reported in Canada between 2005 and 2012. Certain corporations or corporate groups are excluded from the analysis, notably groups with income taxed at the federal small business tax rate, groups in which all members have zero net income, and corporations with no evidence of business activity.^[13] After these exclusions, the subset of corporate income tax files retained for this analysis accounts for 59% of corporate taxable income reported between 2005 and 2012.

This section provides information about the characteristics of corporations identified as potential shifters and those identified as non-shifters. The sample used for the econometric analysis includes roughly 236,000 potential shifters and 335,000 non-shifters (these numbers represent the total for the 2005–2012 period).^[14] Several differences can be observed between these two groups of corporations (Table 1).

Potential shifters account for roughly 42,000 corporate groups while non-shifters account for about 112,000 corporate groups. Potential shifters tend to belong to larger groups than is the case for non-shifters. Groups of potential shifters have 5.6 members on average, compared to 3.0 members for groups of non-shifters. Potential shifters also tend to be larger than non-shifters in terms of average capital stock ($18.8 million versus $2.2 million) and average wage bill ($8.2 million versus $0.9 million).

About 50% of potential shifters have a “foreign connection”, that is own foreign affiliates or reported non-arm’s length transactions with non-residents. This compares to only 7% for non-shifters. Potential shifters are also more likely to carry a balance of unused investment tax credits (12% versus 7%).

Roughly 66% of potential shifters are Canadian-controlled private corporations (CCPCs), compared to 95% of non-shifters.^[15] The taxable income of potential shifters is concentrated in the finance and insurance (29%), manufacturing (17%), and mining and oil and gas extraction (12%) sectors, while that of non-shifters is concentrated in the construction (18%), finance and insurance (13%) and manufacturing (11%) sectors.

Table 1 also compares the net income and loss profiles of potential shifters and non-shifters. Although on average potential shifters reported higher net income for tax purposes than non-shifters ($8.4 million on average versus $0.6 million), they were also more likely to report a current-year non-capital loss.^[16] Roughly 29% of potential shifters reported current-year non-capital losses, compared to 23% of non-shifters.

Potential shifters also have larger beginning-of-year pools of unused non-capital losses than non-shifters. However, expressed as a share of net income, the unused losses of non-shifters at the beginning of year were slightly more important than that of potential shifters (29% versus 25%). While potential shifters are slightly more likely to have unused beginning-of-year losses (43% compared to 40%), they are much more likely to be part of a group where at least one member has unused losses (72% compared to 50%). Potential shifters with group loss pools belong to larger groups on average: 7.2 members compared to only 3.6 members for potential shifters without group losses (not shown in Table 1). Finally, the average taxable income reported by potential shifters is roughly $4 million compared to $0.4 million for non-shifters.

Table 2 presents the relative importance of potential shifters and non-shifters as groups of taxpayers. The taxable income of potential shifters that are part of groups that do not have beginning-of-year unused non-capital losses represented 23% of the corporate income tax base in Canada over the period, while that of potential shifters in groups with beginning-of-year unused losses represented 29%. Non-shifters accounted for a much smaller share of the tax base (5% for non-shifters in groups with no unused losses, and 2% for non-shifters in groups with unused losses). The remainder of the tax base is composed of corporations that are not part of corporate groups and corporations excluded from the sample for reasons explained in Annex 2.

The elasticity of taxable income with respect to provincial CIT rates is estimated using regression analysis that controls for other factors that can influence taxable income, including corporation-specific factors such as the ability to use losses and investment tax credits, the stock of capital, wages and salaries paid, foreign connections, CCPC status, and industry.^[17] The analysis also controls for year-specific factors that are intended to capture shocks that are common to all provinces over time.

The model compares the elasticity of potential shifters to that of non-shifters and measures the extent to which the ability to access pools of unused losses affects the elasticity of taxable income to provincial CIT rates. A corporation is considered to have the ability to access pools of unused losses so long as at least one member of the group of which the corporation is a member carries pools of unused losses at the beginning of the year.

The estimation sample used in this analysis poses certain econometric challenges because taxable income cannot be negative and a large number of corporations report zero taxable income in any given year. To circumvent these challenges, the impact of provincial CIT rates on taxable income is summarized by estimating first the degree to which provincial CIT rates may affect the likelihood that a corporation reports positive taxable income, and then the elasticity of taxable income with respect to provincial CIT rates for those corporations that report positive taxable income. Estimates for these two steps are consistent; for ease of exposition, Table 3 presents the estimates for the second step only, while complete results can be found in Annex 3.

The elasticity of taxable income to provincial CIT rates for non-shifters with positive taxable income is estimated to be -0.2. This implies that a 1% increase in the provincial CIT rate (for example, an increase from 10% to 10.1%) was associated over the sample period with a 0.2% decline in taxable income, after controlling for other factors that affect taxable income. The results also suggest that access to a pool of unused losses does not affect the elasticity of taxable income to provincial CIT rates for non-shifters.

Results for potential shifters with positive income indicate that when a corporation does not have access to a loss pool, a 1% increase in the provincial CIT rate was associated with a 1.1% decline in taxable income over the sample period. This response is notably higher than that of non-shifters. This is consistent with the expectation that taxable income is more sensitive to provincial CIT rates for members of corporate groups that have activities in multiple provinces and that may have the flexibility to enter into related-party transactions to transfer profits across provinces.^[18]

Access to a pool of unused losses reduces the elasticity of taxable income for potential shifters. In this case, the elasticity of taxable income declines from -1.1 to -0.4, a level that is in line with the elasticity observed for non-shifters. This implies that potential shifters with members carrying losses (which account for roughly 70% of potential shifters in the sample) seem to be significantly less sensitive to provincial CIT rates than members of groups without unused losses. This is consistent with the expectation (described at the beginning of this section) that provincial CIT rates may play a secondary role in tax planning decisions when groups have unused losses located in particular provinces that can be used to offset income earned in other provinces.

This paper uses econometric techniques to estimate the elasticity of taxable income to provincial CIT rates. It does so separately for corporations with the potential to shift income between provinces and corporations without interprovincial shifting opportunities.

The results suggest that provincial CIT rates have a negative impact on taxable income—that is, an increase in a province’s CIT rate is generally associated with a decrease in the amount of taxable income reported in that province. This negative relationship is not only driven by interprovincial tax planning, as non-shifters (i.e., corporations without the possibility to engage in interprovincial tax planning) also display a negative elasticity, albeit of a smaller size.

The results suggest that caution may be warranted before concluding that income shifting has an important effect on provincial tax bases. Corporations that display the highest elasticity to corporate income tax rates (potential shifters without access to pools of unused losses) represent only about 22% of the tax base. It is difficult to conclude that interprovincial CIT rate differences for the remainder of corporations affects taxable income significantly, as other potential shifters do not display an elasticity that is very different from that of non-shifters, which have no ability to conduct provincial CIT rates arbitrage. The lower elasticity found for the group of potential shifters with loss pools could be interpreted as implying that the location of losses, rather than (or in addition to) the CIT rate itself, is an important consideration in the direction of interprovincial tax planning transactions. Together, these results suggest that it would be unlikely that any single province is systematically affected, either positively or negatively, by the interprovincial tax planning of corporate groups—for example, when losses are available within a corporate group, the location of losses is likely to be a factor in the direction of tax planning transactions and one that would vary along economic cycles.

Although this analysis addresses limitations from previous studies through the use of micro-level data of corporate groups in Canada, other limitations remain. The model that has been estimated includes many variables to control for differences existing between potential shifters and non-shifters, yet unobserved differences between the two groups may still exist, and these differences could potentially account for some of the observed differences in the behaviours of these two groups. Furthermore, whether a corporate group is classified as potential shifter or non-shifter could be endogenous, as groups that are more responsive to provincial CIT rates may for that reason be more likely to operate in multiple provinces. This study could not control perfectly for such selection biases, and may thus overstate the differences between potential shifters and non-shifters. Some caution in generalizing the results must also be taken. In particular, because the analysis is based strictly on corporations that operate in groups of commonly-controlled corporations, the results are not necessarily representative of all corporations operating in Canada. In addition, as the analysis focused on corporations that are members of corporate groups, and that as such have greater flexibility to undertake tax planning transactions in order to minimize total tax liabilities, the elasticity estimates produced in this analysis are likely higher than they would be for the average of all corporations operating in Canada.

The following corporations and corporate groups are excluded from the sample used for the regression analysis presented in this paper:

• Corporate groups with income from an active business carried on in Canada falling below the business limit for the small business deduction are excluded.
• Corporate groups in which all corporations have negative net income are excluded. As a group, these corporations are not likely to engage in tax planning.
• Groups in which corporations face the same tax rate but operate in different provinces are excluded, because it would not be possible to classify appropriately these groups as either potential shifters or non-shifters. This is so because even when facing the same provincial income tax rate, the incentive to undertake interprovincial tax planning may exist if a rate differential existed in a previous year. This cannot be ascertained as it is not possible to analyze corporate groups over time.
• Corporations with no sign of business activity (those with zero net income, zero taxable income, zero capital, and zero salary and wages payments) are removed from the sample because they are not likely to participate in the interprovincial tax planning activities of the group.
• Tax-exempt entities and Crown corporations are not subject to provincial corporate income tax and are therefore excluded.

A log-linear equation in which taxable income is expressed as a function of the provincial CIT rate and other explanatory variables is used to estimate the elasticity of taxable income. The estimated coefficient for the provincial CIT rate relates to the elasticity of taxable income with respect to provincial CIT rates. This term is expected to be negative.

The potential impact of the ability to use losses on the elasticity of taxable income is accounted for by including an interaction term between the provincial CIT rate and a dummy variable that indicates the presence of a beginning-of-year non-capital loss pool within the group. For this purpose, a group is not considered to have a loss pool unless the sum of group losses accounts for at least 1% of the group net income. The dummy variable is equal to one when a corporation belongs to a group with a loss pool and zero otherwise. The estimated coefficient for this interaction term can be regarded as an adjustment to the elasticity of taxable income when a corporation has access to a group loss pool. Under the assumption that corporations operating in corporate groups with pooled losses are less sensitive to provincial CIT rates, this term is expected to be positive (i.e., it would bring the elasticity closer to zero). The actual elasticity of taxable income for corporations with a group loss pool is the sum of the coefficient on the provincial CIT rate and the interaction term. For corporations without access to a group loss pool, the estimated elasticity of taxable income is captured by the coefficient on the provincial CIT rate. Note that the interaction term between the provincial CIT rate and the group loss dummy is generally insignificant for non-shifters. As such, this term is not accounted for in Table 3 of the text.

Since this study is interested in the sensitivity of taxable income to provincial CIT rates at the time when income transfer decisions are made, loss carry-backs and their subsequent impact on taxable income are deliberately excluded from taxable income. In other words, the measure of taxable income employed in the regression analysis does not reflect losses that are applied to taxable income in prior years.

In addition to controlling for factors such as capital and wages, which measure the productive capacity of a corporation, the regressions control for overall macroeconomic conditions (growth in provincial real GDP and the GDP price index), and corporation-specific factors such as foreign connections, access to accumulated provincial investment tax credits (ITCs), CCPC status, and industry. Finally, the regressions also control for year effects, which are intended to capture shocks that are common to all provinces in a particular year. All monetary values (such as taxable income, capital, and provincial GDP) are expressed in current dollars.

Beyond tax rate differences across provinces, the availability of provincial ITCs varies across jurisdictions. Corporations may change their tax planning behaviour when provincial taxes payable can be offset by provincial ITCs. In order to control for this potential impact, the model includes a variable that reflects the size of total provincial corporate ITCs relative to total corporate taxable income in the province where a corporation is located. This information is not available at the corporation level, as the database does not include provincial CIT information for provinces not part of a Tax Collection Agreement.

The estimation sample used in this analysis poses certain econometric challenges. First, a large number of corporations report zero taxable income in any given year. Second, taxable income is censored at zero (such that negative values are never observed), which implies that the relationship between taxable income and its explanatory factors is not linear. Consider for example the impact of GDP growth on taxable income. Setting aside the role of loss carry-backs, strong GDP growth should on average lead to an increase in taxable income while negative GDP growth could potentially result in losses. However, these losses are only observed as zero values (rather than negative values). As such, the impact of strong GDP growth will not have the same relative impact on taxable income as that of negative GDP growth, making the relationship non-linear.^[19]

The implication is that ordinary least squares (OLS) estimation, which assumes a linear relationship between taxable income and the explanatory variables, may not yield reliable coefficient estimates. To address this issue, a probit regression is used to estimate the impact of provincial CIT rates on the probability (or likelihood) that a corporation reports positive taxable income while an OLS regression is used on the portion of the sample with positive taxable income. Table A3 provides detailed estimation results. Estimates listed in the left panel of the table (for potential shifters and non-shifters) are OLS coefficient estimates that include corporations with zero taxable income as well as those with positive income. The middle panel reports coefficient estimates from a probit regression where the dependent variable is a dummy variable equal to one for corporations with positive taxable income and zero otherwise. These estimates indicate the direction in which a given explanatory variable affects taxable income. The final panel presents OLS estimates for the portion of the sample with positive taxable income. These estimates are presented in Table 3 of the paper and can be interpreted as taxable income elasticity estimates, conditional on having positive income.

Although the specifications employed in this analysis have tried to control for the key factors underlying taxable income (e.g., capital, wages, economic conditions), it is possible that other factors not accounted for are also at play. In particular, the possibility arises that provincial CIT rates may themselves be endogenous to factors that determine taxable income. For instance, to the extent that the decline in the CIT rate in Alberta was coincident with exceptionally strong growth in that economy, the model could be overestimating the impact of provincial CIT rates. However, results obtained when Alberta is removed from the estimation sample do not suggest that this is the case.

With increased convergence in provincial CIT rates, it could be expected that the benefit from interprovincial tax planning would be reduced, in particular if fixed costs must be incurred to engage in interprovincial tax planning. To test whether the importance of interprovincial tax planning has decreased in recent years, the above analysis was repeated over two separate periods, 2005–2009 (a period during which provincial CIT rates converged) and 2010–2012 (a period when provincial CIT rate differentials were at a minimum). Table A4 presents the results from this additional analysis.

The results shown in Table A4 suggest that as expected, potential shifters have been somewhat less responsive to provincial CIT rates since 2010 than during the earlier period. Interprovincial CIT rate differentials should not be driving the tax planning of non-shifters, and as such little difference would be expected in the estimates for non-shifters for the two periods. The results seem to indicate some changes in the impact of provincial CIT rates on non-shifters between these periods, more specifically a reduction in the impact of provincial CIT rates on the probability of reporting positive taxable income, but an increase in the elasticity of taxable income to provincial CIT rates for those corporations reporting positive taxable income that do not have access to unused losses. These results are not in line with expectations, and may reflect post-crisis dynamics that are not properly accounted for in the model.

Taxable income: Taxable income for tax purposes, as calculated on the corporate income tax return, before revisions for loss carry-backs. Taxable income is non-negative. To include observations with zero taxable income in the full-sample OLS estimates, the log of taxable income is calculated as log(taxable income + 1).

ProvCIT: Provincial CIT rate. For corporations that are not in the manufacturing industry, the general statutory CIT rate is used. For corporations in the manufacturing industry, the preferential rate for manufacturing and processing income is used if there is one in place. For multi-jurisdictional corporations, a weighted average rate is calculated using taxable income allocated to each jurisdiction as weights.

Group loss dummy: Equal to one if at least one member of the corporate group has unused non-capital losses at the beginning of the year that are no less than 1% of the group’s net income, and zero otherwise.

Wages: Employee wages and salaries paid by the corporation, taken mainly from T4 reports and supplemented with information from tax returns for multi-jurisdictional corporations. The log of wages and salaries is calculated as log(wage bill + 1).

Capital: Stock of the corporation’s tangible capital (such as buildings and machinery) and intangible capital (such as patents and research and development). This information is extracted from the General Index of Financial Information, and supplemented with information on end-of-year undepreciated capital costs as reported on schedule 8 of the T2 corporation tax return. The log of capital is calculated as log(capital + 1).

Ownership concentration: The percentage of the corporation’s share capital that is owned by other members of the corporate group.

GDP growth: GDP growth by province in real dollars.

GDP price index: GDP price index by province.

Provincial tax credits: Total provincial corporate investment tax credits, as a percentage of total corporate taxable income in the province where a corporation is located.

ITCs: Stock of unused provincial investment tax credits. The log of ITCs is calculated as log(ITCs + 1).

Non-CCPC: Equal to one if the corporation is not a Canadian-controlled private corporation, and zero otherwise.

Foreign connection: Equal to one if any corporation of a group has a foreign affiliate or has non-arm’s length transactions with non-residents, and zero otherwise.

Industry: Industry of operation as reported on the corporate income tax return.

Lachance, Renaud and Lucie Plante (1994). “Fiscalité et mobilité interprovinciale des bénéfices : Qui gagne perd”, Canadian Tax Journal, vol. 42(3), pp. 843-901.

Mintz, Jack and Michael Smart (2004). “Income Shifting, Investment, and Tax Competition: Theory and Evidence From Provincial Taxation in Canada”, Journal of Public Economics, vol. 88, pp. 1149-1168.

¹ From this point forward, references to “provinces” also include the territories.

² The federal government has Tax Collection Agreements for corporate income tax with all provinces except Quebec and Alberta. The basic foundation of the Tax Collection Agreements is that the federal government agrees to collect and administer provincial taxes, in exchange for which the provinces agree to a common tax base.

³ Table A1 in Annex 1 provides the CIT rates by province over the 2000 to 2014 period.

⁴ The federal tax system allows capital and non-capital losses to be carried over to other taxation years. Unused capital and non-capital losses can be carried back to the previous three taxation years. Non-capital losses can be carried forward for up to 20 years, while capital losses can be carried forward indefinitely. The carry-forward period for non-capital losses was seven years before March 23, 2004 and was 10 years after March 22, 2004 and before 2006.

⁵ The allocation formulas are defined under the Income Tax Regulations and apply to income earned by corporations located in provinces that have signed a Tax Collection Agreement. The corporate income allocation rules of the provinces that have not signed a Tax Collection Agreement for corporate income tax (Quebec and Alberta) are generally harmonized with the federal rules. The definition of permanent establishment in the context of the corporate income allocation between provinces draws significantly from that used in international tax treaties, which refers to a fixed place of business of a corporation and generally implies the performance of at least minimal economic activities. However, the definition used for allocation purposes between provinces is broader than that of tax treaties, making it easier for provinces to establish a nexus vis-à-vis a particular corporation. For example, unlike the tax treaty definition, a corporation owning land in a province would be deemed to have a permanent establishment even where that corporation does not otherwise have a fixed place of business in that province.

⁶ Table A2 in Annex 1 provides information about the shares of taxable income reported by corporate groups, and compares these shares to shares of GDP by province in 2005 and 2012.

⁷ The share of taxable income refers to Quebec’s share of total taxable income reported by the manufacturing sector in Canada. Analogously, the share of production refers to Quebec’s share of Canadian GDP in the manufacturing sector.

⁸ The CIT rate used in this study is specified as the net-of-tax rate, which is equal to one minus the combined federal and provincial statutory CIT rate. More specifically, the estimation results suggest that a 1% increase in the net-of-tax rate is expected to result in a 4.9% increase in taxable income for “potential shifters” compared to only 2.3% for “non-shifters”.

⁹ It is not expected that groups with losses have a completely inelastic response to tax rates: for example, when more than one member of a group are profitable, corporations may first seek to use the losses against the income of the member located in the province with the highest tax rate.

¹⁰ All corporations that are part of a group of commonly-controlled corporations with combined assets exceeding $200 million, revenue exceeding $80 million, or debt obligations or equity owing to non-residents exceeding a net book value of $1 million are required to report their financial and ownership information to Statistics Canada on an annual basis. Prior to 2007, these thresholds were $15 million, $10 million, and $200,000 respectively. For more information see the Corporations Returns Act.

¹¹ Income from manufacturing and processing activities is subject to a lower CIT rate in the Yukon, Saskatchewan, Ontario and Newfoundland and Labrador.

¹² The provincial CIT rate faced by multi-jurisdictional corporations is calculated as the average of the CIT rates of the provinces in which the corporation has permanent establishments, and is weighted according to the corporation’s allocation of taxable income across provinces.

¹³ See Annex 2 for a detailed explanation of these exclusions.

¹⁴ Since data for the years 2005 to 2012 are pooled together into one sample, corporations that operated in more than one year (or over the whole period) will appear in the data several times.

¹⁵ A CCPC is a corporation that resides in Canada and is not public or controlled by public corporations or non-residents.

¹⁶ Net income for tax purposes differs from taxable income when certain additional tax deductions are claimed, chiefly the deductions for charitable donations, taxable dividends and loss carry-forwards.

¹⁷ See Annex 3 for a complete list of variables included in the regression analysis. Given that the purpose of this paper is to detect interprovincial tax planning, the CIT rate variable is based on the provincial rate only, instead of the combined federal and provincial CIT rate. Interpreting the elasticity of taxable income to provincial CIT rates requires caution, and the results presented in this paper are not directly comparable to the results presented in Mintz and Smart (2004) and other studies that use combined federal and provincial CIT rates.

¹⁸ Some authors, rather than using the tax rate as one of the explanatory variables, use a variable defined as one minus the tax rate (see footnote 8). A coefficient of -1.1 for the elasticity of taxable income with respect to provincial CIT rates translates into an elasticity of 3.3 with respect to one minus the tax rate. In comparison, Mintz and Smart (2004) obtained an elasticity of 4.9; note however that the elasticity obtained in this study is in respect of provincial CIT rates only, while the elasticity estimated in Mintz and Smart (2004) was in respect of the combined federal and provincial CIT rate.

¹⁹ Although it would be possible to estimate a linear specification (as opposed to a log-linear specification) and thereby be able to include negative income values directly in the model, the log-linear specification has the attractive property that the coefficients can be interpreted directly as elasticity estimates.

Table of Contents