Abstract
Empirical estimates are presented of the policy and structural determinants of the natural rate of unemployment in Canada. The structural features of the economy that impinge on the adjustment of real wages to their equilibrium level are discussed, and estimates are presented showing how the generosity of the unemployment insurance system is related to past levels of unemployment. The empirical results indicate that government policies have been largely responsible for changes in the natural rate, and, hence, they can contribute to a reduction in the natural rate in the medium term.
There are two important policy issues concerning the natural rate of unemployment. The short-run macroeconomic issue concerns the level of the natural rate as an indicator of likely inflationary or disinflationary pressures emanating from the labor market. The long-run issue concerns the structural determinants of the natural rate and the extent to which economic policies can contribute to a reduction of the natural rate.
The aim of this paper is to assess empirically how policy variables and other structural aspects of the Canadian economy may have affected the natural rate since the early 1970s. Such an assessment provides a basis for judging the extent to which structural policies might contribute to a reduction in the natural rate in the medium term. A reduction in the natural rate is an important goal, since it would ease the constraints on macroeconomic policies by reducing inflationary pressures while at the same time tending to increase employment and output.
The paper is organized as follows. Section I presents a general discussion of the structural determinants of the natural rate of unemployment. One of the more important structural determinants of the natural rate is the generosity of the unemployment insurance system, which is discussed in Section II. Section III presents a number of alternative estimated unemployment rate equations and derives an equation for the natural rate. Based on this equation, Section IV discusses changes in the natural rate of unemployment since the early 1970s and the associated policy implications.
I. Structural Determinants of the Natural Rate
Friedman”s (1968, p. 8) definition of the natural rate of unemployment provides a convenient starting point:
The “‘natural rate of unemployment,” in other words, is the level that would be ground out by the Walrasian system of general equilibrium equations, provided there is imbedded in them the actual structural characteristics of the labor and commodity markets, including market imperfections, stochastic variability in demands and supplies, the cost of gathering information about job vacancies and labor availabilities, the costs of mobility, and so on.
Many of the structural characteristics of labor markets will be relatively stable and hence will not contribute to changes in the natural rate. The following discussion focuses on those structural or institutional features of the Canadian labor market which have changed over the last 15 to 20 years and for which data, or reasonable proxies, are available.
Market imperfections include those structural features of the economy that might prevent or impede real wages from adjusting to levels consistent with full employment. The deleterious effects of these imperfections on unemployment are likely to be more apparent when there are large shocks, a point that will be returned to below. Two structural aspects of the labor market that might distort real wages, and thereby affect the natural rate, are minimum wages and the degree of unionization of the labor force, both of which have changed substantially since 1971 in Canada (Figure 1): from the mid-1970s to the mid-1980s, the ratio of minimum wages to commercial wages declined from about 50 percent to less than 40 percent; the proportion of the labor force that was unionized rose from about 33 percent in 1971 to almost 40 percent in the early 1980s, and has subsequently declined somewhat.
Canada: Structural Aspects of the Labor Market, 1971-88
Citation: IMF Staff Papers 1990, 001; 10.5089/9781451956863.024.A003
Taxes are another structural feature of the economy that might affect the natural rate of unemployment by (1) increasing the cost of labor relative to other factors of production, thereby affecting the demand for labor; (2) affecting the decision to invest and produce; or (3) distorting the choice between work and leisure. In Canada average indirect and personal tax rates have been broadly stable since 1971. There have, however, been large increases in employers’ contributions for social security and pensions, particularly from the early 1970s to the early 1980s. (See Figure 1.) Similar increases in payroll taxes have occurred in other industrialized countries, and the possible impact on the natural rate has often been stressed in analyses of European unemployment.1
The role that variability in demands for. and supplies of, goods and labor may have played in increasing the natural rate of unemployment has received considerable attention.2 In the empirical work discussed below, two measures of variability have been used: (1) the coefficient of variation of employment across industries, which has increased since the mid-1970s, and (2) the coefficient of variation of employment across provinces, which declined from the early 1970s to the early 1980s and subsequently rose. (See Figure 1.) Changes in occupational or geographical mobility would also be expected to affect the natural rate of unemployment. The variability of employment across industries might serve as a proxy for occupational mobility, and geographical mobility might be proxied by gross interprovincial labor force flows. (See Figure 1.)
There are empirical, as well as conceptual, problems with these ex post proxies for variability and mobility. One is that some form of wage rigidity would seem to be necessary for changes in the variability of output to affect the natural rate. Another is that increased measured mobility may accompany, or be a manifestation of, increased variability, in which case the impact on the natural rate would be uncertain. Finally, measured variability or mobility may be highly correlated with the business cycle and have little to do with structural changes in the economy.3 Perhaps for these reasons, there does not appear to be a consensus on the empirical significance of mobility or structural dispersion in determining the natural rate in Canada.
Information on the cost of job search, another structural characteristic noted by Friedman, is not readily available. Data are available, however, for some important aspects of the unemployment insurance system that can be expected to affect the reservation wage, and hence the cost of job search. As discussed in the following section, there have been important changes in many aspects of the unemployment insurance system that may have affected the natural rate in Canada.
Demographic factors are often emphasized as important structural determinants of the natural rate of unemployment, and, as in other countries, there have been relatively large changes in the composition of the labor force in Canada. (See Figure 1.) Although increases in specific segments of the labor force—such as traditionally high-unemployment groups or groups with a weak attachment to the labor force—may tend to increase the natural rate, calculations holding labor market shares constant indicate that the magnitude of this compositional effect is small,4 For demographic changes to have large impacts on the natural rate, some form of rigidity preventing the real wages of the various demographic cohorts from adjusting to their equilibrium levels would seem to be necessary.
This highlights a general point referred to above. Supply or demand shocks, such as changes in oil prices or the demographic composition of the labor force, may be associated with changes in unemployment that may last for some time if real wages fail to adjust. Should there be an impact on the natural rate, however, the ultimate cause would not be the shocks per se, but rather those structural aspects of the economy that prevent the adjustment of real wages. In the empirical analysis that follows, emphasis is placed on identifying structural features of the economy that might directly impinge on the flexibility with which real wages adjusted to shocks.
Although those determinants of the natural rate most often included in empirical studies for Canada have been discussed above,5 there might, of course, be others. The possibility that the natural rate of unemployment is positively related to the past history of unemployment—in which case the natural rate is said to have the property of hysteresis—has received increasing attention in discussions of European unemployment.6 A casual examination of the unemployment rate data suggests hysteresis may also be important in Canada: from 1966 to 1974 the annual unemployment rate averaged less than 5 percent and never exceeded 5½ percent; since 1975 the annual unemployment rate in Canada has averaged almost 9 percent and has remained well above 7 percent during the current sustained expansion of output. Recent empirical analyses suggest that hysteresis may account, at least in part, for the persistence of high unemployment in Canada.7 This conclusion is supported by the empirical analysis reported below examining the relationship between natural rate of unemployment and the generosity of the unemployment insurance system, which, in turn, depends on past rates of unemployment.
II. Unemployment and Unemployment Insurance
As mentioned earlier, the generosity of the unemployment insurance (UI) system is an important determinant of the natural rate of unemployment. In 1971 and 1977 there were a number of significant changes that increased the generosity of the UI system.
Changes in the Unemployment Insurance Program8
In 1971 the coverage of the UI program was increased from about 67 percent to 90 percent of the labor force, and the ratio of the average benefit to earnings was raised by about 50 percent. The maximum unemployment insurance replacement rate, adjusted for the proportion of the labor force covered by the system, is shown in the top panel of Figure 2. Other changes made in the unemployment insurance program in 1971 and 1977 concerned the number of weeks of work required to qualify for UI benefits and the number of weeks in the benefit period.
The minimum period of work required to qualify a person for unemployment insurance benefits was lowered in 1971 from 30 weeks in the preceding 2 years to 8 weeks in the preceding year. The basic benefit period of 15 weeks was replaced with a four-phase benefit period, which was conditional on both national and regional unemployment rates. The first two phases, which had to be applied for separately, provided a total of 18 weeks of benefits. Phase III provided for an additional 4 weeks if the national unemployment rate exceeded 4 percent, or an additional 8 weeks if it exceeded 5 percent. Phase IV extended the benefit period in those regions where the unemployment rate exceeded the national average: benefits were extended by 6 weeks if the regional unemployment rate was 1-2 percentage points higher than the national average, by 12 weeks if it was 2-3 percentage points higher, and by 18 weeks if it was more than 3 percentage points higher. From 1971 to 1977. the national unemployment rate was always above 5 percent, so the basic benefit period was 26 weeks. This could be extended to 44 weeks in regions where the unemployment rate exceeded the national average by at least 3 percentage points.
Canada: Measures of Generosity of the Unemployment Insurance (111) System, 1970-88
Citation: IMF Staff Papers 1990, 001; 10.5089/9781451956863.024.A003
a The calculation of the adjustment for work requirements and benefit periods is discussed in the text.b Two-period moving average.In 1977 the unemployment insurance program assumed its current form. Both the qualifying and the benefit periods were made to depend upon the level of regional unemployment rates, rather than the differential between regional and national unemployment rates. Benefits became payable in three phases for a maximum benefit period of 50 weeks per year. A variable minimum work requirement of from 10 to 14 weeks was introduced, depending on the unemployment rate in the region: 10 weeks if the regional unemployment rate was 9 percent or more, and 14 weeks tf the regional unemployment rate was 6 percent or less. Repeat claimants were required to work 16 to 20 additional weeks to qualify for further benefits—a requirement that was waived in regions with unemployment rates over 11.5 percent. With regard to benefit periods, in the first phase the claimant receives 1 week of benefits for each week worked, up to a maximum of 25 weeks of benefits. In the second phase, the claimant receives 1 week of benefits for every 2 weeks worked beyond 25 weeks, up to a maximum of 13 weeks of benefits. In the third phase, the claimant can receive from 2 to 32 weeks of regional extended benefits, depending on the regional unemployment rate—2 weeks of benefits for every half a percentage point that the regional unemployment rate exceeds 4 percent.
Unemployment and Generosity of the Unemployment Insurance System
These changes have made the generosity of the unemployment insurance system dependent on past developments in national and regional rates of unemployment. For example, a claimant with 10 weeks of previous work in a region with an unemployment rate over 11.5 percent would be eligible for 42 weeks of benefits (10 under Phase I and 32 under Phase III), and a claimant with over 50 weeks of previous work in a region with an unemployment rate of less than 4 percent would be eligible for 38 weeks of benefits (25 under Phase I and 13 under Phase II).
The way in which the unemployment insurance system interacts with actual unemployment to affect the length of the work requirement and the benefit period is illustrated in the second panel of Figure 2, which presents estimates of the minimum work requirement for unemployment insurance benefits (MINQ) and extended regional benefits (ERE), as well as the maximum benefit period (MAXB) for someone meeting the minimum work requirement—all measured in weeks per year. The estimates in Figure 2 are weighted averages of calculations for each of the ten provinces, with the weights based on provincial shares of total unemployment.9 For each province (denoted by an “i”) the calculations are as follows, where U is the aggregate national unemployment rate and Ui is the provincial unemployment rate, both in the preceding quarter:
Prior to 1971
MINQi = 15
ERBi = 0
MAXBi = 15
1971 to 1977
MINQi = 8
MAXBi = 26 + ERBi
1978 to 1988
MAXBi = MINQi + ERBi.
Although these calculations only approximate average lengths of qualifying and benefit periods, it is clear that after 1978 developments in extended regional benefits, and hence the maximum benefit periods, were closely related to past developments in the unemployment rate.
The number of weeks which must be worked in a year to qualify for unemployment insurance benefits (MINQ) and the maximum number of weeks in a year for which a person fulfilling the minimum work requirements would be eligible (MAXB) are important dimensions of the generosity of the unemployment insurance system. To incorporate these elements into the estimated equations presented in the next section, the unemployment insurance replacement rate, adjusted for the proportion of the labor force covered by the unemployment insurance system, has been multiplied by an adjustment factor defined as (52- MiNQ)/(52-MAXB)10 If MINQ = MAXB, as it did prior to 1971, the adjustment factor is one; if MAXB is greater than MINQ, the adjustment factor is greater than one, indicating an increase in the generosity of the UI system; and vice versa if MAXB is less than MINQ. The UI replacement rate with and without this adjustment factor is shown in the top panel of Figure 2. From 1982 to 1984 the extension of regional benefits, linked to increases in the unemployment rate, had the effect of sharply increasing the generosity of the UI system. Since 1984 the average weeks of regional extended benefits has declined with the drop in unemployment rates.
III. Unemployment Rate Equations
There are a number of ways to estimate the natural rate of unemployment. Perhaps the most common method is to estimate a wage and/or price equation as a function of the unemployment rate and then solve for the implicit natural rate as that unemployment rate which is consistent with stable inflation. A second, somewhat more direct approach is to estimate an unemployment rate equation as a function of cyclical and structural variables and estimate the natural rate from its structural determinants.11
Estimation Results
The analysis below follows the second approach based on an aggregate unemployment rate equation estimated using quarterly data from the first quarter of 1971 to the second quarter of 1988. a period during which there were large changes in the Canadian unemployment rate. The hiring and firing practices of firms suggest that the unemployment rate will react to changes in its cyclical and structural determinants with a lag. Accordingly, the estimation strategy, which differs somewhat from that followed in most empirical studies in the literature,12 was to start with a genera] autoregressive distributed-lag specification (of up to four quarters) and then test down by dropping insignificant variables. Few restrictions were placed on the lag distributions, although in some cases the free-estimation results suggested obvious patterns (for example, first differences and moving averages) which were imposed to conserve degrees of freedom.13
A number of alternative equations, estimated by ordinary least squares, are reported in Table 1. The notation for the variables is as follows:14 U denotes the unemployment rate, CU capacity utilization in goods-producing industries, GDPGAP the logarithm of the ratio of real gross domestic product (GDP) to its quadratic trend, RELPREN the logarithm of the consumer price of energy relative to the total consumer price index (CPI), RELPREX the logarithm of the deflator for exports relative to the deflator for imports, UIRR the maximum unemployment insurance replacement rate adjusted for the proportion of the labor force covered (with UIRR also incorporating, in equation (1), the adjustment for work requirements and benefit periods described above), RELMW the logarithm of the minimum wage relative to the average commercial wage, LF%UNION the percentage of the labor force which is unionized, TAXSIP the average payroll tax rate for employers’ contributions for social security and pension funds, MOBILITY the average of inter-provincial labor force inflows and outflows as a percentage of the labor force, DISPROV the coefficient of variation of employment across provinces, and DISIND the coefficient of variation of employment across industries. The policy and structural variables are displayed in Figures 1 and 2.
Canada: Unemployment Rate Equations
In equation (1), the unemployment insurance replacement ratio (UIRR) is adjusted for the proportion of the labor force covered by the unemployment insurance system and for the length of the qualifying and benefit periods; in equations (2)-(4), UIRR is adjusted only for coverage.
Canada: Unemployment Rate Equations
| Equation Number | ||||
|---|---|---|---|---|
| Equation | (1) | (2) | (3) | (4) |
| Constant | -3.767 | -2.875 | -10.810 | -15.883 |
| (1.8) | (1.3) | (2.8) | (2.4) | |
| CU - CU-1 | -0.097 | -0.110 | -0.103 | -0.109 |
| (4.0) | (4.4) | (3.9) | (4.8) | |
| GDPGAPMsub>-1 | -0.193 | -0.203 | -0.171 | -0.234 |
| (6.3) | (6.3) | (5.3) | (7.6) | |
| RELPREN | 0.029 | 0.031 | 0.018 | 0.035 |
| (3.7) | (3.7) | (2.4) | (4.9) | |
| RELPREX-| | -0.042 | -0.039 | -0.025 | |
| (4.1) | (3.9) | (2.8) | ||
| UIRRa | 0.004 | 0.014 | 0.020 | 0.013 |
| (2.4) | (2-3) | (2.7) | (1.9) | |
| RELMW-1 | 0.020 | 0.025 | 0.023 | |
| (2.2) | (2.4) | (2.1) | ||
| LF% UNION | 0.169 | 0.141 | 0.096 | 0.144 |
| (3.3) | (2.6) | (1.7) | (2.3) | |
| TAXSIP-1 | 0.367 | 0.317 | 0.631 | 0.533 |
| (3.5) | (3.2) | (3.8) | (4.3) | |
| LF%55+ | 0.493 | |||
| (2.8) | ||||
| MOBILITY-1 | -0.106 | |||
| (1.2) | ||||
| DISPROV-2 | 0.053 | |||
| (3-4) | ||||
| DISIND -2 | -0.041 | |||
| (3.9) | ||||
| U-1 | 0.737 | 0.779 | 0.900 | 0.680 |
| (6.3) | (7.0) | (8.1) | (6.5) | |
| U-2 | -0.322 | -0.269 | -0.317 | -0.210 |
| (3.8) | (3.0) | (3-4) | (2.5) | |
| RBSQ | 0.988 | 0.988 | 0.987 | 0.991 |
| S.E. | 0.238 | 0.238 | 0.251 | 0.214 |
| F-statistics: | ||||
| Autocorrelation (lags 1-4) | ||||
| (CV = 2.5) | 1.96 | 2.20 | 1.59 | 2.81 |
| ARCH (4 lags) | ||||
| (CV = 2.6) | 2.65 | 1.87 | 1.01 | 0.58 |
| Break First Quarter of 1984 | ||||
| (CV = 2.0) | 0.88 | 0.92 | 1.00 | 0.98 |
| Break Second Quarter of 1982 | ||||
| (CV = 1.8) | 1.83 | 1.65 | 1.65 | 1.33 |
In equation (1), the unemployment insurance replacement ratio (UIRR) is adjusted for the proportion of the labor force covered by the unemployment insurance system and for the length of the qualifying and benefit periods; in equations (2)-(4), UIRR is adjusted only for coverage.
Canada: Unemployment Rate Equations
| Equation Number | ||||
|---|---|---|---|---|
| Equation | (1) | (2) | (3) | (4) |
| Constant | -3.767 | -2.875 | -10.810 | -15.883 |
| (1.8) | (1.3) | (2.8) | (2.4) | |
| CU - CU-1 | -0.097 | -0.110 | -0.103 | -0.109 |
| (4.0) | (4.4) | (3.9) | (4.8) | |
| GDPGAPMsub>-1 | -0.193 | -0.203 | -0.171 | -0.234 |
| (6.3) | (6.3) | (5.3) | (7.6) | |
| RELPREN | 0.029 | 0.031 | 0.018 | 0.035 |
| (3.7) | (3.7) | (2.4) | (4.9) | |
| RELPREX-| | -0.042 | -0.039 | -0.025 | |
| (4.1) | (3.9) | (2.8) | ||
| UIRRa | 0.004 | 0.014 | 0.020 | 0.013 |
| (2.4) | (2-3) | (2.7) | (1.9) | |
| RELMW-1 | 0.020 | 0.025 | 0.023 | |
| (2.2) | (2.4) | (2.1) | ||
| LF% UNION | 0.169 | 0.141 | 0.096 | 0.144 |
| (3.3) | (2.6) | (1.7) | (2.3) | |
| TAXSIP-1 | 0.367 | 0.317 | 0.631 | 0.533 |
| (3.5) | (3.2) | (3.8) | (4.3) | |
| LF%55+ | 0.493 | |||
| (2.8) | ||||
| MOBILITY-1 | -0.106 | |||
| (1.2) | ||||
| DISPROV-2 | 0.053 | |||
| (3-4) | ||||
| DISIND -2 | -0.041 | |||
| (3.9) | ||||
| U-1 | 0.737 | 0.779 | 0.900 | 0.680 |
| (6.3) | (7.0) | (8.1) | (6.5) | |
| U-2 | -0.322 | -0.269 | -0.317 | -0.210 |
| (3.8) | (3.0) | (3-4) | (2.5) | |
| RBSQ | 0.988 | 0.988 | 0.987 | 0.991 |
| S.E. | 0.238 | 0.238 | 0.251 | 0.214 |
| F-statistics: | ||||
| Autocorrelation (lags 1-4) | ||||
| (CV = 2.5) | 1.96 | 2.20 | 1.59 | 2.81 |
| ARCH (4 lags) | ||||
| (CV = 2.6) | 2.65 | 1.87 | 1.01 | 0.58 |
| Break First Quarter of 1984 | ||||
| (CV = 2.0) | 0.88 | 0.92 | 1.00 | 0.98 |
| Break Second Quarter of 1982 | ||||
| (CV = 1.8) | 1.83 | 1.65 | 1.65 | 1.33 |
In equation (1), the unemployment insurance replacement ratio (UIRR) is adjusted for the proportion of the labor force covered by the unemployment insurance system and for the length of the qualifying and benefit periods; in equations (2)-(4), UIRR is adjusted only for coverage.
The estimated equations explain developments in Canada’s unemployment rate well. All of the equations pass the diagnostic tests for the absence of serial correlation (except equation (4)) and for parameter constancy over time. More important, the equations contain a relatively large number of policy variables and other structural determinants of the natural rate. This reflects, in part, the relatively unconstrained estimation procedure. The pattern of the estimated coefficients on the two lagged dependent variables means that the impact of each of the independent variables declines, and then rises somewhat before tailing off to zero.15
The cyclical movements of the unemployment rate are determined by two indicators of capacity utilization: the deviation of output from trend,16 which is for the total economy; and capacity utilization in the goods-producing sector. The cyclical variables are supplemented by the relative price of energy and the terms of trade in order to capture additional short-run impacts of supply shocks on unemployment. Although these supply-shock variables may have temporary impacts on the rate of unemployment consistent with stable inflation (that is, on the non-accelerating-inflation rate of unemployment, the NAIRU). they have not been used in the calculation of the natural rate. To the extent that these variables fully account for the cyclical changes in the unemployment rate, the secular movements in the natural rate will be determined by the structural variables.
The estimation results point to the importance of four structural determinants of the natural rate: (1) the unemployment insurance replacement rate, adjusted for coverage and qualifying and benefit periods. (2) the relative minimum wage, (3) the percentage of the tabor force that is unionized, and (4) payroll taxes.17 Of these structural determinants, only the degree of unionization of the labor force is not under the direct influence of government policy. Equations (1) and (2) are identical, except that in the latter the unemployment insurance replacement rate is not adjusted for the effects of the minimum number of weeks of work required to qualify for unemployment benefits and the maximum number of weeks of unemployment benefits (as described at the end of Section II),18 Possible effects from effective personal income taxes and indirect taxes were also tested, but the estimated coefficients were insignificant. This suggests that a revenue-neutral tax change reducing payroll taxes and increasing other taxes would tend to lower the natural rate of unemplfigoyment.
Various demographic variables were also included in the estimated equations. The percentage of the labor force which is young (aged 15 to 24) and/or female gave either insignificant or perversely signed (that is, negative) estimated coefficients. Equation (3) includes the percentage of the labor force aged 55 and older and excludes the terms of trade variable. The estimated coefficient is positive, implying that an increase in the share of older workers will be associated with increased structural unemployment. This may reflect less flexibility on the part of older workers in responding to changes in output and demand, but seems inconsistent with the greater experience of, and thus the greater human capita] embodied in, an older work force.
Equation (4) includes measures of geographical mobility and the dispersion of industrial and provincial employment. The estimated coefficient on the mobility variable is correctly signed but not significantly different from zero. The estimated coefficient on the variation of employment across provinces is positive and significant, and the estimated coefficient on the variation of employment across industries is negative and significant.19 Entered separately, however, neither variable was significant.
Equation for the Natural Rate
An equation for the natural rate can be extracted from the dynamic equations reported in Table 1 by solving for the long-run steady-state relationship between the unemployment rate and its structural determinants. This long-run focus seems particularly appropriate for an equation describing one of the key structural features of the economy. Applying this procedure to equation (1), the natural rate can be calculated as follows:
where the (-statistics reported (in parentheses below the coefficients) are based on approximate large-sample standard errors.
The constant can be calculated by making explicit assumptions about the cyclically neutral level of the GDP-gap variable and the shock-free levels of the two supply-shock variables. Two assumptions have been made: in the first case, these variables were set at their sample-period averages; in the second case, the GDP gap was set at its average value from mid-1987 to mid-1988—a period when most observers consider output to have been at or near capacity.
The two resulting estimates of the natural rate, which bracket recent unemployment rates, are shown in Figure 3. Taking the average of the two estimates suggests that the natural rate was about 8 percent in early 1988. This estimate of the natural rate of unemployment—which is sensitive to the assumptions used concerning the cyclical and supply-shock variables—is broadly consistent with other recent estimates that tend to be in the 7-9 percent range.20 Needless to say, large confidence intervals are likely to be associated with any point estimate of the natural rate.
IV. Changes in the Natural Rate and Policy Implications
The estimation results suggest that there were steady increases in the natural rate of unemployment in Canada from 1971 to 1978. After cyclical movements in 1973 and 1974, the increases in the Canadian unemployment rate from the mid-1970s to the late 1970s largely reflected increases in the natural rate. Although the time profiles of both the actual and the natural rates were similar in the following decade, the large increase and subsequent decline in unemployment during 1981-87 primarily reflected cyclical factors rather than changes in the natural rate.
The contributions which each of the structural determinants is estimated to have made to changes in the natural rate since 1971 are shown in Figure 4, The main factors affecting the natural rate of unemployment are estimated to have been the following.
• The increase in unemployment insurance benefits relative to wages—adjusted for coverage and qualifying and benefit periods—is estimated to have increased the natural rate by almost 1 percentage point from 1970 to mid-1972.21 From 1977 to 1981, the natural rate may have decreased by about 0.3 percentage point, owing to reductions in the adjusted replacement rate. Increases in the adjusted replacement rate, primarily due to large increases in extended regional benefits, are estimated to have increased the natural rate by about 1 percentage point from 1981 to 1983. Thereafter, the decline in regional benefits associated with the decline in unemployment rates may have lowered the natural rate by about 0.5 percentage point. The net effect of changes in the adjusted unemployment replacement rate from 1970 to 1988 may have been to increase the natural rate by about 1 percentage point.
Canada: Natural Rale of Unemployment, 1971-88
Citation: IMF Staff Papers 1990, 001; 10.5089/9781451956863.024.A003
a The two estimates of the natural rate are based on alternative assumptions of the GDP-gap variable and the cyclically neutral levels of the supply-shock variables. The higher estimate is based on sample-period averages; the lower estimate is based on the average value of the GDP gap from mid-19S7 to mid-1988.Canada: Contributions to Changes in the Natural Rate of Unemployment, 1970-88
Citation: IMF Staff Papers 1990, 001; 10.5089/9781451956863.024.A003
Note: The contribution of each variable to changes in the natural rate of unemployment has been normalized to be zero in the fourth quarter of 1970.• The steady decline in relative minimum wages (that is, minimum wages relative to average commercial wages) is estimated to have reduced the natural rate by about 1 percentage point from the mid-1970s to the mid-1980s. Since the mid-1980s, relative minimum wages have been roughly stable.
• The increased unionization of the labor force from 1971 to 1978 is estimated to have increased the natural rate by between 1.5 and 2 percentage points. The degree of unionization moved cyclically from 1978 to 1983, and the net effect may have been to increase the natural rate by almost 0.5 percentage points. Subsequent declines in unionization may have decreased the natural rate of unemployment by about 1 percentage point.
• The rise in payroll taxes is estimated to have increased the natural rate by 1.5 percentage points from 1971 to the late 1970s, and by another 1 percentage point since then.
The factors accounting for the drop in the natural rate since the early 1980s are the continuing declines in the proportion of the labor force which is unionized and the generosity of unemployment insurance benefits. The latter decline is related to the shortening of the extended regional benefit period associated with the decline in unemployment rather than to reductions in the unemployment insurance replacement rate or the proportion of the labor force covered by the unemployment insurance system. This hysteresis type of effect, whereby the natural rate is dependent on past levels of unemployment through the working of the unemployment insurance system, may have increased the natural rate of unemployment by as much as 1 percentage point in 1982 and 1983, when unemployment rates were at postwar highs; and while the subsequent decline in the actual unemployment rate served to reduce the adverse effect of the extended regional benefits, it is estimated that the natural rate of unemployment is still almost 0.5 percentage point higher at present than it would have been in the absence of such benefits.
The policy implications can be summarized briefly. If government policies contributed to increases in the natural rate of unemployment during the 1970s, and there seems to be broad agreement that they did,22 policies should be able to contribute to decreases in the natural rate in the period ahead,23 Government policies, of course, have multiple objectives, including the reduction of fiscal deficits. This suggests that emphasis might be placed on reductions in the generosity of the unemployment-insurance system, particularly through the elimination or reduction of extended regional benefits,24 and on continued declines in relative minimum wages. These types of structural policies, by reducing the natural rate, would ease the constraints on macroeconomic policies by contributing to lower unemployment and inflation in the medium term.
REFERENCES
Abraham, Katharine G., and >Lawrence F. Katz, “Cyclical Unemployment: Sectoral Shifts or Aggregate Disturbances?” Journal of Political Economy, Vol. 94 (June 1986).
Adams, Charles, and David T. Coe, “A Systems Approach to Estimating the Natural Rate of Unemployment and Potential Output for the United States,” Working Paper No. 89/89, International Monetary Fund. October 26, 1989.
Ashenfelter, Orley, and David Card, “Why Have Unemployment Rates in Canada and the U.S. Diverged?” NBER Working Paper No. 1840 (Cambridge, Massachusetts: National Bureau of Economic Research, February 1986).
Bean, Charles R., Richard Layard, and Stephen Nickell, “The Rise in Unemployment: A Multicountry Study,” Economica, Vol. 53 (Supplement 1986).
Blanchard, Olivier J., and Lawrence H. Summers, “Hysteresis and the European Unemployment Problem,” NBER Macroeconomics Annual (National Bureau of Economic Research), Vol. 1 (1986).
Chan-Lee, James, David T. Coe, and Menahem Prywes, “Microeconomic Changes and Macroeconomic Wage Disinflation in the 1980s,” OECD Economic Studies (Organization for Economic Cooperation and Development), No. 8 (Spring 1987).
Coe, David T., “Hysteresis Effects in Aggregate Wage Equations,” in Unemployment, Hysteresis and the Natural Rate Hypothesis, ed. by Rod Cross (Oxford, England: Basil Blackwell, 1988).
Ford, Robert, and David E, Rose, “Estimates of the NAIRU Using an Extended Okun’s Law,” Bank of Canada, Working Paper 89-3 (Ottawa, November 1989).
Friedman, Milton, “The Role of Monetary Policy,” American Economic Review, Vol. 58 (March 1968).
Green, Christopher, and Jean-Michel Cousineau, Unemployment in Canada: The Impact of Unemployment Insurance (Ottawa: Economic Council of Canada, 1976).
Grubel, Herbert G., Dennis Maki, and Shelley Sax, “Real and Insurance-Induced Unemployment in Canada,” Canadian Journal of Economics, Vol. 8 (May 1975).
Layard, Richard, and Stephen Nickell, “The Labour Market,” in The Performance of the British Economy, ed. by Rudiger Dornbusch and Richard Layard (Oxford, England: Clarendon Press, 1987).
Lilien, David M., “Sectoral Shifts and Cyclical Unemployment,” Journal of Political Economy, Vol. 90 (August 1982).
Lindbeck, Assar, and Dennis Snowcr, “Cooperation, Harassment, and Involuntary Unemployment: An Insider-Outsider Approach,” American Economic Review, Vol. 78 (March 1988).
Milbourne, Ross D., D.D. Purvis, and W.D. Scones, Unemployment Insurance and Unemployment Dynamics, Institute for Economic Research, Discussion Paper No. 750 (Kingston, Ontario: Queen’s University, June 1989).
Pelletier, Jacques, “Estimations du NAIRU avec la Courbe de Phillips,” Bank of Canada Working Paper (Ottawa, forthcoming 1990).
Rose, David E., “The NAIRU in Canada: Concepts, Determinants and Estimates,” Bank of Canada, Technical Report No. 50 (Ottawa, December 1988).
Samson, Lucie, “A Study of the Impact of Sectoral Shifts on Aggregate Unemployment in Canada,” Canadian Journal of Economics, Vol. 18 (August 1985).
David T. Coe is a Senior Economist in the Current Studies Division of the Research Department. He is a graduate of the University of Maryland and the London School of Economics and holds a doctorate from the University of Michigan. This paper was written when he was in the North American Division of the Western Hemisphere Department while on leave of absence from the Organization for Economic Cooperation and Development.
See, for example. Bean. Layard, and Nickell (1986) and Layard and Nickell (1987). For an international comparison of developments in nonwage labor costs as a percentage of wages and salaries since the mid-1960s, see Table 5 on page 63 of Chan-Lee, Coe, and Prywes (1987).
See Lilien (1982) and Samson(1985).
See, for example, Lindbeck and Snower (1988) and Blanchard and Summers (1986).
See Milbourne, Purvis, and Scones (1989) and Coe (1988).
The following discussion, which is based on Green and Cousineau (1976) and Ashenfelter and Card (1986), only concerns changes in the unemployment insurance (UI) system between 1971 and 1988. In the spring of 1989 the Canadian Government announced changes to the UI system, including some tightening of eligibility requirements and a shortening of benefit periods.
The unemployment insurance system is administered in 48 regions, rather than the 10 provinces used in the calculations presented here.
If either the difference between, or the ratio of, MAXB and MINQ is used as the adjustment factor, the estimation results are very similar to those reported below.
Alternative approaches to estimating the natural rate of unemployment are discussed in Adams and Coe (1989).
For a review of the empirical literature, see Rose (1988). The sensitivity of natural rate estimates to alternative methodologies and data is discussed in Pelletier (1990).
When alternative lag specifications gave conflicting signals as to the significance of specific variables—for example, because of multicollinearity between the relatively large number of explanatory variables—preference was given to keeping structural variables in the estimated equations.
Variables expressed as the logarithms of ratios have been multiplied by 100, so that all variables are entered in what are effectively percentage terms, UIRR and TAXSIP are entered as two-period moving averages. Data sources are as follows: The unemployment insurance replacement rate adjusted for coverage, minimum wages, and commercial wages is from the Bank of Canada. The average rate for employers’ contributions (for social security and pension funds), expressed as a percent of total wages and salaries, is from the Organization for Economic Cooperation and Development’s Standardized National Accounts and is interpolated from annual data. The percentage of the labor force which is unionized is from Labour Canada’s annual Directory of Labour Organisations in Canada (Ottawa) and is interpolated from annual data; the 1979 observation is not available and was set equal to the average of 1978 and 1980. All other data are from Statistics Canada’s CANSIM data base.
That is, the lag distribution initially resembles a third-order polynomial, with the peak in the third quarter, before asymptotically approaching zero.
When the squares of the quadratic-gap variable were added to the equation, in order to test for nonlinearities, its estimated coefficient was marginally significant, but there was little difference in the overall estimation results. Using a linear trend with a break in 1973, rather than a quadratic trend, also had little overall effect, although the significance of the estimated coefficient on the unemployment replacement rate was reduced somewhat.
Entering separate tax rates for employers’ contributions to social security and to pension funds resulted in significant estimated coefficients of almost identical magnitude on each variable, but had little impact on the overall regression results. The significance, but not the sign, of the estimated coefficients on relative minimum wages and the replacement rate was sensitive to the inclusion of the employers’-contribution variable.
The smaller estimated coefficient on the UIRR variable in equation (1), compared with those in equations (2)-(4), reflects the fact that the adjustment factor for qualifying and benefit periods increases the magnitude and variance of the UIRR variable. (Compare Figure 2.)
Increased dispersion across industries may be a proxy for occupational mobility tending to decrease the natural rate, whereas increased dispersion across provinces may be a proxy for structural changes tending to increase the natural rate.
See Rose (1988) and the references cited therein. The higher estimate in Figure 3 is based on sample-period averages for the cyclical and supply-shock variables. If the average relative price of energy during 1979-88 were used instead of the sample-period average, the estimate of the natural rate would be correspondingly higher.
This estimate is similar to the 0.8 percentage point estimate of Grube 1, Maki. and Sax (1975, p. 187).
The estimation results suggest, for example, that each of the following policies might contribute to lowering the natural rate by about ½ of 1 percentage point during 1989—93: continued gradual declines in relative minimum wages at the same rate as occurred, on average, from the mid-1970s to 1988; the elimination of regional extended benefits for unemployment insurance benefits; or the gradual reduction in payroll taxes to about their 1982 levels.
The proposed reforms to the UI system, announced in early 1989, reduce extended original benefits somewhat, but only in regions with low unemployment rates.
