Accounting for regional differences in per capital personal income growth: an update and extension

Accounting for regional differences in per capital personal income growth: an update and extension

Daniel H. Garnick

Accounting for Regional Differences in Per Capita Personal Income Growth: An Update and Extension

AFTER five decades of narrowing, regional differences in per capita personal income (PCPI) as a percent of the national average widened in the 1980’s. From 1929 to 1979, PCPI increased from 64 to 90 percent of the national average in the low-income regions (Southeast, Southwest, Plains, and Rocky Mountain) and declined from 127 to 107 percent in the high-income regions (Mideast, Far West, New England, and Great Lakes). From 1979 to 1988, the low-income regions slipped back to 88 percent of the national average, and the high-income regions rose to 109 percent (chart 6 and tables 1 and 2).

[Tabular Data Omitted]

This article updates and extends the analytical measures that had been used to account for regional differences in PCPI growth in the SURVEY OF CURRENT BUSINESS in 1982.(1) It explores the variations over time in factors of production, factor returns, and transfer payments and in the rates of change in income and population. It concludes that the percentage-point shift in the 1980’s, although small, was the product of divergent regional changes that were highly significant.

The article first presents an overview and then analyzes the components of PCPI change in the three following sections. The first of these sections describes the methodology of decomposing PCPI change into industry mix, differential regional earnings, job ratio, working-age ratio, property income ratio, and transfer payments ratio. The second compares the regional patterns of these six components in the 1970’s and 1980’s. The third describes the methodology of decomposing PCPI change into components capturing the “lift” effect of income and the “drag” effect of population, and it reviews the effects for six timespans in 1929-88.


In the earlier article, PCPI was decomposed into components, and their statistical contributions to the narrowing of regional relative differences were estimated. In 1940-79, more uniform regional industrial composition of jobs (industry mix) accounted for about one-half of the narrowing of regional relative differences in PCPI growth. Reduced differential regional earnings (after adjusting for regional industry mix) accounted for about one-tenth, and more uniform regional ratios of jobs to working-age population (job ratio) accounted for another one-tenth. The residual one-fourth was about equally accounted for by more uniform regional distributions of personal dividend, interest, and rental income (property income) per capita and of transfer payments per capita.

In the 1980’s, the widening of regional relative differences in PCPI growth was more than accounted for by the three components of job-related income: Differential regional earnings accounted for about one-half, the job ratio for about four-tenths, and industry mix for one-eighth of the three-component total. More uniform regional distributions of property income per capita and of transfer payments per capita partly offset the components contributing to regional divergence in the 1980’s (chart 7).

New England and the Mideast more than accounted for the relative PCPI gain in high-income regions in the 1980’s. In these regions, there were relative gains in each of the three components related to job income. The Southwest, Rocky Mountain, and Plains regions more than accounted for the relative PCPI loss in low-income regions. In these regions, there were relative losses in each of the three components (chart 8).

In the last section of this article, further analysis demonstrates that regional population growth rates in the 1980’s reinforced the divergence of regional PCPI growth. In the Southwest and Rocky Mountain regions, population growth rates were 198 percent and 133 percent, respectively, of the national average; in the Mideast and New England, the rates were only 31 percent and 55 percent, respectively.

It has been an article of some faith (and some empirical analysis) that working-age population flows and job creation move in parallel, reflecting competitive forces, the only question being whether working-age population moves to the locus of jobs or jobs to the locus of working-age population. In Hicks’ theory of wages, regional convergence would result from (1) immigration to high-wage areas putting downward pressure on wage rates and (2) outmigration from low-wage areas depleting the labor reserves and putting upward pressure on wage rates.(2) In a 1960 study, Borts found that migration flows from low-to high-wage regions in the United States had occurred in 1919-53 but that the flows had not been large enough to result in substantial wage-rate convergence. He concluded that continued migration in the “right direction” was a necessary condition for convergence.(3)

In the earlier article, the 1929-79 period was broken into subperiods, of which the 1940’s and the 1970’s were the only decades in which the reduction in differential regional earnings contributed to regional PCPI convergence. The direction of migration was “right” in the 1940’s but not in the 1970’s, when, apart from the Far West, very large working-age population flows were from high- to low-income regions. (The reduction in differential regional earnings in the 1970’s coincided, to a degree, with regional convergence in relative costs of living. The rising cost of living in low-income regions might have been expected to slow immigration, but it did not.(4))

In the 1980’s, population flows continued in the “wrong” direction for the most part, despite differential regional earnings well above and unemployment rates well below the national average in New England and the Mideast.(5) (Regional costs of living diverged as well, reflecting sharply divergent relative housing costs.)

The Hicks-Borts proposition regarding the direction of labor force migration and the regional equalization of factor returns appears to have become less and less descriptive since mid-century. The agricultural revolution and the subsequent industrial and geographic reallocation of farm labor reserves had particular force in low-income regions in the 1940’s, but progressively diminished thereafter. Technological change in other industries has, in general, permitted them to be more footloose and to locate in closer correspondence with the dominant long-term population migration trends. The resulting greater regional dispersion of industry in recent decades is reflected in the decreasing importance of industry mix in explaining regional PCPI change.

The industry-mix contribution in the 1980’s reflected, in part, a reversal of unsustainable developments in the 1970’s, particularly developments affecting factor returns in the oil and gas and related industries. The relatively large contributions of differential regional earnings and of the job ratio to the high-income regions in the 1980’s may, in turn, be unsustainable and subject to reversal; that is what the competitive model would suggest.(6) However, the existence of external economies connected with particular locations – suggested by the fact that New England and Mideast experienced divergent differential regional earnings in the 1950’s and 1960’s as well as in the 1980’s – and inertia in regional population migration patterns may militate against the smooth working of the competitive model.

Per Capita Personal Income

Components and Their


Table 3 shows PCPI by type of payment for 1929, 1940, 1950, 1959, 1969, 1979, and 1988.(7) Payment types have grown at different rates, nationally and regionally, over each decade, and shares of PCPI accounted for by the payment types have varied, as shown in table 4. As an example, nationally, the share of property income was higher in 1988 than in any other year except 1929, reflecting unusually high interest rates and growing indebtedness in the 1980’s. Regionally, more uniform distributions of property income in the 1980’s partly offset factors contributing to regional disparities.(8) However, the main focus in this study is not on property income but on income that is related to jobs, because this kind of income accounts for the bulk of the change in regional PCPI as a percent of the national average.

[Tabular Data Omitted]

Per capita income components

The formulation for assessing detailed component contributions to changes in regional PCPI as a percent of the national average has been streamlined and the presentation clarified, compared with that in the 1982 SURVEY article. The formulation is

(1) TPI/N = H/J x E/H x J/[N sub.w] x [N sub.w]/N x FI/E x TPI/FI

where TPI is total personal income, N is total population, H is hypothetical earnings, J is the number of jobs, E is earnings, [N.sub.w] is the working-age population, and FI is factor income (property income plus earnings).

When equation (1) is expressed as differences between time periods and is transformed into logarithmic form, the right-side components are additive, and the percent change on the left side is precisely equal to the sum of the percent changes on the right.

[mathematical expression omitted]

The detailed components are as follows. Industry mix (H/J): This detailed

component is the ratio of hypothetical

earnings to jobs, where hypothetical

earnings is the wages and

salaries, other labor income, and

sole proprietors’ and general partners’

income that would have originated

in a region if all jobs in

each industry in the region had

been compensated at the national

average rate in the corresponding

industry.(9) When this component

is calculated for two or more

regions, the national distribution

of earnings by industry is multiplied

by each region’s distribution

of jobs by industry. Thus, regional

differences in this component reflect

regional differences in the

distribution (mix) of jobs among

industries with varying earnings

rates nationally. The industrial

comparisons are made at approximately

the two-digit Standard Industrial

Classification (SIC) level

of detail, as in the table “Personal

Income by Major Sources”

for States, usually published in


BUSINESS. Differential regional earnings

(E/H): This detailed component

is the ratio of earnings to hypothetical

earnings, where earnings

is the actual wages and

salaries, other labor income, and

sole proprietors’ and general partners’

income originating in a region.

When this detailed component

is calculated, the region’s industry

mix of jobs is multiplied

by the region’s actual earnings by

industry in the numerator, and

the same industry mix of jobs is

multiplied by the corresponding

national earnings in the denominator.

Thus, this detailed component

reflects regional-national

differences in industrial earnings

rates that abstract, for the most

part, from those due to regional-national

differences in the industry

mix of jobs. Job ratio (J/[N.sub.w]): This detailed component

is the ratio of the count of

jobs of wage and salary workers

by place of work and of sole proprietors

and general partners to

the resident working-age population

(ages 18-64 years). Working-age ratio ([N.sub.w]/N): This detailed

component is the ratio of

the working-age population to total

population. The difference between

this ratio and unity equals

the share of population that will

be referred to as the “nonlabor

share”; any downward change in

the working-age population share

implies an increase in the nonlabor

share, and conversely. Property income ratio (FI/E): This

detailed component is the ratio of

property income plus earnings – that

is, factor income – to earnings. Transfer payments ratio (TPI/FI):

This detailed component is the

ratio of total personal income – that

is, factor income plus non-factor

income (mainly transfer

payments) – to factor income.

Contributions of the components

Table 5 shows, for each region, average annual percent changes in PCPI by detailed component for 1969-79 and 1979-88. For each region in each decade, these measures provide the basis for evaluating the contribution of the change in each detailed component to the change in PCPI as a percent of the national average. Table 5 shows TPI/FI in two versions, one designated as a place-of-work estimate and the other designated as a place-of-residence estimate. The two designations yield somewhat different estimates, but they do not materially affect the values of the specified component contributions at the regional level.

[Tabular Data Omitted]

Table 6 shows, for each region, the percentage-point difference from the national average percent change in PCPI for each detailed component. The signs of the regional-national differences may be either positive or negative. Using these differences, the procedure for accounting for convergence and divergence of the high- and low-income regions consists of two steps: (1) Subtracting the high-income regional-national difference from the low-income regional-national difference for each detailed component and (2) grouping the results of step (1) into two parts – those contributing to convergence and those contributing to divergence. For example, in the 1970’s, the results of step (1) were 0.41 for industry mix, 0.40 for differential regional earnings, 0.29 for the job ratio, -0.16 for the working-age ratio, 0.08 for the property income ratio, and -0.06 for the transfer payments ratio. When the results with positive signs are grouped, the industry mix, differential regional earnings, job ratio, and property income ratio together contributed 1.18 percentage points toward convergence. When the results with negative signs are grouped, the nonlabor share and transfer payments together contributed -0.22 percentage point toward divergence. The net result – 0.96 – indicates convergence between high- and low-income regions.

[Tabular Data Omitted]

The 1970’s and 1980’s

Of the total of the group of detailed components contributing to regional PCPI convergence in the 1970’s, (1) industry mix accounted for 35 percent, (2) differential regional earnings for 34 percent, (3) the job ratio for 25 percent, and (4) the property income ratio for 7 percent. A relative increase in the nonlabor share and a relative decrease in the transfer payments ratio in low-income regions, which benefited from falling unemployment rates, partly offset the other component contributions.

Of the total of the group of detailed components contributing to regional PCPI divergence in the 1980’s, (1) industry mix accounted for 12 percent, (2) differential regional earnings for 47 percent, and (3) the job ratio for 41 percent. Relative increases in the property income ratio and in the transfer payments ratio in low-income regions partly offset the other component contributions.

Industry mix

During the two most recent decades, industry mix contributed considerably less to PCPI convergence than during the 1940-79 period overall. Individual decade patterns have varied, but the very large industry mix contributions in the 1940’s dominated the 1940-79 period. With the onset of the revolution in farm technology, a major reallocation of redundant farm labor to other industries occurred within the context of large increases in total jobs. As noted in the 1982 SURVEY article, from 1940 to 1979, farm jobs as a percent of total jobs in low-income regions declined more than 25 percentage points; two-fifths of this decline occured in the 1940’s, when regional PCPI differences narrowed more than in any other decade. Regional differences continued to narrow in the subsequent three decades, but the reallocation of the diminishing pool of redundant farm workers appears to have accounted for no more than one-fifth of the subsequent convergence.

Much of primary commodities production remains concentrated in low-income regions, however. These regions experienced compounded relative advantages with respect to both the industry-mix and differential-regional-earnings contributions to PCPI convergence in the 1970’s. Owing to the relative price inelasticity of demand for many primary commodities, prices and incomes tend to be more volatile than jobs. In the early 1970’s, supply shocks led to soaring prices for petroleum, grain, and oilseed, and industrial hoarding led temporarily to soaring prices for copper and other industrial raw materials perceived to be in short supply. The initial supply response slowed the decades-long attrition of farm households and temporarily turned around the long-term decline in jobs in a number of mining industries in these regions. The rising earnings per job in the primary commodities industries affected both hypothetical and actual earnings per job in the 1970’s, to the advantage of the low-income regions.(10)

These same industries were disadvantaged in the 1980’s. In the Southwest and Rocky Mountain regions, weakness in these industries and in construction canceled out most of their industry-mix gains of the 1970’s. The Plains region, however, retained most of its industry-mix gains of the 1970’s mainly because high Federal Government crop payments to farm proprietors in 1988 (not characteristic of earlier in the 1980’s) boosted the region’s end-year hypothetical earnings. In the Southeast, the gains from industry mix increased modestly in the 1980’s. The Great Lakes region continued to experience negative contributions in the 1980’s, and the other high-income regions approximately regained their losses of the 1970’s.

Differential regional earnings

The relatively very large differential regional earnings contribution to PCPI divergence in the 1980’s partly reflects, as just mentioned, a reversal of 1970 developments in the primary commodities industries in the 1970’s. It also reflects substantial regional differences in earnings among industries that are more nearly uniformly distributed among regions. Rising relative unemployment rates in the Southwest, Rocky Mountain, and Plains regions and falling relative unemployment rates (and labor shortages in many labor market areas) in the New England and Mideast regions in the 1980’s had opposite effects on relative increases in wage rates. Often, the least skilled, entry-level jobs in the larger labor market areas in the latter regions commanded substantial premiums above the Federal minimum wage rate, in contrast with the conditions in the former regions. As well, contrasting phases of the construction cycle in the two groups of regions through 1988 resulted in contrasting receipts of overtime premium payments in the relatively high-wage construction industry.

The industry-mix component, as measured, does not completely filter out industry-mix differences among the regions. As an important example, nonelectrical machinery manufactures (SIC 35) includes oilfield and farm equipment (in which the Southwest and Plains regions, respectively, specialize) and computers. Employment and wage premiums in oilfield and farm equipment increased in the 1970’s and declined in the 1980’s, and employment and wage premiums in computers increased in both decades. New England was particularly advantaged through much of the 1980’s because of the increasing penetration of the market by minicomputers in which the region specializes.

The occupational composition of industries also varies among regions. Central administrative offices, in general, and corporate headquarters, in particular, tend to employ a higher proportion of professional, technical, and managerial workers than do the operating units, and these administrative units are more regionally concentrated. During the 1980’s, there was a wider divergence of job remuneration between professional, technical, and managerial workers and other job classifications than in earlier decades. This divergence particularly benefited the New England and Mideast regions, which are host to a disproportionate, though declining, share of administrative units.

That is not the whole story: Money-center banks and nonbank financial institutions in major labor market areas in high-income regions provide an example of the working of locational external economies in the 1980’s. Capital flows to and from these centers multiplied, and the proliferation of financial restructuring was the source of very large bonus payments, indeed. Junk bond packagers, arbitragers, and related corporate lawyers and tax accountants profited more than substantially in the money centers, while jobs in financial backroom” operations in low-income regions, such as those recording credit card transactions in some medium-sized cities in the Plains region, were not similarly remunerated.(11)

Job ratio and working-age ratio

The job ratio was also a major component contributor to convergence in the 1970’s and to divergence in the 1980’s. In the 1970’s, total jobs increased at almost twice the rate in low-income as in high-income regions – at average annual rates of 2.95 percent and 1.68 percent, respectively. At the same time, the increase in working-age population in low-income regions was 1.7 times that of high-income regions – at average annual rates of 2.4 percent and 1.4 percent, respectively. In the 1980’s, the average annual rates of job growth were much closer – at 2.03 percent and 1.71 percent, respectively. The increase in working-age population in low-income regions was 1.6 times that in high-income regions – at average annual rates of 1.6 percent and 1.0 percent, respectively. Thus, in the 1980’s, the job ratio turned in favor of the high-income regions, notably in New England and the Mideast, where the population growth rate remained well below the national average in both decades. (Population growth in the Southeast and Southwest regions in the 1970’s and in the Rocky Mountain and Southwest regions in the 1980’s included a higher-than-national-average nonlabor share, resulting in divergent contributions from the working-age ratio component from those regions.)

The national business cycle masks somewhat different cyclical patterns among individual regions. In the Southwest and Rocky Mountain regions, the boomlike patterns of the 1970’s continued through the mid-1980’s, when the collapse in international oil prices sent the regional economies into sudden shock: Construction, trade, and the finance-insurance-real estate group were battered, along with oilfield development and services activities. The job ratios in these regions were further depressed because working-age population immigration continued, although it trailed off when unemployment rates began to soar after the mid-1980’s.

Property income ratio and transfer payments ratio

More uniform regional distributions of property income contributed slightly to regional PCPI convergence in the 1970’s and partly offset divergence in the 1980’s. In the 1970’s, unemployment rates that were falling relative to the national average in low-income regions resulted in a small relative decline in transfer payments, and that decline partly offset regional PCPI convergence in that decade. In contrast, in the 1980’s, unemployment rates that were rising relatively in low-income regions resulted in a relative increase in transfer payments, and that increase partly offset regional PCPI divergence.

Income and Population

Effects, 1929-88

Ever since the first wave of European immigration, this Nation’s population has exhibited a much higher degree of regional mobility than that of other industrially advanced countries. Since the Nation’s early settlement, population has shifted west. In the 1929-88 period, the three westernmost regions have consistently had high relative population growth rates; the New England, Mideast, and Plains regions have had low rates. The Great Lakes region has had increasing population outmigration since the 1950’s. For about a century following the Civil War (except during the 1930’s), the Southeast had outmigration, but since the 1960’s, the region has had increasing inmigration.

When viewed in terms of comparative statics, regional population growth exerts a “drag” on relative PCPI growth, and income growth exerts a “lift,” But regional population and income growth are not independent of one another, nor are they perfectly covariant. In this section, PCPI change is decomposed into lift and drag effects – termed the “income effect” and the “population effect,” respectively – over each of six approximate decade timespans in 1929-88.

A measure of the income and population effects can be derived by differentiating PCPI with respect to time (t), as shown in equation (2).

[Mathematical Expression Ommitted]

where y is TPI, n is total population,

[delta] is the change in value over a timespan, and the subscript (-1) is the value at the initial date of a timespan. The first term on the right side of the final equality sign, [delta]y/n, hereafter called the income effect, measures the per capita increment in TPI during a timespan. The second term, Y-1/n-1 X [delta]n/n, hereafter called the population effect, is the product of the initial value of PCPI (the initial condition) and the population growth rate during the timespan. The sign of the population effect is negative, which implies a drag on PCPI change when population is growing. The initial condition thus modifies the drag of the population growth rate on PCPI growth: It mutes the drag in low-income regions and amplifies the drag in high-income regions.(12)

Table 7 shows the change in PCPI, the income effect, and the population effect and its components – initial-year PCPI and population growth rate – for each timespan in 1929-88. In each of the timespans, the absolute dollar change in PCPI in low-income regions fell below the national average, while the dollar change in the high-income regions exceeded it. When there are large differences among regions in the absolute dollar values of PCPI, it is not unusual for absolute (columns 1-6) divergence and relative (columns 7-12) convergence to occur simultaneously. This situation is especially likely when inflation accounts for a large part of income growth and affects income growth rates more or less uniformly across regions. Relative convergence of low-income regions will occur when their PCPI growth rates exceed the national average (and when those of high-income regions fall short of it). The components of PCPI behave in the same way as the total with respect to divergence when treated in absolute terms and with respect to convergence when treated in relative terms (except in the 1980’s, when, in both terms, PCPI and its components diverged regionally).

[Tabular Data Omitted]

Income and population effects compared

In each timespan except the 1930’s, the income effect provided the expected lift. In the 1930’s, PCPI declined in the Nation and in each region. Both the income and population effects were negative, except in the Southeast and Far West regions; in those regions, small positive income effects offset about one-third and one-quarter, respectively, of the negative population effects (column 1). The national population growth rate was the lowest recorded, and thus it exerted a smaller-than-average drag. The negative income effect took a greater toll on high-income regions; it accounted for more than one-half of the PCPI decline in the high-income regions, compared with less than one-third of the decline in low-income regions. In this decade of the Great Depression, the industrial Great Lakes, New England, and Mideast regions experienced population outmigration for want of jobs. The Far West, on the other hand, had a population growth rate 230 percent of the national average (column 7). Three of the four low-income regions also had greater-than-average population growth; in the Plains region, drought and dust storms ravaged farms and spurred large-scale outmigration.

World War II and the postwar conversion were powerful engines of growth during the 1940’s; the income effect swamped the population effect even though the national population growth rate exploded at the start of the baby boom (column 2). During this decade, the PCPI disparities narrowed more than in any other: PCPI more than tripled in low-income regions and more than doubled in high-income regions. The income effect was 50 times larger than the population effect in the Plains region, reflecting both huge gains in farm productivity and continued very high rates of population outmigration. The Southeast, with an income effect 21 times larger than the population effect, returned to its pre-1930’s pattern of outmigration for similar reasons.

Industrialization was taking hold in both the Plains and Southeast regions during the 1940’s. The Plains region continued exporting population during each of the subsequent decades, through good farm years and bad; the Southeast experienced a turnaround of migration (and remigration) starting in the 1960’s. The Southwest and Rocky Mountain regions experienced high to very high rates of population inmigration over each decade in the face of very uneven income effects. It was only in the two most recent decades that, with high population growth (including that in the Southeast) and with PCPI approaching the national average, the population effect in low-income regions exerted greater-than-national-average drag on PCPI growth.

The high-income regions, except the Far West, experienced an opposite population effect. The population growth rate in the Far West has averaged more than twice the national average during the decades under study. New England and the Mideast have experienced varying outmigrations – in the last two decades quite sharp, in the face of both economic reversal in the 1970’s and substantial recovery in the 1980’s. Since the 1950’s, the Great Lakes region has experienced increasing population outmigration, and its income effects have fallen below the national average since the 1960’s. By the end of the 1980’s, the Great Lakes region had fallen from above to below the national average PCPI.

Thus, although each region appears to be governed by its own culture of population migration, columns 7-12 in table 7 show a pattern of rising relative population effects in the low-income regions and declining relative population effects in the high-income regions over the decades under study. These patterns, in turn, reflect the compound effects of both the initial condition – monotonically rising (declining) initial year PCPI as a percent of the national average in the low-income (high-income) regions – and a somewhat uneven pattern of relative population growth rates.

Through mid-century, the Hicks-Borts proposition on the regional equalization of factor returns had some verisimilitude; it also had demonstrated relevance for explaining regional PCPI convergence as a percent of the national average. Since mid-century, however, modern technology has permitted industries to become increasingly footloose. Industries established in the New England, Mideast, and Great Lakes regions through the early decades of the century were based on a mechanical technology and were locationally bound to dense labor markets and by high transportation costs. The shift over time to lightweight materials, to miniaturization, to reduced numbers of moving parts in equipment, and, more generally, to the widespread substitution of electronic for mechanical processes has reduced the role of transportation costs and of large, skilled labor pool requirements in the production and distribution processes. Advances in telecommunications, more efficient transmission of power, and relatively cheaper, faster, and more convenient transportation have increasingly overcome the impedance of distance in the provision of producer services as well as in the production and distribution of goods. Overall, the economies of proximity to inputs and to market areas for manufacturing industries and for some producer services industries appear to have been weakening over the last three decades.(13)

Where there is population, there is a demand for consumer services. To that extent at least, jobs flow with population. This has clearly been the case with respect to retirement communities and recreation areas. The amenities that draw population to these areas also lure enterprises other than those connected with the provision of local consumer services, given the local cost of production relative to the prices the enterprises receive. In addition, the new industries – developed in the last three decades of rapid technological change – have tended to locate in the regions experiencing relatively rapid population growth; old industries located in the earlier established industrial belt have gone through or are currently undergoing substantial technological restructuring, and the redundant labor in these high-income or formerly high-income regions have tended to gravitate toward regions where the jobs are located. In spite of this gravitation toward jobs, it has been shown that locational external economies are associated with a degree of stickiness in factor returns in certain industries and locations (e.g., financial services in money centers) and with divergent differential regional earnings.

All the above explains much of the apparently “wrong” direction of regional population flows. However, not all job-related migration is implemented with perfect knowledge, and not all migration is in search of jobs. So, it can be assumed, regionally specific inertia in population movement and locational external economies may well continue beyond the present decade and continue to confound oversimplified theories of wages and regional PCPI convergence.

( 1). Daniel H. Garnick and Howard L. Friedenberg, “Accounting for Regional Differences in Per Capita Personal Income Growth, 1929-79,” SURVEY OF CURRENT BUSINESS 62 (September 1982): 24-34.

( 2). J.R. Hicks, The Theory of Wages (London: Macmillan, 1932).

( 3). G.H. Borts, “The Equalization of Returns and Regional Economic Growth,” American Economic Review L (June 1960): 319-47. Whereas Hicks specifically referred to real wage-rate differentials, Borts utilized available statistical measures, which were prepared in nominal terms.

( 4). In particular, housing costs increased more in the 1970’s in the low-income, fast-growing regions than in the high-income, slow-growing regions. Regional costs of living and wage-rate differentials are not independent because wages account for a large fraction of the costs of production of housing and services, which are consumed mainly in the vicinity of their production.

( 5). Intercensal estimates of regional population growth are subject to substantial revision after decennial population censuses. Any revision after the 1990 census, therefore, would also entail revisions of current estimates of regional labor force and unemployment rates and, of course, net migration rates.

( 6). Lynn E. Browne comes to this conclusion in “Shifting Regional Fortunes: The Wheel Turns,” New England Economic Review (May/June 1989): 27-40. Ms. Browne uses conceptually different and statistically more aggregate measures of “industry mix” and “earnings per job” than the equivalent measures used for this article. Also, the end year of Ms. Browne’s analysis is 1987 rather than 1988. Nonetheless, there is substantial agreement where the methodology overlaps in the two articles.

( 7). The choice of the first 3 years is based solely on data availability, the choice of the next 3 years is based on national business cycle peaks, and 1988 is the most recent year for which data are available.

( 8). Whatever the effect of the rising property income share on intraregional or intrastate differences in income distribution, interregional and interstate differences in the shares have continuously narrowed over all the decades under study. In part, this narrowing is explained by the share of population aged 65 and over relative to total State population, because property income tends to be a more important income source for this age group than for others. The simple cross-section correlation coefficient for the share of population aged 65 and over and the share of property income among States was 0.490 in 1969, 0.602 in 1979, and 0.604 in 1988 – all significant at the 1-percent level. At the same time, the coefficient of variation for the share of population aged 65 and over for all States has declined over the three decades. (This decline is explained, in part, by the higher propensity to migrate for the population aged 20-40 than for other age groups, including that for the population aged 65 and over, which has the second highest propensity to migrate.)

( 9). Estimates of the number of wage and salary workers, sole proprietors, and general partners are on a job-count basis, derived from tax and administrative records; that is, each job, whether full- or part-time, is counted (unlike a person-count estimate, usually derived from household surveys). There is no adequate basis on the subnational level for converting these job counts to a full-time equivalent measure. While the job-count estimates in this study generally agree with those in the Regional Economic Information System tables, the estimates for sole proprietors and general partners differ from current estimates in anticipation of changes to be introduced as part of the next comprehensive revisions to the State personal income estimates.

(10). Inasmuch as low-income regions have a disproportionate share of the affected primary industries, higher relative earnings rates in these industries in the 1970’s strongly affected the earnings rates used in calculating hypothetical earnings (H) as well as actual earnings (E).

(11). For a more geographically detailed discussion of the reversal of economic patterns in the 1970’s and 1980’s and the implications for regional economic theory, see Daniel H. Garnick, “Shifting Patterns in the Growth of Metropolitan and Nonmetropolitan Areas,” SURVEY OF CURRENT BUSINESS 63 (May 1983): 39-44; “Patterns of Growth in Metropolitan and Nonmetropolitan Areas: An Update,” SURVEY 65 (May 1985): 33-38; and “Growth in Metropolitan and Nonmetropolitan Areas: An Update,” SURVEY 69 (April 1989): 37-38. The methodology underlying the 1983 SURVEY article, while comprehensive, did not isolate differential regional earnings; hence, it did not identify the locational external economies pertaining to, for example, money-center financial institutions in the 1980’s (as well as in the 1950’s and 1960’s).

(12). The modifying effect of the initial PCPI value can be illustrated in relative terms. Consider two regions with initial PCPI’s of 125 percent and 75 percent of the national average, respectively. If both regions experience population growth rates equal to the national average, the high-income region will experience a drag two-thirds greater than that for the low-income region. Thus, to the extent that income growth and population growth are not covariant, the initial condition requires greater lift to keep high-income regions aloft relatively, and therein lies the tendency toward PCPI relative convergence, all else being equal.

(13). Daniel H. Garnick, Reappraising the Outlook for Northern States and Cities in the Context of U.S. Economic History, Working Paper Number 42 (Cambridge, MA: Joint Center for Urban Studies of Massachusetts institute of Technology and Harvard University, May 1978).

COPYRIGHT 1990 U.S. Government Printing Office

COPYRIGHT 2004 Gale Group