INTERkATIONAL BANK FOR RECONSTRUCTION AND DEVELOPMENT

DEVELOPMENT DEPARTMENT

Public Disclosure Authorized URBAN AND REGIONAL ECONOMICS DIVISION

URBAN AND REGIONAL REPORT NO.75-1 Public Disclosure Authorized REGIONAL LABOR FORCE PARTICIPATION AND THE DISCOURAGED WORKER

HYPOTHESIS: SOME EMPIRICAL EVIDENCE

BERTRAND RENAUD

DUC-TU PHAM Public Disclosure Authorized

These materials are for internal use only and are circulated to stimulate discussion and critical comment. Views are those of the authors and should not be interpreted as reflecting the views of the World Bank. References in publications to Reports should

Public Disclosure Authorized be cleared with the authors to protect the tentative character of these papers. REGIONAL LABOR FORCE PARTICIPATION AND THE DISCOURAGED

WORKER HYPOTHESIS: SOME EMPIRICAL EVIDEN CE

by

Bertrand RENAUD and Duc-Tu PHAM"

International Bank for Reconstruction and Development (World Bank), on leave from the University of Hawaii.

Formerly with the Bank of Vietnam. REGIONAL GROWTH AND LABOR FORCE PARTICIPATION (Abstract)

The purpose of this paper is to clarify the economic determination of the regional labor supply. In explaining the regional level of labor force participation two models are compared: the " model" based on the "discouraged worker" hypothesis and the neoclassical model of labor force participation. It is shown that a migration model complements the neoclassical model and provides an alternative interpretation of the discouraged worker hypothesis, this reinterpretation provides an explanation for the failure of the "unemployment model" at the regional level when it is quite successful at the national level. I. INTRODUCTION

Current and anticipated conditions of the labor market are always dominant concerns in the discussion of regional problems and regional poli- cies. To forecast the rate of growth of the regional labor supply it is

necessary to understand the behavior of three variables: the rate of natural population increase, labor force participation rates and net migration flows.

This paper originates in an effort to explain the aggregate labor supply in

Hawaii. It focuses on the determinants of labor force participation rates

to explain why the labor force of the state has increased at simple annual

rates of 2.8 percent during the period 1950-1970 and 3.1 percent over 1960-

1970 when the state population has increased only at average rates of 2.7 and 2.2 percent over the same period.

To explain labor supply and the behavior of labor force partici-

pation, a first model focuses on the level of the unemployment rate. It is

based on the concept of the "discouraged worker" who leaves the labor force

and stops looking actively for a when opportunities decline

and he fails to find employment. It is found that the "discouraged worker"

hypothesis does not explain well the behavior of the various age groups

which comprise the regional labor force in Hawaii. Similarly, the use of

this model at the regional level for the regions of the by

Corry and Roberts has yielded mixed results (3). These findings contrast with the uniformly successful use of the model at the national level by

Dernburg, Strand, Tella, Wachter, (4), (8), (10), (11), (13) and others.

Thus, it becomes necessary to analyze the behavior of labor force partici- pation rates at the regional level with a different model. It is also necessary to explain the poor explanatory power of the discouraged worker hypothesis and of the unemployment rate at the subnational level and to clarify the economic determination of the regional labor supply. 2

After a brief description of the Hawaii labor market, we describe the test of the "discouraged worker" hypothesis through the use of a model focusing on unemployment rates. Given the failure of this first model, the neoclassical model of labor force participation provides an alternative rationale for the determination of labor force participation. It is tested in two versions, one of which embodies again "the discouraged worker" hypothesis with little success. Then we show how a migration model comple- ments the neoclassical model, provides an alternative interpretation of the discouraged worker hypothesis and may explain its failure at the regional level when it is effective at the national level. Finally, we discuss the implications of our findings.

II. THE DATA BASE AND THE REGIDNAL LABOR MARKET

The Department of Labor and Industrial Relations of the State of Hawaii publishes monthly and yearly statistics on the total labor force but this data does not include a breakdown by age nor sex. For the analysis, yearly estimates by age-sex cohorts have been derived from a two-step pro- cedure based on yearly estimates of population by age-sex cohorts obtained by the authors previously (6), and the information on the age-sex composition of the labor force given by the U.S. Bureau of the Census for 1950, 1960 and 1970. Further adjustments were required because of differences in reporting methods between the U.S. Bureau of the Census and the State. In particular, the Census figures for Hawaii include a very large military sector which constitutes with its dependents between 10 and 15 percent of the state popu- lation depending on the year. To insure homogeneity, the data used here relates to the civilian labor force and population. Because of the procedures used to obtain the age-sex cohort estimates for the population in the 3

earlier demographic study (6), the series covers only the 21 years from 1950 to 1970. But it was a period during which a wide range of unemployment rates were experienced by the state.

As we have noted, the labor force in Hawaii has grown more rapidly than the total population. This is due to two factors: changing age-specific labor force participation rates and a changing age structure of the population. The contribution of the changing age structure of the population does not explain all the changes in the size of the labor force (6). On the other hand, the age-sex specific labor force participation rates have varied sig- nificantly according to age.

In testing the various models of this paper we make the usual dis- tinction between primary workers and secondary workers. Primary workers are defined here as males between the age of 20 and 64 and secondary workers are males between 14 and 19 6 or over 5,or females 14 years or older. This tradi- tional distinction is based on the concept of the household as a unit of labor supply, see for instance Wachter (13). It is quite important because labor force participation rates for primary workers are very high and remained stable over time while changes in the aggregate level of labor force partici- pation are most directly related to the behavior of secondary workers. The latter's participation rates have fluctuated significantly as can be seen in Table 1 where the coefficient of variation for cohorts in this group are three to four times larger than for primary workers.

III. TEST OF THE DISCOURAGED ORKER HYPOTHESIS WITH THE UNE4MLOYMENT MODEL

The first model used to explain labor force participation rates focuses on the level of the unemployment rate. Dernburg and Strand (4) TABLE 1. LABOR FORCE PARTICIPATION RATES BY AGE AND SEX (1950-1970).

Mean Standard Coefficient Value Deviation of Variation (percent) (2+1)

SECONDARY WORKERS Males 14-19 33.45 2.05 .061 Males 65+ 7.53 .75 .100

Females 14-19 17.30 2.71 .157 Females 20-24 45.93 5.07 .110 Females 25-34 39.10 5.82 .149 Females 35-44 45.24 6.30 .139 Females 45-64 39.90 7.15 .179 Females 65+ 7.53 .75 .100

PRIMARY WORKERS Males 20-24 80.63 3.21 .039 Males 25-34 85.35 3.79 .044 Males 35-44 86.32 4.11 .047 Males 45-64 78.90 4.06 .051 first developed the concept of the "discouraged worker" which relates par- ticipation in the labor force (either in aai employed or an unemployed status) to the difficulty of finding employment as reflected by the unemployment rate in the market. At the national level, they found a very significant negative relationship between the unemployment rate and labor force participation rates for both male and female workers. Their work and that of Tella (10), (11), played a significant role in the calculation of labor force estimates and in particular "hidden unemployment". The basic unemployment model used to test the discouraged worker hypothesis at the national level is: (1) (L/P) = a + al UE +a2

where the parameter a is expected to be negative.

In this first model the labor force participation rate (L/P). is measured separately for each specific age-sex cohort of the population while variable UE refers to the overall annual state unemployment rate measured by the State Department of Labor and Industrial Relations. The time trend variable is expected to reflect the significance of non-economic factors affecting labor force behavior over time. The results of the test are almost uniformly negative. The coefficient of variable UE is insignificant except in the case of male workers 14-19 and female workers 65 and over where it is significant at the 5 percent level and has the expected sign. To test for the possible existence of a non-linear response of the labor force to the magnitude of the unemployment rate, the same model with variables in logarithmic form was also tested, but yields uniformly non- 6

significant parameters for the unemployment rate.i The results are presented dn Table 2 where the values of the intercepts have been omitted to save space.

These intercepts are not particularly informative and simply reflect the average level of labor participation of each age-sex cohort which are already presented in Table 1.2

THE NEDCLASSICAL MODEL OF LABOR FORCE PARTICIPATION

An alternative model of labor force participation has been proposed

by Mincer who objects to the unemployment model and to the concept that

potential entrants into the labor force respond to job opportunities instead of the rate. What must be used is a labor supply model (in the form of labor force participation), which responds to the real wage rate.(). Then, the problem becomes one of estimating labor supply curves by age groups.

The actual supply of labor with respect to the real wage rate reflects the net effect of the trade-off between income and leisure. With a rising real wage the opportunity cost of leisure time, in terms of the goods which could be acquired with more work, increases and induces additional entry into the labor force; this is the substitution effect. But rising incomes provide workers with an opportunity to raise both their levels of consumption of goods and leisure, which is the income effect.

1/ It must be noted that the labor variable explicitly symbolized in the expression (L/P)i refers to a specific cohort while the labor variable implicit to the calculation of unemployment rates refers to a different variable which is the totel state labor force. Estimation of the model in the present form is statistically legitimate. Similarly, the esti- mation of the model in logarithmic form is valid and a rather common procedure, see Wachter (13). 2/ Complete tables including intercepts, F-values, and Durbin-Watson tests are available from the authors on request. TABLE 2. LABOR FORCE PARTICIPATION RATES 1950-1970 THE UNEMPLOYMENT MODEL FOR SECONDARY AND PRIMARY WORKERS*

Age Class Unemployment Time 2 Unemployment Time R2 Rate Trend Rate* Trend* SECONDARY WORKERS Males 14-19 -. 490 .150 .523 .002 .066 .718 (-1.79) (2.34) (.06) (4.82) Males 65+ .449 -.097 .371 -.014 -.080 .507 (1.38) (-1.45) (-.20) (-3.19) Females 14-19 .363 .433 .776 .029 .132 .497 (1.47) (7.46) (.24) (3.13) Females 20-24 1.226 .633 .380 .018 .039 .074 (1.59) (3.50) (.15) (.95) Females 25-34 .460 .945 .881 .022 .148 .674 (1.18) (10.37) (.24) (4.51) Females 35-44 .128 .994 .922 .017 .153 .799 (.38) (12.48) (.23) Females (6.18) .039 1.139 .967 .028 .212 .876 (.15) (19.34) (.39) (8.32) Females 65+ -. 100 .097 .866 .001 .123 .925 (-1.86) (7.75) (.03) (10.63) PRIMARY WORKERS Males 20-24 .393 .298 .234 -.005 .013 .104 (.724) (2.35) (-.11) (.87) Males 25-34 .746 .331 .181 -.001 .007 .030 (1.12) (2.13) (-.02) (.42) Males 35-44 .818 .413 .245 -.001 .013 .055 (1.18) (2.56) (.02) (.68) Males 45-64 .738 .444 .305 -.00004 .018 .090 (1.13) (2.90) (-.001) (.93)

*Variables in logarithmic form, t-values in parentheses. 8

In addition to being conceptually richer, the neoclassical model

of labor force participation also permits the testing of the "discouraged worker"hypothesis and can include both unemployment and wage effects. In this formulation, unemployment increases the cost of joining the labor force

to the marginal worker beyond the opportunity cost il terms of leisure which would be observed at equilibrium without unemployment. A rising unem- ployment rate will shift the labor supply downward, below its previous level,

because of the rapidly increasing cost of job search. The general form of the model is:

(2) (L/P)i = f(UE, W, T)

where W is the state real wage level obtained by dividing the total annual

private industry wage bill by the total number of employees in private industries deflated by the state consumer price index. (It is not possible to calculate cohort-specific wage rates.)

The estimates of this alternative model for all cohorts, both in

linear and logarithmic form, are reported in Table 3. They do not support

the hypothesis that labor force participation rates are jointly determined

by the real wage rate and the unemployment rate. Once again, the "dis- couraged worker" hypothesis is not supported by these state estimates. In

the linear form, the unemployment rate is weakly significant at the 10 per-

cent level of significance for the two categories of males 14-19 and females 65 and over. Its coefficient is insignificant for all other cohorts ard has

the wrong sign in seven out of twelve equations. When estimated in loga- rithmic form, the model yields uniformly negative results for the discouragod worker hypothesis. TABLE 3. NEOCLASSICAL MODEL INCLUDING UNEMPLOYMENT

Unemployment Annual Real Time 2 Unemployment Annual Real Time 2 Rate Wage Trend Rate* Wage* Trend* SECONDARY WORKERS Males 14-19 -. 354 -2.432 .621 .613 -. 006 .451 -, 039 .380 (-1.39) (-2.21) (2.81) (-.172) (3.06) (-1.84) Males 65+ .187 4.709 -1.008 .783 -. 016 .225 -. 158 .630 (1.08) (6.32) (-6.75) (-.26) (2.65) (-4.31) Females 14-19 .130 4.180 -.376 .929 .021 .661 -.096 .877 (.90) (6.71) (-3.01) (.36) (8.13) (-2.76) Females 20-24 .521 12.656 -1.817 .781 .011 .576 -1.600 .635 (1.10) (6.19) (-4.43) (.15) (5.68) (-3.67) Females 25-34 ~096 6.542 -.321 .961 017 .529 -.034 .936 (.42) (6.66) (-1.63) (.40) (9.26) (-1.42) Females 35-44 -.135 4.738 .076 .958 .013 .379 .022 .945 (-.53) (4,28) (.34) (.34) (7.44) (1.01) Females 45-64 -.138 3.177 .524 .980 .025 .420 .068 .980 (-.68) (3.61) (2.97) (.82) (10.27) (3.86) Females 65+ -.089 -.194 .134 .869 .001 .045 .107 .929 (-1.64) (-.03) (2.86) (.02) (1.02) (5.64) PRIMARY WORKERS Males 20-24 -. 041 7.795 -1.210 .613 -.006 .151 -.039 .380 (-.10) (4.54) (-3.51) (-.172) (3.06) (-1.84) Males 25-34 .171 10.324 -1.667 .657 -.003 .197 -.(j61 .411 (.39) (5.41) (-4.35) (-.07) (3.68) (-2.64) Males 35-44 .214 10.832 -1.684 .691 -. 001 .219 -. 063 .466 (.47) (5.51) (-4.27) (-.03) (4.02) (-2.70) Males 45-64 .162 10.344 -1.558 .722 -. 003 .235 -. 063 .503 (.38) (5.62) (-4.21) (-.06) (4.19) (-2.63)

*Variables in logarithmic form. 10

Clearly, in the case of this regional economy the level of unemploy-

ment does not have the explanatory power it has at the national level in the

analysis of labor force participation. A neoclassical model which does not

include the unemployment rate can be tested in the general form:

(3) (L/P)i = f (W,T)

The results are reported in natural and logarithmic form in Table 4. The

wage rate variable is highly significant for all cohorts of secondary and

primary workers. As expected, the level of explained variation R2 is higher

for secondary than for primary workers because of the smaller level of

variation in primary worker labor participation rates over time.

In addition to being superior to the unemployment model in explain-

ing the behavior of labor supply, the neoclassical model yields much richer

insights into the supply response of secondary workers to changing market

conditions. Given the fact that the household unit is the major tie between

the behavior of primary workers and secondary workers, one can first explain

the behavior of the primary wage earners and then see how the secondary wage

earners take into account the behavior of the primary income earner of their household in their adjustment to market conditions.

As mentioned earlier, the neoclassical model predicts that in a

situation of rising the substitution effect between leisure and work will be positive and the income effect negative in the case of primary workers.

The model tested in logarithmic form in Table 4 shows that the supply elas-

ticity with respect to wages for primary workers increases steadily from .151 for males 20-2 to .235 for males 45-64 indicating that the substitution

effect dominates over the income effect and that its significance increases with age. 11

The secondary workers in the household in responding to labor

market conditions will take into account the earnings of their primary

earner. While we also expect the supply of secondary workers to rise with

real wages, the magnitude of the supply response should vary according to

the individual wage level experienced by the primary worker in the household.

The income level of a household changes over the life-cycle and increases

with age because of human capital accumulated through experience and other

institutional factors. We expect a declining response with age in the form

of smaller supply elasticity coefficients with respect to a given real wage

in the case of secondary workers. This would happen because the cross-

substitution effects(between the primary worker's wage and that of the secon-

dary worker) are negative and larger for older age groups, compensating more

and more for the positive direct response to rising wages.

The empirical results for the state age cohorts confirm this a

priori expectation. The elasticity of supply with respect to real wages

declines continuously for the female cohorts from .662 for females 1h-19 to

.420 for females 45-64. In the case of young male workers the cross-

substitution effect with the primary workers of higher wages appears to

overshadow the direct effect and we have a negative supply elasticity of

-.111. Such an interpretation camot be used for workers over sixty-five

because they are considered secondary workers whatever their sex and, in

their case, the supply elasticity for male workers is lower than for female workers (.225 against .298).

The values of the time trend parameter estimates are also worth examining. First, they are all negative except for males 14-19 where the trend is positive and females 35-4 and 65+ for which there is no significant 12

trend. These results are consistent with national findings for the male

primary workers (13). The results for female workers would appear incon-

sistent with the interpretation of the time trend T as an indicator of

sociological changes of preferences generally attributed to urbanization,

, and declining fertility. Such changing preferences are inducing women to leave home and to enter the labor force. The results may be particular to Hawaii where demographic, economic and cultural factors have made it the State with the highest level of female labor force participation

6 among the fifty States. By 19 0,Hawaii had already reached the level reached nationally in 1970. Demographically, the State has had a very high

proportion of single adult women and a very low proportion of women 65 and

over. In addition, the State's industry mix is heavily concentrated in the

service industries. Living costs are considerably higher than the U.S.

average. Finally, culturally and historically there has not been a strong resistance to the participation of married women, the working wife or mother is a well established tradition among non-caucasian groups.

The results of Table 4 in logarithmic form which reports small time trends for females over 25 suggest that socio-cultural adjustments to female labor force participation may have already been made for female workers 25 and over. A downward adjustment is still operating for female workers 14 to 24, presumably because of extended schooling. Among male secondary workers the positive trend for males 14-19 is unexpected but the strong declining trend for males 65+ is consistent with national trends and the development of systems. The behavior of the primary workers is remarkably uniform and indicates only a small downward trend in labor force participation. TABLE >. NEOCLASSICAL MODEL OF LABOR FORCE PARTICIPATION

Annual Time 2 Annual Time 2 Real Wage Trend Real Wage* Trend*

SECONDARY WORKERS Males 14-19 -2.801 .741 .577 -. 111 .103 .776 (-2.51) (3.50) (-2.32) (5.66) Males 65+ 4.903 -1.072 .771 .225 -. 153 .629 (6.60) (-7.59) (2.64) (-4.71)

Feales 14-19 4.135 -.420 .926 .662 -.102 .877 (7.00) (-3.59) (8.11) (-3.28) Females 20-24 13.198 -1.994 .768 .576 -. 163 .634 (6.47) (-5.14) (5.68) (-4.21) Fenales 25-34 6.641 -. 354 .961 .529 -. 039 .935 (6.94) (-1.95) (9.23) (-1.79) Fe=ales 35-44 5.226 -- .958 .423 -- .942 (21.64) (18.47) Females 45-64 3.033 .571 .979 .420 .061 .979 (3.51) (3.48) (10.13) (3.86) Females 65+ .560 -- .763 .298 -- .793 (8.22) (8.95) PRIMARY 1'RKERS Males 20-24 7.753 -1.197 .613 .151 -. 037 .379 (4.65) (-3.77) (3.05) (-1.98) Males 25-34 10.502 -1.725 .655 .197 -. 060 .911 (5.64) (-4.88) (3.68) (-2.93) Males 35-44 11.055 -1.757 .688 .219 -. 062 .466 (5.76) (-4.82) (4.02) (-3.02) MalIs 45-64 10.153 -1.613 .721 .235 -. 063 .503 (5.86) (-4.73) (4.19) (-2.92)

*Variables are in logarithmic form. .. 14

III. THE ROLE OF UNEMPLOYMENT IN A REGIONAL ECONOMY

The results of the analysis suggest very strongly that the level

of unemployment is not a significant factor in the determination of the

supply of labor at the regional level. The lack of significance of the

unemployment variable cannot be attributed to statistical collinearity and a 3/ joint interaction between unemployment and the trend variable.- However,

the total failure of the disoouraged worker hypothesis to explain any of the

variation in labor supply except for male workers 14-19 and female workers

65+ (see Table 2) is so different from findings at the national level by

Dernburg, Strand, Tella, Wachter (4), (8), (10), (11), (13) and others that it requires further investigation.

The explanation for the lack of significance of unemployment rates may lie in the fact that national models are tested for a close economy where the only way to withdraw from the labor force is to stay at home, thus reducing the national labor force participation rate (L/P). because L declines when P remains unchanged. On the other hand, when we deal with an open state economy there is the possibility of relocating to another regional labor market under adverse circumstances. In such a case, both the numerator and the denominator of the labor force participation rate (L/P)i are adjusted downward leading to only a snall variation in the value of the rate. The estimation of a labor supply model at the regional level requires the esti- mation of net migration equations in addition to labor force participation

3/ This is clearly shown by a comparison of the two regression models (L/P)= f(UE,W) (L/P) = f(W,T) which are not reported here to save space. Results are available from the authors. 15

equations. In a separate study the.authors have estimated annual net migration flows by age group and the correlation between these flows and the Hawaii unemployment rate is high (6). Using net migration rates to eliminate the effect of varying age-group sizes and the difference between the unem- ployment rate in Hawaii and on the mainland (DDUE1 ), lagged one year to

account for delays in migration decision, we find that migration is quite sensitive to the rate of unemployment.- The model includes a popu- lation size variable P to trace out the possible effect of crowding on the desire to migrate, this variable is of minor relevance to our present con- cern. There is no reliable way of estimating migration flows by sex in Hawaii and in Table 5 we have reduced the age breakdown into 5 classes. The results show that for the most significant groups between 25 and 64 the level of unemployment is a significant factor in explaining net migration flows and that at the regional level the "discouraged worker" hypothesis may translate itself into a change in migration levels rather than a downward adjustment of labor force participation rates.

IV. CONCLUSIONS

The results of the analysis show that changes in the labor supply

through adjustments in level of labor force participation of secondary workers are much better explained by the neoclassical model. This model is definitely superior to the unemployment model which has been more commonly used, possibly because of greater data availability. The inability of the

.V The migration rates used in our analysis are the ratio of net immigra- tion (positive or negative) to total. civilian resident populAtion, 16

TABLE 5. NET MIGRATION RATES AND UNEMPLOYMENT

Age Group Intercept DDUE T P R -1

15-19 15718.0 -- 554.78 -43.03 .627 (5.23) (5.35) 20-24 -254.66 -12.48 58.15 -16.58 .836 (.88) (2.27) (-1.92) 25-44 405.25 -10.50 13.68 -1.07 .850 (-3.99) (4.65) (-5.26) 45-64 428.76 -4.55 15.96 -1.16 .787 (-1.63) (5.12) (-5.46) 65+ 504.30 -- 21.10 -1.45 .752 (7.02) (-6.37)

Source: B. Renaud, T. D. Pham [6], p. 41. 17 unemployment rate to explain labor force participation raises serious doubt

about attempts to calculate estimates of "hidden unemployment" by region. If

there is a "discouraged worker" hypothesis operating at the regional level,

its effect is certainly much weaker than the impact of regional unemployment on interregional mobility. It would be an ineffective operation to adjust regional labor force data to obtain estimates of " labor

force" for the region (see (8)). A fallacy of composition problem arises here: what is found true at the national level is found incorrect within a region which is a part of the national economy; but it can be dispelled.

The results observed for the small region of Hawaii are possible because the local economy fluctuates at a different pace from the national aggregate economy.

In addition to the immediate policy value of explaining the behavior of labor force participation rates, one reaches a better understanding of the relationship between regional growth and income convergence among regions.

As stated by Richardson, "... .labor participation rates may vary because of differences in unemployment and in the ratio of labor force to total popu- lation. In the , however, a reduction of interregional differ- ences in labor force participation rates has been a factor promoting convergence in per capita income." ((7), p.354). The neoclassical model of labor force participation provides a good explanation for the lower aggregate rates observed in low income stagnant regions. An increase in the real wage rate will draw additional secondary workers into the labor force. Thus labor force participation rates are useful indicators for interregional comparisons when data on regional roal wagn levoln are not avri I bi I o. 18

BIBLIOGRAPHY

1. Bank of Hawaii, Annual Economic Reports (3 vols.), 1968, 1969, 1972.

2. Brown, A. J. The Framework of Regional Economics in the United Kingdom, Cambridge: Cambridge University Press, 1972, pp. 205-249.

3. Corry, B. A. and J. A. Roberts. "Activity Rates and Unemployment, the U.K. Experience: Some Further Results." Applied Economics, 1974, 6, 1-21.

4. Dernburg, Thomas and Kenneth Strand. "Hidden Unemployment 1953-62: A quantitative Analysis by Age and Sex." American Economic Review, Vol. 56, March 1966, 71-95.

5. Mincer, J. "Labor Force Participation and Unemployment: A Review of Recent Evidence" in Prosperity and Unemployment, Robert A. Gordon and Margaret S. Gordon, eds. (New York, John Wiley & Sons, Inc., 1966).

6. Renaud, B. and T. D. Pham. Population Dynamics in Hawaii, Economic Research Center Monograph, University of Hawaii, Honolulu, January 1973, 143 p.

7. Richardson, Regional Economics, New York: Proeger Publishers, 1969.

8. Simler, N. J. and A. Tella. "Labor Reserves and the Philipps Curve." Review of Economics and Statistics, Vol. 50, February 1968, 32-49.

9. State of Hawaii Department of Labor and Industrial Relations, Labor Force Estimates, 1950-1970 (Annual).

10. Strand, Kenneth and Thomas Dernburg. "Cyclical Variation in Civilian Labor Force Participation," Review of Economic Statistics (Nov. 1964).

11. Tella, Alfred. "The Relation of Labor Force to Employment," Industrial and Labor Relations Review (April 1964).

12. U.S. Bureau of the Census, Census of Population 1950, Characteristics of the Population, Vol. II, Part 51-54: Territories and Possessions, Washington, D.C., 1953.

13. Wachter, Michael L. "A Labor Supply Model for Secondary Workers," The Review of Economics and Statistics, May 1972.