You are on page 1of 33

ANALYSIS AND EVALUATION OF THE FAIRNESS RESULTING FROM THE JURY SELECTION PROCESSES IN THE FEDERAL DISTRICT COURT

OF PUERTO RICO

by Elas R. Gutirrez, Ph.D. September 2001

1. INTRODUCTION This essay describes the results of our analysis of certain data relevant to a challenge of the process established by the Federal District Courts system in the District of Puerto Rico for jury selection. The challenge is based on a question of fairness. That is, on whether the process followed by the Federal Court in the District of Puerto Rico guarantees that there is an equal and unbiased chance of selection of juries, independent of certain socioeconomic characteristics. Or whether the process systematically discriminates against certain population groups on the basis of specifically prohibited characteristics. This article reports on the research performed and the results obtained. Three official data sources have been used. These are: 1) Data collected by the US Department of Commerce, Bureau of the Census, and published available in magnetic media, as part of the 1990 Census of Population; 2) a special set of detailed cross-tabulations for Puerto Rico (CPH-L-155) based on unpublished data collected for the 1990 Census of Population, produced by the Decennial Programs Coordination Branch, Population Division, US Bureau of the Census; 3) the data obtained from all questionnaires sent out by the Court, filled-out by potential jurors and returned by mail. Data from persons selected for the Master Jury Wheel (MJW) and for the Qualified Jury Wheel (QJW) were obtained through questionnaires sent out by US mail by the Court. However, the Court was unable to provide tabulations, or data previously transferred to magnetic media, on basic characteristics such as sex, age, educational attainment, occupation, or degree of English language proficiency. Therefore,
Page -1

examination of filled-out questionnaires and entry of data into magnetic media was necessary. Data on occupation or professional status is requested (in the questionnaires) in a format that made its retrieval from returned and filled-out questionnaires unfeasible. Other characteristics, such as personal or family income, are not asked from potential jurors. 2. JURY SELECTION PROCESS The jury selection process is managed in strict compliance with the provisions of the Amended Plan for the Random Selection of Grand and Petit Jurors Pursuant to the Jury Selection and Service Act of 1968, as Amended. The process begins when the Court selects citizens at random, from all registered voters from all the precincts of the Commonwealth of Puerto Rico that are comprised within the District of Puerto Rico, to serve as grand or petit jurors. The Courts Clerk will maintain what is known as a Master Jury Wheel (MJW) consisting of the names of all the persons randomly selected from the list of registered voters for the Commonwealth of Puerto Rico. The number of names is a minimum of 0.5% of the total of all names in the list. For the District of Puerto Rico there are fifty thousand (50,000) names in the MJW. The names are chosen through a stratified sampling process that strives to obtain a fair geographical representation according to reported place of residence. After obtaining the total number of names needed for maintaining the MJW, the names are grouped into sets of 2,000 names each, following procedures described in the Amended Plan. The initial sets of 2,000 names each and any additional sets of 2,000 each to be drawn from the MJW, is publicly selected, at random, by manually drawing by lot from numbered balls or cards placed in a jury drum or box. This process also follows a determined procedure described in the Amended Plan. After selection, the Clerk then prepares a list of the names in each of the sets. The Clerk mails to every person, whose name is on the sets, a Juror Qualification Form accompanied by instructions to fill out and return. The returned questionnaires go through an initial screening. Respondents not meeting certain criteria, such as US-citizenship, or the minimum-age requirement, are excluded from further consideration. These groups of candidates to jury service are further reduced in size through a process by which certain occupations or personal circumstances are basis for rejecting or excusing candidates. Discrimination for reason of sex, race, national origin, or economic

Page -2

status is explicitly forbidden.1 All those persons found to be qualified by the jury administrator at the time of the initial screening of the questionnaires are summoned to a Jury Orientation/Qualification session. In this session the Court will interview each person to ascertain his or her English language proficiency in addition to a general jury orientation. All courts proceedings are conducted in English in the Federal District Court of Puerto Rico. Proficiency in English is required by statute, i.e., 28 U.S.C. Section 1861 (2). English language proficiency must enable a potential juror to read, comprehend and satisfactorily fillout a questionnaire and to be able to exercise all functions as a juror.2 If language proficiency is determined, the person is qualified as a juror and is part of the Qualified Jury Wheel (QJW). Certain conditions, such as occupations, or personal hardships, may be raised at that point as reason for being excused from service as a juror. As stated previously, the MJW is selected through a statistical sampling procedure that is designed to assure a fair geographical representation of the population of Puerto Rico, subject to the qualifying minimum-age requirement. The MJW and the QJW should, thus, be representative samples of the Districts voting population. Participation in general elections has been historically high in Puerto Rico, usually approaching 90% of the eligible population. Thus, the MJW should fairly

Amended Plan for the Random Selection of Grand and Petit Jurors Pursuant to the Jury Selection and Service Act of 1968, as Amended, United States District Court for the District of Puerto Rico, September 1, 1996. See page 1, section 2. Discrimination Prohibited. No citizen of the United States shall be excluded from service as a grand or petit juror in this Court on account of race, color, religion, political beliefs, sex, national origin or economic status. 2 Ibid. Page 13, section 13, B. All those persons found to be qualified by the jury administrator at the time of the initial screening of the questionnaires shall be summoned to a Jury/Orientation session. In this session the Court will interview each person to ascertain his or her English language proficiency in addition to a general jury orientation. (Our district has the unique situation where English is not the primary language spoken; therefore, the Court takes this opportunity to insure that each person qualified for jury service has sufficient English language comprehension ability pursuant to 28 U.S.C. Sec. 1861(2).)
1

Ibid. Page 2, section 6, (a). Every person shall be deemed qualified to serve on grand and petit juries in this Court if he or she is a citizen of the United States, eighteen years of age or older, has resided for a period of one year within the judicial district of Puerto Rico, and is able to read, write, speak, and understand the English language with a degree of proficiency sufficient to fill out satisfactorily the juror Qualification Form and to render satisfactorily jury service in this Court.

Page -3

reflect the socioeconomic characteristics of the population of the jurisdiction above 18 years --which is the minimum voting age. 3. FUNDAMENTAL SOCIOECONOMIC CHARACTERISTICS OF PUERTO RICO Puerto Ricans are a Spanish-speaking people. Spanish and English have been recently designated official languages of the Commonwealth of Puerto Rico. However, Puerto Rico is a Spanish-speaking jurisdiction. The US Bureau of the Census has reported that, out of a total population of 3,522,000 in 1990, there were 3,219,765 persons, 5 years and over, that spoke Spanish at home.4 Approximately one-half of the Spanish speaking population 18 years and older are classified by the Census Bureau as unable to speak English.
3

Approximately one out of every four Spanish speakers in Puerto Rico are reported by the Census as being able to speak English easily. The remaining three out of four are reported as being able to speak English only with difficulty.5 Table 1 and Table 2 provide relevant data.

The World Almanac and Book of Facts 1997 reports the following salient facts about Puerto Rico. Geography: Total area: 3,508 sq. mi. Land area: 3,427 sq. mi. Population (1994 estimate): 3,801,977 (and about 2.7 million more Puerto Ricans reside in the mainland US). Population density: (1990) 1,035 per sq. mile. Urban population (1990): 66.8%. Ethnic distribution of the population: (1990): 99.9% Hispanic, with a primarily Hispanic culture. Language: On Jan. 28, 1993, the Government of Puerto Rico declared Spanish and English joint official languages. Federal government civilian employment: (in 1992): 10,000.
3 4 US Bureau of the Census, 1990 Census of Population and Housing, Summary Tape File part 3A and unpublished data. 5 Persons 5 years and over who reported that they spoke a language other than English in question 15a of the 1990 Census Form were also asked in question 15c to indicate their ability to speak English based on one of the following categories: Very well, Well, or Not at all. The data on ability to speak English represents the persons own perception about his or her own ability or, because census questionnaires are usually completed by one household member, the responses may represent the perception of another household member. The instructions that were mailed to households did not include any information on how to interpret the response categories in question 15c. See the 1990 Census of Population Guide, Part B. Glossary, page 42. Moreover, the Guide explains goes further to warn on the limitations of the data, by stating that: Persons who speak a language other than English at home may have first learned that language at school. However, these persons would be expected to indicate that they spoke English Very well. Persons who speak a language other than English, but do not do so at home, should have been reported as not speaking a language other than English at home. The extreme detail in which language names were coded may give a false impression of the linguistic precision of these data. The names used by speakers of a language to identify it may reflect ethnic, geographic, or political affiliations and do not necessarily respect linguistic distinctions. The categories shown in the tabulations were chosen on a number of criteria, such as information about the number of speakers of each language that might be expected in a sample of the United States population.

Page -4

Table 1 Spanish Speakers' Ability to Speak English Age Bracket Easily With Difficulty
5 to 17 18 to 64 Over 64 Total 105,850 581,134 60,496 747,480 162,334 531,680 68,020 762,034

Unable 565,317 884,383 10,445 1,460,145

Total 851,614 2,026,092 342,059 2,969,659

Table 2 Spanish Speakers' Ability to Speak English Age Bracket Easily With Difficulty 5 to 17 3.6% 5.5% 18 to 64 19.6% 17.9% Over 64 2.0% 2.3% Total % 25.2% 25.7%

Unable 19.0% 29.8% 0.4% 49.2%

Total % 28.7% 68.2% 11.5% 100.0%

The ability to speak English is likely to depend on the following factors: 1) the parental or language spoken at home; 2) educational attainment measured in terms of the number of completed school years; and, 3) years of residence in an English-speaking jurisdiction. The US Census classifies the population of Puerto Rico according to place of residence. The basic units of classification are Puerto Ricos legal subdivisions, i.e., municipios. The concept is closer to the county used in the US mainland. Puerto Rico has 78 municipalities. Municipalities differ widely with respect to the prevailing socioeconomic conditions. The number of inhabitants also varies between municipios. Municipios are thus, heterogeneous social, demographic, and economic geographic subdivisions. However, the US Census Bureau has subdivided the Puerto Ricos geographical area into smaller subdivisions denominated as census tracts. Census tracts are defined to represent a more homogeneous population group in terms of its socioeconomic characteristics. Given Census Bureau methodology, census tract population totals vary within a narrow range. The population within a census tract varies in number from approximately 2,000 to approximately 8,000 resident persons. In certain municipios, especially those that are less densely populated, the necessary density conditions set by the Census Bureau to establish census tract subdivisions are not found. In such cases, the municipio is subdivided into smaller subdivisions denominated as blocks. Groups of blocks are called block numbering areas. Blocknumbering areas are used instead of census tracts to perform analysis

Page -5

with geographically distributed Census data, whenever census tracts have not been defined. Socioeconomic conditions vary within and between municipios. Socioeconomic conditions vary to a higher degree between than within census tracts because of the way they are defined by Census Bureau methodology. Income is a key socioeconomic variable. There are several measures of the typical income value within a geographic subdivision. Per capita income, i.e., total personal income divided by the number of resident population, is a recognized indicator of living standards and economic status. Although widely used, this measure is equivalent to an arithmetic mean. Arithmetic mean values are more sensitive to extreme values within a distribution than other measures such as the median or the geometric mean. The median is particularly useful in cases where data have been grouped in classes. Some of these groups may be open classes. For example: less than, or more than, a particular number. This is how the distribution of income is usually presented. 4. STATISTICAL DESCRIPTION AND ANALYTICAL RESULTS 4.1. THE WORKING HYPOTHESIS

The jury selection process must guarantee against any systematic socioeconomic bias. Statistical methods make use of systematic procedures stated in terms of hypothesis that are then subjected to tests. The working hypothesis tested here is that the jury selection process followed by the Federal District Court in Puerto Rico does not incorporate a systematic bias that loads the potential jury pools against any population group in terms of characteristics specifically defined as discriminatory and, thus, prohibited. Specifically, that the English language proficiency requirement does not result in a biased, or loaded, pool from which potential jurors are then chosen. 4.2. THE MODEL

In general, a model specifies a priori relationships between certain variables. These relationships are derived from the logic of some theory, or behavioral construct of reality, i.e., or a model. A model strives to predict or explain the occurrence of values in a dependent variable, subject to values for other explanatory variables. If a theory states that there is a linear dependence between a set of variables, then the expected value, or conditional probability, of variable y, is subject to the values of variables x1, x2 and xn, and can be mathematically expressed by a linear regression model. Equation (1) is such an expression. (1) E (y|x1, x2,, xn) = + 1 x1 + 2 x2++ n xn
Page -6

Equation (1) is a conditional probability statement. The distribution of values of a random variable y is expressed as conditional to the values of other explanatory random variables. If values for the variables in question can be observed (measured), the model can then be tested for statistical significance using sample data. Often used statistical methods have been developed for such purpose. Income differences are an indicator of socioeconomic differences. US Census data from the 803 census tracts or block-numbering-areas (BNA) that constitute the total census sub-divisions used during the 1990 Census of Puerto Rico have been obtained. Differences in income per capita, in income per capita of persons in households net of those persons living in quarters-- and in median family income, were observed among these comparable and approximately homogenous sub-divisions. Let: Yi = income per capita for resident population in census tract i; Eei = population 18 and older resident able to speak English easily as a proportion of total population in census tract i; Edi = population 18 and older resident able to speak English with difficulty as a proportion of total population in census tract i; uI = a non-observable residual or disturbance term; i = identify observations for a census tract or blocknumbering area; and finally, (i=1,2,,803);

, 1, and 2 are coefficients.


Then the model can be specified as the regression equation (2): (2) Yi = + 1 Eei + 2 Edi + ui Equation (2) is a linear model that states that differences in income per capita observed between census tracts can be explained by differences in the proportion the of population of age that is able to speak English with ease and with difficulty. The values for the coefficients of this model can be estimated by applying the method of ordinary least-squares (OLS). Through this
Page -7

method, quantitative estimates can be obtained for the coefficients of equation (2). Equation (3) below, expresses equation (2) in terms of OLS bi coefficients. These coefficients are the estimators of i coefficients. (3) Yi =a + b1 Eei + b2 Edi + ei The error term (ei) accounts for differences between values of Yi calculated by the quantified model, compared with the observed corresponding value. This can be expressed mathematically as follows by equation (3a): (3a) ei = Yi (a + b1 Eei + b2 Edi) The method of OLS generates estimates for the regression coefficients that assure that the sum of the errors squared is a minimum. The sun of the square of errors is used because errors are both positive and negative deviations around the observed values. Different values for the coefficients would yield different sums of squared errors. Thus, in mathematical terms the sum of squared errors is a function of the estimates of the regression coefficients. The method of OLS calculates values for the coefficients assuring that Minimum{ a,, b1, b2}

=1
I

ei2.=

The working (null) hypothesis to be tested can be expressed as follows.

0:

, = 1 = 2 = 0;

Where: The coefficients a, b1, and b2 are the OLS estimates of the true coefficients , 1, and 2 . The estimation procedures carried out with census data, and the tests for statistical significance of the estimated coefficients, yield results that are not supportive of the null hypothesis. In statistical terms the alternative hypothesis is expressed as follows.

1:

, 1 2 0.

Given the results obtained, the null hypothesis is rejected while the alternative hypothesis is maintained, i.e., on the basis of observed data, the coefficients are statistically different from zero and a linear

Page -8

relationship explaining economic status in terms of English language abilities of the population cannot be attributed to chance. The language of statistics in this instance is much like the constitutional requirement known as the presumption of innocence, i.e., an accused is innocent until proven guilty beyond a reasonable doubt. In other words, a theory (the accused) stands (innocent) until observed data provides sufficient grounds (evidence) to permit the conclusion that whatever relationship between the variables stated by the theory is not merely explainable as due to chance ( reasonable doubt). In this particular case there is a linear relationship that explains a statistically significant portion of the observed variation in income per capita across census tracts in Puerto Rico. This linear relationship is so strong that it cannot be concluded that the relationship is merely due to chance. Therefore the working hypothesis (i.e., the null hypothesis) must be rejected. The alternative hypothesis is thus sustained by census data across census tracts. It is concluded that the ability of the population to speak English (easily and with a degree of difficulty) explains differences in economic status measured in terms of income per capita. There are other measures of economic status. These include median family income and per capita income of persons living in households excluding those that are institutionalized. The model was estimated using census tract data on median family income and per capita income for persons not living in quarters. The results obtained are consistent with and confirm those obtained using per capita income. Certain factors determine the degree of language proficiency that is attained by a population. Educational attainment is obviously a primary factor explaining proficiency in any second language. Due to labor market realities, English language proficiency translates into differences in economic status. Labor market conditions differ for men and women. In order to explore this hypothesis, gender differences have been introduced explicitly in the model. Thus, the linear relationship observed between economic status and English language proficiency has been further refined. The refinement consists in taking gender differences in English language proficiency into account. The results obtained when gender was used to refine the model are now discussed. The English language ability variable (measured in terms of the proportions of population 18 years and older) has been decomposed into
Page -9

the male and female populations. Equation (4) expresses the model with appropriate use of mnemonics. (4) Yi =a + b1 Eei + b2 Edi + ei Where: Yi = median family income for census tract i;

Eei = resident population 18 and older able to speak English easily as a proportion of total population in census tract i; Edi = resident population 18 and older able to speak English with difficulty as a proportion of total population in census tract i; The results of the econometric analysis are presented in three summary outputs. These contain key regression analysis statistics. The summary outputs are divided into three sections each. The first section lists statistics used to evaluate the goodness of fit of the regression hyper plane. In other words, how well the model fits the statistical data. This section also provides information as to what percentage of total variation is explained by the model. The second part of each output presents key values related to the analysis of variance used to establish whether the model passes a statistical significance test. That is, whether the results can be sustained at some pre-established level of probability and are not merely explained by chance. The pre-established level of confidence for this analysis is 95%. The F-statistic provides a test of significance for the model. The third section of the summary outputs provides the (bi) coefficients for the multiple regression equations and the calculated t-statistics for these coefficients. The calculated t-statistics, which relate the coefficients to their respective standard errors, provide a test of significance for each variable in the equation. At 95% of required significance most coefficients pass the significance test. A number are statistically significant even the very stringent 99% confidence level.

Page -10

Explanatory variables are identified in the summary outputs by labels. The variable mnemonics and labels used to estimate the regression coefficients for equation 4 are listed below. Eei = ENG18+A/POP. Edi = ENG18+C/POP.

SUMMARY OUTPUT Median Family Income in 1989 (P084A001) Regression Statistics Multiple R 0.708 R Square 0.502 Adjusted R Square 0.500 Standard Error 4246.882 Observations 803 ANOVA df Regression Residual Total SS 2 14,517,177,305.67 800 14,428,803,507.76 802 28,945,980,813.43 Standard Error 536.1 1,167.0 2,907.8 MS 7,258,588,652.83 18,036,004.38 F Significance F 402.45 1.1398E-121

Intercept ENG18+A/POP ENG18+C/POP

Coefficients 3,731.0 31,869.3 5,706.4

t Stat 6.96 27.31 1.96

The econometric results robustly support the model. The adjusted coefficient of determination, denoted by R2 (Adjusted R Square in the summary outputs) is a measure commonly used to describe how well the sample regression line (or hyper plane) fits the observed data. This coefficient cannot be negative or greater than one, i.e., 0 R2 1. It can be interpreted also as the portion of total variation in the dependent variable explained by the model. Equation 4 fitted to census tract data explains 50% of total variation in median family income. Furthermore, all regression coefficients have the correct signs and are found to be statistically significant in light of the calculated t statistics.

Page -11

SUMMARY OUTPUT Median Family Income in 1989 (P084A001) Regression Statistics Multiple R 0.825 R Square 0.681 Adjusted R Square 0.680 Standard Error 3397.506 Observations 803 ANOVA df Regression Residual Total SS 3 19,723,083,605.93 799 9,222,897,207.50 802 28,945,980,813.43 Standard Error 924.6 2,846.7 4,303.4 3,677.3 MS 6,574,361,201.98 11,543,050.32 F 569.55 Significance F 6.769665E-198

Intercept ENG18+A_F/POP ENG18+C_F/POP UN_ENG18+_M/POP

Coefficients 6,886.3 72,283.2 -16,652.1 -8,064.3

t Stat 7.45 25.39 -3.87 -2.19

The model was further refined to account for gender differences and labor market realities. Equation (5) completes the specification of the model. (5) Yi =a + b1 FEei + b2 FEdi + b3 MEUi + ei Where: FEei = resident female population 18 and older able to speak English easily as a proportion of total population in census tract i; FEdi = resident female population 18 and older able to speak English with difficulty as a proportion of total population in census tract i; MEUi = resident male population 18 and older unable to speak English as a proportion of total population in census tract i; Once more, it is noted that explanatory variables are identified in the summary outputs by labels. The variable mnemonics and labels used to estimate the regression coefficients for equation 5 are listed below. FEei = ENG18+A_F/POP; FEdi = ENG18+C_F/POP; MEUi = UN_ENG18+_M/POP Equation 5 represents a marked improvement in terms of explanatory
Page -12

power over equation 4. When gender differences, and a variable incorporating the proportion of the male population unable to speak English, are explicitly introduced, and the equation is fitted to census tract data, the model explains 68% of total variation in median family income. Furthermore, all regression coefficients have the correct signs are found to be statistically significant in light of the calculated t statistics. There is no doubt that census data demonstrates that differences in economic status are dependent on the ability to speak English easily. Furthermore, as the proportion of males that are unable to speak English increases, family incomes decrease across census tracts in Puerto Rico.
SUMMARY OUTPUT Population 18 years and over able to speak English easily, as percent of total population Regression Statistics Multiple R 0.945 R Square 0.893 Adjusted R Square 0.891 Standard Error 0.043 Observations 803 ANOVA df Regression Residual Total 16 786 802 SS 12.34 1.47 13.82 MS 0.77 0.00 F Significance F 411.37 0

Intercept P0460001/POP P0460002/POP P0460003/POP P0460004/POP P0460005/POP P0460006/POP P0460007/POP P0460008/POP P0460009/POP P0460010/POP P0460011/POP P0460012/POP P0460013/POP P0460014/POP P0460015/POP P0460016/POP

Coefficients Standard Error -0.2 0.0 0.0 0.1 0.1 0.1 1.0 0.0 1.0 0.1 1.0 0.1 0.5 0.2 2.0 0.1 2.0 0.2 0.0 0.1 0.6 0.1 0.6 0.1 0.2 0.1 0.4 0.1 0.7 0.2 0.5 0.1 1.0 0.2

t Stat -9.43 0.27 1.31 22.73 15.84 17.48 2.78 15.91 10.87 0.26 5.31 5.02 1.97 3.88 4.81 3.85 5.16

Page -13

Educational attainment and English language ability are strongly correlated. This is demonstrated by a regression fit that tests such a model. Equation (6) specifies the model. The dependent variable is the resident population, 18 and older, that are able to speak English easily as a proportion of the total population in census tract i. (6) Eei = a + b1 SAm1i + b2 SAm2i + + b8 SAm8i + + b f16 + e 16 SA i i Where: SAmji = male population 18 years and older that have completed j years of school completed (mj = 1,2,8 and mf = 9,10, , 16) as a proportion of population in census tract i; SAfji = female population 18 years and older that have completed j years of school completed (mf = 9,10, , 16) as a proportion of population in census tract i. Eight levels of educational achievement are used by the Census Bureau to classify the male and female population groups across census tracts. Less than 7th grade 4th to 9th grade 10th to 12th grade, no diploma High school graduate (includes equivalency) Some college, no degree Associate degree Bachelor's degree Graduate or professional degree

+ b9SAf9i+

Following the same terminology used previously to explain the results of the regression analysis for equations 4 and 5 equation, the following can be said. The ability of the population to speak English is dependent on achieving a certain level of educational attainment.
Page -14

The equation has been fitted to census tract data. The model explains 89% of total variation in the ability to speak English easily. Regression coefficients have the correct signs and are found to be statistically significant in light of the calculated t statistics, except for the first two levels for males and the first level for females. There is no doubt that census data demonstrates that differences in economic status are dependent on the ability to speak English easily. Furthermore, as the proportion of males that are unable to speak English increases, family incomes decrease across census tracts in Puerto Rico. The ability to speak English is predominantly correlated by educational attainment. 5. SOCIOECONOMIC CHARACTERISTICS OF THE POPULATION OF PUERTO RICO There is a substantial gap in per capita income between Puerto Rico and the mainland USA. Adjusted for variations in the cost of living between the two, the per capita household income of Puerto Rico in 1990 was 44.3 percent that of the US. That is, US per capita income is 2.26 times Puerto Ricos. The per capita household income adjusted by cost of living differences of Puerto Ricans on the island was 74.4 per cent that of Puerto Ricans on the mainland in 1990. That is, real per capita household income of mainland Puerto Ricans is 1.34 times that of Puerto Ricans residing in this District. The degree of income inequality in Puerto Rico is very high. The poorest 40 percent of all families on the island received only 7.5 percent of all family income in the island in 1989. In contrast, in the US, the poorest 40 percent of the families received 15.2 percent of the nations family income. The extent of inequality in Puerto Rico in 1989 was closer to that of developing countries like Brazil, where the poorest 40 percent of the population received only 7.8 percent of all income, or Panama, where the corresponding figure was 8.3 percent. Inequality also appears to have risen sharply in Puerto Rico since the islands early period of economic growth in the 1950s and 1960s. The type of household a person lives in has an effect on that persons socioeconomic status. The population living in households headed by women (no spouse present) confronted the worse economic conditions, having an average per capita household income of $2,846 in 1990. With 30 percent of all women 15 years of age or older living in female-headed households, gender constitutes a significant force in the islands income distribution. Indeed, poverty rates in 1990 were substantially higher for

Page -15

women than for men: close to 70 percent of Puerto Rican women were living in poverty in 1990, compared to 52 percent of men. See Table 3 below.

Table 3 Selected Socio-economic Indicators: Puerto Rico and U.S.: 1990 Indicator PR US Number of Families 889,998 65,049,428 Number of Families Under Poverty Line 492,025 6,487,515 Percent of Families Under Poverty Line 55% 10%

There is deep regional inequality in Puerto Rico. The gap between the income of the municipios with the highest and the lowest per capita income in 1989 was substantial: Guaynabo had close to four times the income per capita of Adjuntas. This inequality follows closely along metropolitan non-metropolitan areas. The economic conditions of these municipios are declining and they have comparatively high unemployment rates. The proportion of persons living below the poverty income line6 (the poverty rate) declined both between 1970 and 1980, and between 1980 and 1990. In 1970, 62.8 percent of the population lived in families with income below the poverty level. In 1980, 59.8 percent was below the poverty level. By 1990, the poverty rate had dropped to 57.3 percent. To a large degree, federal transfer payments to Puerto Rico explain this decline in poverty indicators. The highest poverty rates in Puerto Rico are observed among the youngest and oldest populations. For those under 25 years of age, the poverty rate in 1990 was 70 percent. In contrast, among persons 46-55, the poverty rate was 49 percent. For persons over 75, it was 69 percent. Because of these differences in poverty rates by age group, demographic changes can strongly influence the overall poverty rate. Among Puerto Ricans the age distribution has shifted significantly toward the middle range of the age distribution. This may have contributed to a drop in overall poverty rates in the 1970s and 1980s.
6 The poverty threshold for a family of four persons (two adults and two related children) in 1989 was $12,674. These thresholds do not take into account the differences in cost of living between the mainland and Puerto Rico. For this reason, the incidence of poverty in Puerto Rico may not be accurately measured by the census data.

Page -16

Table 4 compares family income distribution in Puerto Rico with the distribution of the US families. Median family incomes are also presented. Tabulations are based on Census data.

Table 4

Family Income Distribution:


Rico & United States 1989 Income Class
Less than $5,000 $5,000 to $9,999 $10,000 to $14,999 $15,000 to $24,999 $25,000 or more Median Family Income

Puerto
PR 25% 25% 17% 18% 16% 9,988 $ US 4% 6% 7% 16% 67% 35,225 Ratio PR/US 6.28 4.45 2.29 1.07 0.24 0.28

Percent Family Income Distribution by Income Class Puerto Rico and United States 1989

70.0 60.0 50.0 40.0 30.0 20.0 10.0 0.0 Puerto Rico United States
$25,000 or more $10,000 to $14,999 $15,000 to $24,999 $5,000 to $9,999 Less than $5,000

Percent of Families

Income Class

Page -17

6. CHARACTERISTICS OF THE POPULATION OF PUERTO RICO BY ABILITY TO SPEAK ENGLISH A set of special cross-tabulations produced by the Census Bureau provide strong evidence of the observable correlation between the degree of English language abilities and economic status of the population of Puerto Rico. 7 The Decennial Programs Coordination Branch, Population Division, US Bureau of the Census prepared cross-tabulation CPH-L155. These are included as an Appendix to this memorandum. Tables 7 and 8 of CPH-L-155 provide a statistical basis to analyze social and economic characteristics of the population of Puerto Rico, by ability to speak Spanish and English.8 Table 7 of CPH-L-155 cross-tabulates several social characteristics, i.e., age, household type and kinship, residence in 1985, school enrollment status, type of school, educational attainment and ability to read and write. Table 8 of CPH-L-155 cross-tabulates labor force status, occupation, work status, earnings and income of persons 5 years and over, poverty status and income below poverty level (for males and females). Both tables organize the data using the following taxonomy. Persons in households All Households Members Speak Spanish Total One of more members Speak English Households in which one or more persons are unable to speak English Total One or more members Speak English Persons able to speak Spanish Total Speak English Easily (SEE) Speak English with Difficulty (SEWD) Unable to Speak English (USE)

US Bureau of the Census, Decennial Programs Coordination Branch Population Division, 1990 Census Detailed Cross-tabulations for Puerto Rico, CPH-L-155, Washington, DC, August 1994 8 These tables must not be confused with Tables 7 and 8 of this memorandum.
7

Page -18

Unable to Speak Spanish Total Speak English Easily (SEE) Speak English with Difficulty (SEWD) Unable to Speak English (USE) 6.1. RATIO ANALYSIS

There are differences in economic status between groups of the population of Puerto Rico when these are classified according to ability to speak English. See Tables 5 and 6.

Table 5 Economic Pueto Rico 1989 Males 15 years & over Total With no income With Income Median Income Mean Income

Speak English Easily With Difficulty Unable 348,368 306,637 542,828 66,620 70,196 152,633 281,748 236,441 390,195 $ 10,599 $ 7,147 $ 4,778 $ 16,417 $ 8,315 $ 6,012

Total 1,197,833 289,449 908,384 $ 6,717 $ 10,099

Table 6 Economic Characteristics by Ability to Speak Spanish & English Pueto Rico 1989
Females 15 years & over Total Wlith no income With Income Median Income Mean Income

Speak English Easily With Difficulty 333,470 345,171 109,380 126,782 224,090 218,389 $ 7,665 $ 5,516 $ $ 10,034 $ 6,759 $

Unable 647,174 255,032 392,142 2,987 4,090

$ $

Total 1,325,815 491,194 834,621 4,285 6,384

Page -19

Median and Mean Income by Ability to Speak English Males 15 Years and Over Puerto Rico 1989

$18,000 $16,000 $14,000 $12,000 $10,000 $8,000 $6,000 $4,000 $2,000 $Median Income Mean Income Easily With Difficulty Unable

Educational attainment, occupational, and income differences between Spanish speaking persons, unable to speak English, and those able to speak English easily, or with difficulty, were analyzed. Analysis, based on ratios, yields significant differences between population groups.

Income

Page -20

Median and Mean Income by Ability to Speak English Females 15 Years and Over Puerto Rico 1989

$10,000

$8,000

Easily With Difficulty Unable

Income

$6,000

$4,000

$2,000

$Median Income Mean Income

For each variable (and its sub-categories) two (2) ratios were calculated: 1) the number of persons that speak English easily, divided by the number of persons that speak English with difficulty (SEE/SEWD); and 2) the number of persons that speak English easily, divided by the number of persons unable to speak English (SEE/USE). The ratios thus obtained are pure numbers, i.e., they have no dimension. This mathematical property permits comparisons independent of units of measurement. 6.1.1. Educational Attainment The ratios obtained for educational attainment are presented in Table 7 of this memorandum.

Page -21

TABLE 7 RATIO ANALYSIS OF EDUCATIONAL ATTAINMENT


EDUCATIONAL ATTAINMENT 1 2 3 4 5 6 7 8 9 10 11 LESS THAN 5
TH

SEE/SEWD

SEE/USE

GRADE

0.41 0.49 0.81 0.84 1.33 1.16 1.38 2.09 4.08 4.90 7.32

0.05 0.16 0.42 0.68 2.13 1.90 2.84 6.58 19.22 14.7 27.97

5TH GRADE TO 8TH GRADE 9


TH

GRADE TO 12

TH

GRADE, NO DIPLOMA

HIGH SCHOOL GRADUATE SOME COLLEGE, NO DEGREE ASSOCIATE DEGREE, OCCUPATIONAL PROGRAM ASSOCIATE DEGREE, ACADEMIC PROGRAM BACHELORS DEGREE MASTERS DEGREE PROFESSIONAL SCHOOL DEGREE DOCTORATE DEGREE

A clear pattern is identifiable. There is a direct and positive correlation is clear between higher levels of educational attainment and ability to speak English easily. Both, (SEE/SEWD) and (SEE/USE), ratios increase in direct correlation to educational attainment. For instance, for every person that speaks English with difficulty and has attained a Masters Degree, there are approximately four (4) persons that have attained the degree but speak English easily. On the other hand, for every person that is unable to speak English and has attained a Masters Degree, there are more than nineteen (19) persons that have attained that degree but speak English easily. 6.1.2. Income Calculated income memorandum. ratios are presented in Table 8 of this

Page -22

TABLE 8 RATIO ANALYSIS OF INCOME


INCOME CATEGORY NO INCOME WITH INCOME $1 TO $199 OR LOSS $1,000 TO $1,999 $2,000 TO 2,999 $3,000 TO 3,999 $4,000 TO $5,999 $6,000 TO $7,999 $8,000 TO $9,999 $10,000 TO $14,999 $15,000 TO $24,999 $25,000 TO $49,999 $50,000 OR MORE SEE/SEWD SEE/USE

1 2 3 4 5 6 7 8 9 10 11 12 13

0.89 1.11 0.81 0.78 0.72 0.79 0.79 0.88 0.98 1.21 1.99 4.08 7.69

0.43 0.65 0.34 0.27 0.24 0.29 0.39 0.56 0.83 1.63 3.67 9.87 14.28

Inspection of Table 8 above yields a similar pattern. Higher income categories are correlated with higher values for ratios (SEE/SEWD) and (SEE/USE). For example, there are twice (1.99) as many persons that have incomes between $15,000 and $25,000 per year that speak English easily, than those with the same income but can speak English only with difficulty. This ratio increases to 7.7 at the $50,000 or more income category. At that income category the ratio of persons that speak English easily to those that are unable to speak English reaches 14.3. 6.1.3. Occupation The ratios calculated for occupational categories are presented in Table 9. TABLE 9 RATIO ANALYSIS OF OCCUPATION
OCCUPATION (Broad Categories) Managerial and professional Technical, sales, and administrative support Service Farming, forestry and fishing Precision, production, craft and repair Operators, fabricators and laborers SEE/SEWD SEE/USE

2.02 1.77 0.85 0.87 0.59 0.72

4.35 4.30 0.47 0.69 0.17 0.36

Page -23

Inspection of Table 9 yields the same pattern observed previously. That is, higher skill levels are positively correlated with the ability to speak English easily. The broad occupational categories that appear in Table 9 have been sorted in descending order according to likely earnings. The ratios exhibit a consistent trend with those previously discussed. Table 8 of CPH-L-155 (not to be confused with table 8 of this memorandum) presents data on mean earnings for males and females 16 years and over (of labor force age) for 1989. The mean earnings of males that speak English easily are reported as $17,737. For those that speak English only with difficulty mean earnings drop to $10,233. For males unable to speak English, mean earnings were reported as $7,055. Or 39.7% of mean earnings for those that could speak English easily. The mean earnings for women 16 years and over that speak English easily are reported as $11,733. For those that speak English only with difficulty mean earnings drop to $8,426. For females unable to speak English, mean earnings were reported as $6,234. Or 53.1% of mean earnings for those that could speak English easily. 6.2. FINDINGS DERIVED FROM RATIO ANALYSIS In general, persons that speak English easily have higher incomes than do persons unable to speak English, or than those persons that speak English with difficulty. Higher SEE/SEWD and SEE/USE ratios are strongly associated with higher educational attainment, higher hierarchy in the occupational structure, and higher income levels. Ratio analysis clearly demonstrates that there are significant differences among population groups with respect to three economic status indicators, i.e., educational attainment, occupation, and income. The distinguishing characteristic defining these population groups is the ability to speak English. In conclusion, ratio analysis performed on data from the special cross-tabulation obtained from the Census Bureau yields a significant and pervasive pattern that is clearly identifiable. All economic status indicators examined and analyzed are directly related to the ability to speak English.

Page -24

6.3. LORENZ CURVE ANALYSIS Lorenz curves are useful for income distribution analysis. For a view of inequality with respect to all income groups, the cumulative distribution of income is usually plotted as the Lorenz curve. The larger the area included between the straight diagonal line of perfect equality and the curves describing the cumulative distribution of income, the higher the degree of inequality. The analysis is illustrated graphically by plotting Lorenz curves for the male and female Spanish speaking population groups, classified according to English language ability. Again, the same classification has been used. Namely, able to speak English easily, able to speak English with difficulty and unable to speak English. The special cross-tabulation of census data for Puerto Rico, prepared by the Office of the Census, through Lorenz curve analysis, provides irrefutable evidence of the strong correlation between English language ability and economic status in Puerto Rico. These data evidences the inequality of the income distribution in Puerto Rico. Furthermore the data shows the inverse relationship between English language ability and income distribution inequality. The data also shows that gender intensifies the inequality. The distribution of income for the Spanish-speaking population of males is unequal to a high degree. For those that speak English easily, 7.2% of their income is received by 50% of this population group. For those that speak English with difficulty, 7.2% of their income is received by 74.7%. For those that are unable to speak English, 7.2% of their income is received by 82.2%. The case for women is extreme. The distribution of income for the Spanish-speaking population of females is such that, for those that speak English easily, 8.9% of their income is received 73.3% of this population group. For those that speak English with difficulty, 8.9% of the income is received by 84.7%. For those that are unable to speak English, 8.9% of their income is received 95.8%. Thus, not only are living standards, measured by per capita or median income are less than US averages, but the distribution of income is biased against women and against those not having English language abilities. When both characteristics coincide, the negative effects are compounded. So strong is this correlation that the English language requirement undoubtedly biases the jury pools in Puerto Rico by income and, thus, by economic status. In fact, the English language

Page -25

requirement acts like a filter against the lower income strata of the population. The two graphs presented in this page and the next illustrate how income is distributed for men and women according to 1990 Census data. Each graph illustrates how the lack of English language ability has the effect of worsening the distributions.
Income Distribution of Spanish Speaking Female Population 15 and Over by Ability to Speak English Puerto Rico 1989
100.0% 90.0% 80.0% 70.0% 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0% 0.00% Easily Difficulty Equality Unable

Percent of Population Group

10.00%

20.00%

30.00%

40.00%

50.00%

60.00%

70.00%

80.00%

90.00%

100.00%

Percent of Income

On the basis of the official US Census Bureau data, it must be concluded, that the English language proficiency requirement systematically biases the jury pools towards the higher brackets of the income distribution and, thus, of the socioeconomic strata of the population of Puerto Rico. Because women are poorer than men and exhibit a more unequal distribution of income, and because their income, and its distribution, is negatively correlated to English language abilities, the language requirement is also biased to a higher degree against lowincome women. This leads to a systematic exclusion of low-income groups and women from a fair jury selection process.

Page -26

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 0.00%

Income Distribution of Spanish Speaking Male Population 15 and Over by Ability to Speak English Puerto Rico 1989

Percent of Population Group

Difficulty Unable Equality Easily

10.00%

20.00%

30.00%

40.00%

50.00%

60.00%

70.00%

80.00%

90.00%

100.00%

Percent of Income

7. ANALYSIS OF DATA OBTAINED FROM CLERKS OFFICE 7.1. JURY CANDIDATES DATABASE The Courts Jury Administration Office made all completed and returned questionnaires available for inspection and permitted conversion of data into magnetic media. Caribbean Data Systems, a firm specializing in data collection and entry, converted filled-out questionnaires into a computer file. This file included data from 25,271 completed questionnaires. There were 14,408 questionnaires returned by mail but not completed. These, plus the completed questionnaires total 39,679. A database was built at Corplan, Inc. using a combination of database management tools that included several computer programs. The tools included FoxPro and Microsoft Access. The 25,271 records pertaining to the completed questionnaires were netted of those that had left the citizenship question blank. Given the limited amount of data available from the questionnaires, only two characteristics were feasible for research on potential jurors: 1) average age; and 2) educational attainment.

Page -27

7.2. CERTAIN CHARACTERISTICS OF POTENTIAL JURIES Average age for both sexes, when excluding those 73 and over, is 43.3 years. Average age of males, when excluding those 73 and over, is 44.4 years. Average age of females, when excluding those 73 and over, is 42.0 years. Educational attainment for respondents was grouped in two classes: High school or less and some years of college completed. The distribution is 23.2% with high school or less and 76.8% with some years of college completed. The educational attainment distribution for respondents was compared with the distribution for Spanish-speaking persons 25 years and over living in households in Puerto Rico. The distribution of these persons by educational attainment is 70.8% with high school or less, while those with some years of college represent 29.3%. On the other hand, the educational attainment distribution for Spanish-speaking persons 25 years and over that speak English easily in Puerto Rico is as follows: 42.2% with high school or less, and 57.8% with some years of college. The educational attainment distribution for respondents, which can be considered, at that stage of the selection process, as potential jurors, is significantly different from a broader Spanish-speaking population group. The difference is less striking between the potential jurors and the population group that speaks English easily. Educational attainment positively correlates with income and, thus, with economic status. Here is, therefore, an indicator of the presence of economic status bias, even at this early stage of the jury selection process. Unfortunately no data on occupations or employment status was retrievable from the questionnaires made available by the Court . The way the questionnaires are interpreted and filled out by respondents made the task of obtaining some of that data impractical. However, Census data is useful for the required analysis. Census data, combined with information obtained from questionnaires made available by the Court, corroborates the bias ingrained in the jury selection process. The empirical relationship between scholastic achievement and English language proficiency has
Page -28

already been demonstrated memorandum.

in

section

4.2,

equation

of

this

Following the approach discussed previously in section 4, a model has been specified and estimated with census tract data to determine economic status characteristics, given knowledge of scholastic achievement --measured in terms of completed years of formal education. The model estimates per capita and median household income, as a function of the proportions of residents that have completed high school or less, and those that have at least a year of college. These data was retrievable from the questionnaires made available by the Court. Statistical significance tests, at 95% confidence level, sustain the empirical relationship between scholastic achievement, English language proficiency and economic status --measured by per capita or median family income. Equation (7) specifies the model as a linear regression equation. The model accounts for gender differences. (5) Yi = a + b1 FHSi + b2 FCi + b3 MHSi + b4 MCi + ei Where: Yi = a measure of economic status (e.g., income per capita or median family income).

FHSi = resident female population 18 years and older that have completed high school or less, as a proportion of total population in census tract i; FCi = resident female population 18 years and older that have completed at least a year of college, as a proportion of total population in census tract i; MHSi = resident male population 18 years and older that have completed high school or less, as a proportion of total population in census tract i; MCi = resident male population 18 years and older that have completed at least a year of college, as a proportion of total population in census tract i.

Page -29

The summary output for the regression estimates follows.


SUMMARY OUTPUT

Median Family Income in 1989 (P084A001)


Regression Statistics Multiple R R Square Adjusted R Square Standard Error Observations ANOVA df Regression Residual Total 4 798 802 Coefficients 12003.1 -10284.2 -20920.2 5345.6 50137.9 SS 20,279,933,292.4 8,666,047,521.1 28,945,980,813.4 Standard Error 1373.2 2521.4 3263.3 3140.7 3344.9 MS 5,069,983,323.1 10,859,708.7 F 466.9 Significance F 2.9167E-207 0.837 0.701 0.699 3295.407 803

t Stat 8.74 -4.08 -6.41 1.70 14.99

Intercept M18+_HS_OR_LESS/POP F18+_HS_OR_LESS/POP M18+_YC/POP F18+_YC/POP

Once more, it is noted that explanatory variables are identified in the summary output by labels. The variable mnemonics and labels used to estimate the regression coefficients for equation 7 are listed below. FHSi = F18+_HS_OR_LESS/POP; FCi = F18+_YC/POP;

MHSi = M18+_HS_OR_LESS/POP; MCi = M18+_YC/POP. When gender differences are explicitly introduced, and the equation is fitted to census tract data, a multiple correlation coefficient of 0.84 is obtained. The model explains 70% of total variation in median family income. Furthermore, all regression coefficients have the correct signs are found to be statistically significant, in light of the calculated t statistics. The model has been used to estimate the income level characteristic of questionnaire respondents. These estimates are then compared to Puerto Rico-wide averages. For the Spanish-speaking population 25 years and over, median income is estimated at $7,166 while per capita income is estimated at $9,016.

Page -30

Median family income, for questionnaire respondents 18 to 72 years old,9 is estimated by the model at a level of $28,512.10 That is, 50% of respondents to the questionnaires live in families that have incomes higher equal to or higher than $28,512 per year. The model was applied to generate an estimate of per capita income for the potential jury set, i.e., questionnaire respondents. Per capita income of questionnaire respondents is estimated to be $12,439. As would be expected, given the educational attainment and age characteristics of this group, the potential jury set is found at the extreme end of the income distribution, i.e., among the highest 13.1% for males and 5% for females. This is clearly illustrated by the Lorenz Curves presented in section 6.5 above. These economic status characteristics highlight the resultant bias, produced by the English proficiency requirement for jury service in Puerto Rico, against the low-income groups. Specifically against the 86.9% of males that receive 20.8% of the income and the 95% of women that receive 17% of the income. Therefore, the selection process is systematically excluding 91% of the population, qualified by age, from jury service, on the basis of a seemingly neutral criterion that ultimately constitutes an income, i.e., economic status, criterion. This is certainly not a fair and true cross-section of this community. 8. CONCLUSIONS 8.1. CONCLUSION 1

Statistical analysis demonstrates that differences in economic status, indicated by various measures, among the resident population of Puerto Rico, distributed by census tracts, is found to be strongly associated with English language proficiency, with 95% level of confidence. 8.2. CONCLUSION 2 Statistical analysis demonstrates that differences in English language ability among the resident population of Puerto Rico, are found to be strongly associated with income, gender and scholastic achievement. The level of statistical confidence with which the results were obtained is at least 95%.

10

Questionnaires (798) that were received with a blank in question #1 (citizenship) were excluded. All dollar figures are valued at constant 1989 prices.

Page -31

8.3. CONCLUSION 3 Analysis based on US census data demonstrates that the income distribution in Puerto Rico is: in general, highly skewed; in particular, skewed to a higher degree against women than against men; and, that the inequality increases in inverse proportion to the English language proficiency.

8.4. CONCLUSION 4 Statistical analysis demonstrates that English language proficiency requirement systematically biases the jury selection process towards the upper brackets of the income distribution of Puerto Ricos social structure. 8.5. CONCLUSION 5 The methodology followed by the Court guarantees a fair geographic representation. However, due to the English language requirement, the jury selection process is unable to guarantee equal and fair chance of selection to low-income groups and even less to lowincome women. 8.6. CONCLUSION 6 Given the socioeconomic characteristics of the population of Puerto Rico, the English language requirement results in a jury selection process that systematically discriminates against the majority of the resident population of the District. 9. EXPERT WITNESS OPINION The statistical analysis performed on cross-sectional data collected by the US Census Bureau, at census tract or block-numbering area level, provides unequivocal evidence that there is a direct and positive relationship between English language proficiency and economic status in Puerto Rico. This relationship is supported by 1990 census data, with a 95% level of confidence, as an observable characteristic of the population of Puerto Rico. Therefore, the English language requirement is not neutral as to socioeconomic characteristics of the population from which potential jurors are chosen.

Page -32

Per capita income and per capita income of persons living in households net of the institutionalized population is dependent on English proficiency to a statistically significant degree. This is reflected in a strong, statistically significant, linear relationship between the proportion of census tract populations English language proficiency and economic status, measured by income per capita or median family income Given the high degree of inequality in the distribution of incomes that is prevalent in Puerto Rico; given that women have, on the average, lower incomes than men; given that incomes among women are even more unequally distributed; given that the distribution of income worsens, as English language proficiency decreases; and given that there is a strong association between economic status measured by income and English proficiency, it must be concluded that choosing juries on the basis of English proficiency biases the process by systematically excluding low-income groups of the population. This systematic exclusion is especially true of low-income women. Given that only 21.6% of the Spanish-speaking population of Puerto Rico 18 and older are reported as able to speak English easily, the excluded groups are not only the overwhelming majority of the population but the poorer. The bias produced by the English proficiency requirement for jury service in Puerto Rico against the low-income groups. It systematically works against that 86.9% of males that receive 20.8% of the income and that 95% of women that receive 17% of the income. Therefore, the selection process is systematically excluding 91% of the population, qualified by age, from jury service on the basis of a seemingly neutral criterion that ultimately constitutes an income, i.e., economic status, criterion. This is certainly not a fair and true cross-section of this community.

ERG/ERG

Page -33

You might also like