ABSTRACT
This paper uses 1991 census and 1990 survey data from Brazil to test hypotheses regarding the relationship between population growth (most of which stems from immigration), sex ratios, and women's labor force participation on the Amazon frontier. Strong evidence supports the link between population growth and sex ratios, though significant local variation exists. The relationship between sex ratios and gender roles, especially female labor force participation and occupational mobility, is less clear and appears variable.

INTRODUCTION
On Highway PA-275 near the 1980 gold rush boomtown Curionopolis in the eastern Brazilian Amazon one passes a fading hand painted sign reading: "OLD WEST RANCH." Arriving in the next town Parauapebas by this the only highway to the rest of the country, the first side road one encounters is the "Street of Happiness," where during the peak of construction of the giant Carajás mines between 1981 and 1985 about 250 brothels gave the neighborhood its boisterous reputation. The town's largest bordello, the "Pinga de Ouro" (Drop of Gold), faced directly on the main highway and had in one long barn-like shed fifty-seven wooden stalls. At the peak of the construction phase of Carajás, in the early 1980s 25,000 men were employed in temporary jobs. At the same time the goldrush at the nearby Serra Pelada mine employed up to 100,000 men. In an attempt to bring order to the Serra Pelada gold rush, women, gambling and firearms were excluded from the site of the mine itself. Brothels sprung up on Parauapebas' main commercial street until they were moved by the mayor to this street on the edge of town, now called simply "The Zone." At its peak in 1985 prostitution may have employed well over 500 women, or about one-fifth of the total working population of the town and over one-third of its adult women (SEPLAN, 1988).

Five years later in "The Zone," fading aqua and red paint showed the weathered wood beneath, windows were cracked, and weeds had grown up along the narrow dusty "Street of Happiness." In the mornings small children and dogs wander out into the blazing sun while a few women wash clothes in backyards. While the "Pinga de Ouro" has been converted into a third-class dormitory, a barracks for a subcontracting company, about thirty bordellos continue to function in town with names like "Nightclub Playboy," "Club Him and Her," and "Dreams of Midnight." From our small random survey of households in the town we estimated that at least 200 women continue working full-time in the sex industry, or about one in five working women. Our other systematic survey of male residents of barracks for subcontracting companies for the Carajás mines showed that two-thirds had patronized the red-light district and about a third went about [end p. 91] once a month or more. Because the construction phase of the Carajás mine is over, however, and because the Serra Pelada gold mine is now closed for practical and political reasons, business for prostitutes is thinning. Many owners have closed or moved their businesses on to newer frontier areas. Contemporary natural resource frontiers such as the Brazilian Amazon allow unique natural experiments in which to examine claims that women's paid and unpaid work is significantly affected by the ratio of women to men in a society. The rapid changes in Parauapebas' "Street of happiness" and the town's high rates of female employment-when compared to Brazil as a whole-suggest a tie between rapid population growth, shifts in sex ratios (the number of men per 100 women in a society), and women's work. Our argument is built upon the theoretical and historical arguments of Guttentag and Secord (1983) that gender roles can change with imbalances in the ratio of men and women. Their study, based upon the insights of social exchange theory, hinges on the postulate that those of the sex in shorter supply will perceive greater alternatives outside their relationship and therefore exert more power within it. Such power within relationships can, in the right context, increase the range of roles that women might play in society. Depending on the extent of the shortage of women and the rigidity with which they are controlled, Guttentag and Secord (1985, 114) hypothesize that frontiers sometimes open opportunities for women:

Frontier societies ... are fluid; along with the lack of tradition, their boundaries and populations are constantly changing; new ways of doing things and new attitudes and values are repeatedly being introduced .... The economy of the frontier is a developing one, with an acute scarcity of labor. Under such circumstances, women are not likely to be rigorously excluded from participation in productive activities outside of the home .... Under these circumstances, we would expect the high sex ratio condition, reflecting a severe shortage of women, to be favorable for them and to provide them with a certain degree of social power.

Early stages of migrations to natural resource frontiers are often primarily males of working age who are seeking their fortune in economic activities related to resource extraction, and the sex ratio is highly skewed. Later, the sex ratio of migrants changes as the predominantly male-worker-oriented settlement spawns work for women. However, with the narrowing of economic opportunities that comes with the depletion of the resource or completion of the project that initially attracted the migration, men re-migrate onwards to new resource areas and the frontier society may shift from being relatively bereft of females to the opposite condition.

Understanding these relationships between rapid population growth in areas of intense in-migration and sex ratios, and between sex ratios and the relative power of and roles available to women in turn has important policy implications for anticipating the special needs of women on frontiers or in areas of natural resource extraction and large construction projects. More broadly, however, such findings also help inform our understanding of the social effects of sex ratios and the sometimes dramatic transitions experienced by frontier societies.

We here provide a cross-sectional exploratory study of the relation between population growth, sex ratios and women's work on the contemporary Amazon frontier. To examine the importance of the level at which one aggregates data on sex ratios, we compare the local case study of Parauapebas with municipal, state, and national-level data. We pose two specific questions. First, how closely related is sex ratio to population growth on the frontier? Second, what do great differences in the number of women to men mean to the societies that develop in these towns-specifically, does the shortage of women expand or contract paid employment options for women? While historical studies will be needed to document how sex ratios and roles have changed during the maturation of the Amazon frontier, this study provides a first look at the extent of sex ratio imbalance at several levels in the Amazon and the results of some initial case study work.

The current study attempts to combine macro-level inquiry of sex ratios across municipalities and states in contemporary Brazil with micro-level analysis, to discern how dynamics are worked out at the level of one town. We explore these issues with 1991 census data from Brazil and household surveys from one Amazon frontier town in 1990. We conducted fieldwork in Parauapebas and elsewhere in the Amazon in 1988, 1989-1990, 1992 and 1993. Brazil provides an ideal setting to test the relations of population growth, sex ratio and gender roles because of the presence of both a dynamic frontier [end [92] in its northern Amazon region and settled regions in the South and Northeast.

The subject of the field research portion of this article was a massive World Bank-funded iron, manganese, and gold mining project called Carajás in the Amazon state of Pará which was developed in the early 1980s (Figure 1). Outside the controlled company town of the Carajás mine lies a boomtown named Parauapebas which provided informal commerce and services to the project and where we focused our work. Comparing our 1990 sample with earlier mining company figures we are able to show the reversal of the sex ratio in this frontier town in just five years.

PREVIOUS STUDIES
In the past ten years a growing literature has tied imbalances in the numbers of men and women in a population to a series of important features of that society. In their ground breaking book Guttentag and Secord (1983) related the sex ratio to the valuation and treatment of women. Central to their argument was the proposition that the sex which was in shorter supply would gain "dyadic" or bargaining power through their relative shortage, be able to demand better treatment, and have a greater selection in choosing a mate. However, because a shortage of women often increases the perceived threat of their infidelity, in societies where women are in short supply but men hold the bulk of power, women will be even more tightly controlled in their sexual and work behavior. Since the publication of their book Guttentag and Secord's analysis has been expanded by a series of researchers to examine the effect of sex ratios on average ages at marriage, fertility rates, divorce, suicide and illegitimacy rates, rates of violent crimes, women's participation in the labor force, types of employment open to women, and a series of other social facts (see Gutek and Cohen 1987; Messner and Sampson 1991; O'Brien 1991; South 1988; South and Lloyd 1992; South and Messner 1986; South and Trent 1988; Ward and PampeI1985).

Both Walsh (1981) and Guttentag and Secord (1983) used U. S. census data to gain insights into gender roles on the U. S. frontier in the last century. Guttentag and Secord reported that Colorado in 1860 had a sex ratio of 1659 men for every 100 women and California in 1850 had a ratio of 1228. "Moreover, sex ratios remained far above 100 for at least several decades. By 1900, these states still had ratios as high or considerably higher than the highest ratios previously reported for the midwestern frontier." (1983,126-27). Walsh, however, pointed out that imbalances on the frontier were brief, and concluded that contrary to popular perceptions, while "the nuclear household may not have been readily visible in mining and lumber camps ... it was the basic unit of social structure in most of the West." (1981,64-65). She reports that at an aggregate level the agrarian frontier (the Midwestern USA) showed sex ratios of only 117.4 at their highest, while the mining frontier's (the Far Western USA) was 280.9 at its highest, taking only thirty years to drop to 141.9. [end p. 93] The same data on sex ratios, therefore, can often be interpreted two ways-either women are in great shortage, or the shortage is brief. The current study shows that these two observations need not be contradictory, but depend on the time- and geographic-frames of the observer. Studies of international as well as internal migration have likewise documented changing sex ratios in initial versus later waves of migrants. It is apparent from the Amazon that higher levels of aggregation of both time and space hide substantial local and temporal variation. We argue, however, that these highly local conditions may be more important in determining individual experiences of the frontier for women and men, and that generalization should therefore be undertaken with caution.

Studying contemporary oil boomtowns along the Louisiana coast of the U. S. A. and comparing their results to previous research on women in boomtowns, Gramling and Brabant (1990) found a curious pattern of women's employment and wages. While jobs for men were created in great numbers, employment of women lagged behind in both the periods of boom and bust, rising later and falling later (1990,13). They documented that most of these jobs were low-paying service sector jobs in stores, providing housekeeping, hairdressing, etc. These occupations responded more slowly to the collapse in the price of oil than did direct drilling jobs. In fact, during the height of the boom, when men's labor is often in short supply and wages for them rise, women's wages actually dropped in the boomtowns Gramling and Brabant studied. Competition between the pre-boom female residents and the incoming wives of the predominantly male oil workers gluts the market for those attempting to obtain traditionally "women's jobs" (Gramling and Brabant, 1990, 5; see also an earlier review of this literature in Freudenberg, 1981).

Highly suggestive for rapidly growing frontiers and resource boomtowns is Ward and Pampel's (1985) multivariate time-series study of women's laborforce participation in sixteen countries. They found that sex ratio imbalances were the best predictor of women's labor force participation rates, concluding that "when the sex ratio in the population becomes tilted, the entrance of women into the labor force grows (663)." This was true independent of the country's level of economic development, growth of the service sector (traditionally "female-labeled jobs"), fertility rates, strength of the state, income inequality and changes in the overall demand for workers. However this relationship was strongly positive between females as a percent of the population and laborforce participation. That is, surpluses of males (as is the case in most frontier areas) decreases women's access to jobs.

On the other hand, Freudenberg (1981) raises a quite different question concerning boomtown gender roles: can the rapid social change brought to a new region open up new opportunities for women? Citing the flow of modern values from urbanized areas to rural regions where a big project might be sited, Freudenberg speculated that boomtown growth and rural industrialization might actually help break down the traditionally limited gender roles of women. Although women might continue to be excluded from the core activity such as underground mining and offshore oil rigs (common cross-nationally-see Boserup 1970), in cases where there is a shortage of male workers and a surplus of women, these women may sometimes gain access to traditionally-male jobs.

Many previous studies have been limited to cross-sectional, macro-level examinations of the correspondence between sex ratio and social facts across cities or countries. Most previous studies of frontiers have also failed to examine the link between age structures and sex ratios, thereby missing the possibility that women's social position at any given point in time could also be dependent on surplus or lack of women in certain age groups. The importance of age-specific sex ratios, especially limiting ratios to adults in their key years of family formation (15-44) was documented by Fossett and Kiecold (1991) for urban minority populations in the U. S. Migratory movements are also of course very age- and sex-specific. More importantly, although there is a latent recognition of the relationship between sex ratios and population growth, there has been little attempt to examine the extent, strength or durability of this relationship. Raising these issues initially is important because they force us to consider that gender roles may change continuously with population growth rates and shifting types of migrants on frontiers. [end p. 94]

HYPOTHESES, DATA SOURCES AND METHODS
These previous studies suggest two specific hypotheses on which we bring these data to bear. First, since young men often migrate first, we expect that frontier areas experiencing high population growth rates will have sex ratios highly skewed towards males (>105). We operationalize high population growth rates somewhat arbitrarily as over 3 percent per year, significantly above Brazil's overall growth rate of 2.2 percent annually. We examine the proposition that population growth rates will be positively correlated with sex ratios across Brazil's 26 states and the 105 municipalities in the Amazon frontier state of Pará using 1991 census data. A second, and more elusive proposition to investigate is that more non-traditional types of employment will be open to women in frontier boom towns. Part of the elusiveness of this proposition stems from the difficulty of defining which jobs a society deem to be "traditionally female." Of course such definitions shift, but here by "non-traditional" employment types for women we mean jobs that at a given time are seen as being primarily limited to men. This proposition cannot be tested formally, but descriptive data on women's labor force participation rates and sectors of employment from our one case-study town provide some suggestive qualifications to Freudenberg's (1981) hypothesis and the predictions stemming from Guttentag and Secord's observations quoted above. We examine this proposition by comparing results of our representative household survey in Parauapebas with national-level data on women's employment by age. We also systematically compare female laborforce participation with sex ratios over Brazil's 20 plus states for each decennial census from 1940 to 1980.

The data for this study come from two sources. First, data from the preliminary results of the 1991 national census of Brazil by county (município) are employed to assess the relationship between population growth and sex ratios (IBGE, 1992). To compare sex ratios and rates of female employment over the nation across the 1940-1980 period we employ four previous censuses (IBGE, 1986, 34436, 74). Second, a survey administered by the first author in March and April of 1990 to a simple random sample of 100 of Parauapebas' 5,000 households provided information on sex, age, occupation and a series of other issues for 539 Parauapebas residents. Being able to randomly choose households required assigning each house in town a unique, consecutive number: this sampling design was possible because census bureau (IBGE) workers had counted and recorded the number of houses on each side of each block in November of 1989. Three return visits at different times of the day and week were made for homes with no one present. Of occupied households, where a resident was contacted, the response rate was 100 percent: no individuals declined the confidential interviews.

Respondents provided basic demographic, occupational and migratory information on themselves and all other household members totaling 539 individuals. These surveys allowed th~ construction of an age/sex pyramid of the town's population and the broad characterization women's employment by sector. The interviews covered a series of issues: household structure, migration history, work, education, remittances, networks of mutual aid with neighbors, household division of labor, standard of living, voting, religion, etc. Whichever adult was present was interviewed, but since most surveys were done during weekdays the majority of respondents were women. This was useful because women could better respond to the extensive questions about housework and their own work without the male present. When both partners were present, questions were addressed to both. Respondents were prompted only when they did not understand the questions, and because the study probed economic strategies of households, most questions were very concrete in nature and "loaded" questions could be avoided. More complete presentation of the survey results and women's position in the household and the society can be found in Roberts (1991).

THE RELATIONSHIP BETWEEN POPULATION GROWTH AND SEX RATIOS
The picture painted by national level data comparing population growth and sex' ratios varied substantially, with states hiding much wider variations within their borders. The eight states entirely within the legal Amazon region (Figure 2A), [end p. 95] were the only states in the country experiencing population growth rates over 3 percent per year during the period 1980-1991 (Figure 2B; IBGE, 1992). Maranhão, bordering the eastern Amazon and not completely within the region, has socioeconomic and historical characteristics more closely resembling the country's Northeast region. That state was a strong origin region of migrants to the Amazon and particularly to Parauapebas. The fastest growing Amazon states were Rondônia, Mato Grosso, Roraima and Amapá, all inundated with gold miners in the 1980s and the former two the site of dense small farmer frontiers.

A strong statistical relationship was found at the state level between population growth rates over the inter-censal period 1980 to 1991 and sex ratio. Quite simply, states with higher growth rates had greater shortages of women (r= .838, n=27, p<.01), and the growth rate alone explained 70 percent of the variation in sex ratios in Brazil's states. Brazil as a whole had only 97 men per 100 women (IBGE, 1992). However, all eight Amazon frontier states showed a surplus of men overall, varying from the outlier sex ratio of 123 men per 100 women in gold rush Roraima to 101-102 men per 100 women in Acre, Pará, and Amapá (Figure 2C). The only four states in Brazil with sex ratios over 105 in 1991 were the Amazon frontier states of Amazonas, Mato Grosso and Rondônia and Roraima. Returning to our first hypothesis, we see that of the seven states with population growth rates over 3 percent, four had sex ratios over 105. A chi square adjusted for the small sample size shows that this relationship is significant at the .01 level, and the correlation reported above is strong, but Figure 2 shows that much variation exists and the relationship is not a rigid one, even within the Amazon region of Brazil.

Do higher sex ratios correspond to lower or higher levels oflabor force participation by women? While data on female participation in the labor force by state are not yet available for the 1991 census, we can compare sex ratios and rates of women working for wages in the previous five decennial censuses. While there is substantial variation, especially for states with lower sex ratios, there is an overall negative relationship between the variables. That is, in general, states with higher sex ratios (especially the Amazon region) corresponded to lower women's labor force participation.¹ While [end p. 96] this relationship appears to be weakening somewhat with time, it continues to exist.

The combination of a strong correlation and substantial variation in the relationships between population growth, sex ratio and women's labor force participation in Brazil suggests the need for closer examination at more disaggregated levels. Data on women's labor force participation rates were not available on a smaller scale, so we analyzed municipality-level data for the state where the case study was conducted (Pará) in 1991 for population growth and sex ratio only. In the next section we take up results of our survey in the municipality of Parauapebas with all three variables.

Before the current flux of migration to the Amazon frontier which began around 1960, there was already an existing but sparsely settled European-descent Brazilian population as well as numerous indigenous groups. This makes the case somewhat different from the U.S. West. Still, Pará's population has more than tripled in the last three decades, going from 1.5 million in 1960 to 5.1 million in 1991 (IBGE 1989, 1992; Brazil's population doubled in the period). Comparing annual population growth rates by municipality in the state over the period 1980-1991, truly explosive rates (over 10 percent per year) were experienced in the previously sparsely-inhabited southern part of the state, especially in wildcat gold-mining areas (Figure 3). The fastest growing areas-averaging 60.9 percent annually in Tucumã and 31.7 percent in Ourilândia do Norte-were gold mining regions. Other rapid growth areas were the result of large development and settlement projects of the Brazilian state in extraction of natural resources such as mining, lumbering, hydroelectric power and ranching/farming (for broad patterns of Amazon development see Hecht and Cockburn 1989; Schmink and Wood 1991). Meanwhile, population grew around or below the national rate of growth for the 1980s (1.89 percent) in the Bragan<;a region near the state's capital Belém, an earlier agricultural frontier. Population numbers declined in only a few municipalities.

For the 105 municipalities in the state of Pará, population growth and sex ratios were positively correlated, but not nearly as strongly as in the state-by-state comparison (r=.488, n=105, p.<.01). As predicted, gold-mining and other high-growth areas [end p. 97] saw highly skewed sex ratios (> 110), reaching 150 men per 100 women in Ourilfuldia do Norte and 122 in Tucumã. A large majority of all municipalities outside the major cities showed sex ratios over 105 men per 100 women. The correspondence between population growth and sex ratio was confounded by the fact that several slow-growing municipalities also experienced extremely skewed sex ratios. Specifically, it appears that city ward majority-female migration is occurring in the more settled areas of the northern part of the state. This observation is supported by the strongly female-dominated sex ratios in the major cities Belém, Castanhal and Santarém. Because of poor schools in the interior of the state, a common pattern there is for middle and upper-class families to send some children to study in the capital or the largest possible city nearby. Often male children are kept at home to tend the farm or family business, freeing more females to go. Also, as we shall shortly note, the differential in work opportunities between interior and cities appears far greater for women than for men.

To summarize this section, it appears that frontier states which were experiencing rapid growth rates were attracting males at a more rapid rate than females. These states also had the lowest female labor force participation rates. Within the state of Pará, men outnumbered women in almost all municipalities and most dramatically in those where gold mining took place in the 1980s.

CHANGING SEX RATIOS IN PARAUAPEBAS

The Brazilian census bureau found that population in the Parauapebas municipality (county) over the eleven years 1980 to 1991 grew at an average rate of 18.0 percent per year (IBGE 1992, 44). Gilmore (1976, 536) observed that "in most boom towns a 15 percent growth rate leads to institutional breakdowns in the labor market, the housing market, and the system for financing local public facilities." During some years in the 1980s the town of Parauapebas saw growth rates over 225 percent (Roberts, 1992).

At the town level, the ratio of men to women at Carajás/Parauapebas could appear highly skewed or quite balanced, depending on when one surveyed the population and who one counts. Mining company figures indicate that at Carajás 25,000 workers were employed during the peak of construction in 1981-1983, of which only about 500 were women (Duarte and Botelho, Interview, 2 Nov. 1989). Of these, about 20,000 men were housed in the temporary barracks in the subcontractors camp called the Vila over iron deposits near the mine. On the other hand, in Parauapebas at the end of the peak of construction in 1985, a survey by CVRD's superintendency for the mine found only 6,898 men to 6,705 women, a sex ratio of only 103 (3135 families were interviewed; Meyerhoffer/CVRD 1986). Thus two virtually adjacent towns had vastly different sex ratios, and it appears that Parauabas' sex ratio was already dropping as construction of the mine was completed.

Our survey of households in Parauapebas five years later revealed that a reversal in the balance of sexes in this young frontier town had taken place since 1985. In our random sample of households in the town in 1990 there were only 90.5 men per 100 women (268 men to 296 women): that is, there were about 14,800 women living in town to about 13,400 men---an estimated surplus of about 1,400 women. Brazilian census data are released at the county-level only, where the sex ratio for the Parauapebas municipality (including Carajás,Parauapebas, rural areas and wildcat gold mines) was much higher ---111.4 men per 100 women in 1991 (IBGE 1992,44).

Even with an overall reversal in the sex ratio, some of the "male frontier" literature suggests that women of reproductive age might be proportionally fewer than men of this age in the town. This claim, however, falters upon careful examination of the town's age/sex pyramid. Figure 4 shows the somewhat skewed population pyramid of the Parauapebas households disaggregated by sex and age, comparing its profile with that of projections for Brazil for the same year. Contrary to the widespread perception that young women are not available in the town, women aged 10 to 25 are far more abundant in Parauapebas than in the rest of Brazil. Also surprisingly, men aged 15 to 24 are less common than at the national level. Comparing young men and women based on these data is especially revealing. Seen together, women aged 15 to 29 in Parauapebas made up 19.5 percent of te town's population, while men in these age groups accounted for 13.4 percent, revealing a shortage of 1,600 young men. Further, this imbalance remains, [end p. 98] although less dramatic for women aged 15-39. And in support of Guttentag and Secord's prediction, there were a number of cases of significantly older men married to women in their early or mid-twenties (though this is not uncommon in Brazil).

The population pyramid also reveals that Parauapebas is almost entirely lacking in people over age 55, as is common for areas of new migrants. In fact, all age categories over 40 are under-represented. This means that younger families must pay for, provide themselves, or go without services once provided by parents or other older next-of-kin. This could especially affect women's ability to work outside the house if they bear children, given Brazilian attitudes about household and child-rearing chores. It also suggests the need for these young adult migrants to remit money to support retired kin.

It appears that Parauapebas men often came first to the frontier but soon brought in their families. Of 80 households who responded to the question "when you came to live here did you come alone or with others"? 51 percent reported that the male came first to secure a job and housing and then sent for the family, usually a few months to a year later. In a few of these households males had in turn moved on to other frontier areas to prospect for gold, leaving their families in Parauapebas and were often gone for months or years at a time (half were household heads and half were teenaged sons). Regarding the other households, 38 percent said the family came together to the frontier, and 10 percent were households headed by unaccompanied women. Four of these eight latter dwellings were brothels. The remaining female-headed families appeared to be among the poorest in town, relying on low-paid or unstable domestic or temporary employment. While smaller in number, these latter categories alert us to the fact that women are on the Amazon frontier and many are not financially dependent on men.

It is also important to note that the number of men is potentially understated, since many are still living in virtually all-male company lodgings, especially at Carajás in the "Vila," the subcontractors' camp. It is possible, however, to derive an estimate of their numbers. Those in Parauapebas lodgings number between 200 and 300. Plans, though, are to move all these men and women down from the Vila to Parauapebas. CVRD data state that 60 percent of subcontractor workers were already living in Parauapebas at the time of the survey (Barroso, Public Relations, CVRD, Carajas), but most of these live in houses. Yet if one were to add in the mostly male subcontractor workers in the Vila (numbering about 500-700 in 1992), we would still be left with a picture of an approximately balanced sex ratio in Parauapebas, which forces us to revise renderings of the "male frontier" in this case. The unexpected abundance of women in Parauapebas (which appears unperceived even by most local residents) places increased importance on the analysis of their occupational roles in the town. Have opportunities for women to work in this boomtown increased and decreased with the rise and fall of the incoming population, as Gramling and Brabant (1990) found? Or have early infusions of external influence expanded the range of acceptable work for women, as Freudenberg (1981) and Guttentag and Secord (1983) suggested? The [end p. 99] following section interprets some preliminary data from the household survey on the work women do outside and inside the home: work for wages, in the informal sectors, including work as prostitutes, and their unpaid work sustaining households. The analysis begins with the question of which age groups of women are employed.

WOMEN'S WORK IN PARAUAPEBAS
Our household survey suggests that there is significant variation in women's participation in wage labor by age group (Figure 5). In Parauapebas, two in three women aged 15-19 and 35-39 worked for wages. Women of the most common reproductive ages-25 to 34-were least likely of adult women to work for wages, the average for this age group being closer to one in three women. In Parauapebas the largest numbers of working women were 15 to 24 years old, accounting for over half the female workforce in the town (54 percent). In terms of percent of women in an age group working, the highest levels were for women aged 15-19 and 35-39. Six age classes, however, showed over forty percent labor force participation, and only 25-29 and 45-54 year-olds were below the national average.

Parauapebas women in almost all age groups, therefore, were more likely to work for wages than their counterparts in the nation as a whole were in 1985 (IBGE 1989: 124). Though the sample was small and the results should be seen as exploratory only, they raise three dramatic points. First, that women's labor rates overall were well above national averages; second that women of all ages were employed in the town; and third, that women accounted for about 35 percent of the workforce there. Comparing labor force participation of women to men, women's participation nearly equals men's at ages 15-19, then quickly drops off to half their rate, probably as many get married and begin to bear children. A greater proportion of women in their thirties and early forties are in the labor force, but since this is a cross-sectional view, we do not know if they ever stopped working, or if they worked before having families.

However, an analysis of the birth records for the Parauapebas hospital for the first six months of 1990 showed that 16 to 30 year-old women were most often pregnant and with newborns. The median age of women giving birth in the hospital was 22.5, and 29 percent of all births were to women under 20. Comparing data on childbearing with those on the percent of women in the workforce, we can see a clearly inverse relation between the number of women bearing children and the percent employed. While older children can share the burden of caring for new babies, gender roles and the lack of child-care facilities in Brazil certainly limit women's ability to combine many types of work and fertility.

In what occupations did most women find work?

Almost half of working women were employed in commerce and services (44.3 percent). Many women worked in their family's business, alongside their husbands running the store, hotel, restaurant or construction company. A few had enough education to become nurses and schoolteachers. Women without such opportunities worked especially in "informal" economic activities such as domestic servants, in prostitution, vending in the street or at the train station, sewing clothing in the home, as a maid for the county government or, lowest paid of all, washing clothes. The highest-earning woman encountered in the sample was a smuggler, bringing electronic goods in from Paraguay (she earned four times her husband's salary). Besides smuggling, [end p. 100] Parauapebas women in almost all age groups, therefore, were more likely to work for wages than their counterparts in the nation as a whole were in 1985 (lEGE 1989: 124). Though the sample was small and the results should be seen as exploratory only, they raise three dramatic points. First, that women's labor rates overall were well above national averages; second that women of all ages were employed in the town; and third, that women accounted for about 35 percent of the workforce there. Comparing laborforce participation of women to men, women's participation nearly equals men's at ages 15-19, then quickly drops off to half their rate, probably as many get married and begin to bear children. A greater proportion of women in their thirties and early forties are in the labor force, but since this is a cross-sectional view, we do not know if they ever stopped working, or if they worked before having families.

However, an analysis of the birth records for the Parauapebas hospital for the first six months of 1990 showed that 16 to 30 year-old women were most often pregnant and with newborns. The median age of women giving birth in the hospital was 22.5, and 29 percent of all births were to women under 20. Comparing data on childbearing with those on the percent of women in the workforce, we can see a clearly inverse relation between the number of women bearing children and the percent employed. While older children can share the burden of caring for new babies, gender roles and the lack of child-care facilities in Brazil certainly limit women's ability to combine many types of work and fertility.

Prostitution and domestic service together employed a third of all working women. Prostitution alone still employed one in five female workers in our 1990 survey (21.5 percent) ---or one in every eight adult women in town. Of the women aged 15-25 encountered in the sample, about one-third of those working were in prostitution (15 of 44). Of all women 15-19 years old, 21 percent (9 of 42) were prostitutes. Even excluding prostitutes, the 15-19 age group still had a female workforce participation rate of about 48 percent, well above the national average. Meanwhile, of all 20-24 year-olds in town, 17 percent worked in prostitution and without such employment the town's labor force participation rate in that class dropped well below the national average to about 26 percent. That the bordellos still have some business was documented in a separate systematic survey of 71 male dwellers in subcontracting company barracks (Roberts, 1991). Of these, 65 percent had been at least once to a prostitute in the red-light district of Parauapebas, and 34 went often, about once per month or more.

EVALUATING FRONTIER SEX RATIOS AND THEIR EFFECTS

We can now return to and evaluate our two central propositions and their implications. Based on the observation that migrant streams are often biased towards males in their early stages, and the nature of employment on the extractive frontier of Amazonia, we predicted that population growth rates would be positively correlated with sex ratios. The first specific hypothesis was that areas experiencing high population growth rates (over 3 percent per year) due to heavy in-migration will have sex ratios highly skewed towards males (>105). This hypothesis was strongly supported in a comparison across Brazil's 26 states. Across the 105 municipalities in the Amazon frontier state of Para, this relationship existed but was much weaker. The tendency for local conditions to be far more complex than agglomerated trends reinforces the point made at the outset that attention must be paid to the micro-level in examining sex ratios and their effects on gender roles. As set out in the introduction, we believe that sex ratios measured at the level of the town-even more disaggregated than municipality data--are the most important for the actual experiences of frontier women and men.

Frontiers are heterogeneous, and future work in rural sociology and the sociology of frontier regions might discover patterns in sex ratios and gender roles that vary by which types of commodities are being extracted and the degree to which that frontier growth has clustered in boomtowns. In fact, the Amazon frontier has become a largely urban one. To the surprise of many current social researchers in the Amazon, since the early 1970s hundreds of new towns have emerged and since 1980 the majority of the Amazon population now lives in urban areas (Browder and Godfrey 1990; IBGE 1981). Second, general models of the transition from rural to urban frontiers are beginning to give way to a more nuanced view that towns in the Amazon take starkly different forms based especially on what commodity is being extracted and on the level of government presence, planning and resources (Becker 1990; Browder and Godfrey 1990; Godfrey 1992; Roberts 1992; Sawyer 1987). The heterogeneity of the frontier should not be seen as a new development, however; Alistair Hennessey (1978), for example, distinguished nine types offrontiers in Brazil's early history alone. This examination of Parauapebas underlines the need to combine case study data which raise questions and check the validity of census data with comparisons at different levels of aggregation.

The more elusive proposition of Freudenberg (1981) and Guttentag and Secord (1983)----that more non-traditional types of work will be open to women in frontier boom towns because of the lack of established traditional society and shortage of male workers-was not supported but cannot be refuted by our fieldwork in Parauapebas. The timing of the study is an important factor since our survey also documented that sex ratios in the town appear to have reversed between 1985 and 1990. Although we must rely on a mining company census for 1985 and interviews with government and mining company officials, it appeared that women's work opportunities and labor force participation were virtually unchanged while sex ratios had reversed.

Defining non-traditional work also remains problematic, and while women's labor-force participation was higher in Parauapebas across almost all age classes when compared to Brazil as a [end p. 101] whole, we found that most women workers there were confined in 1990 to low-paying service sector jobs and prostitution. We have seen in this maturing mining boom town that even though prostitution is in decline it continues to employ one in five female workers. We must therefore confront the question: what outstanding features of life in Parauapebas sustained widespread prostitution?

By requiring large numbers of male workers for temporary or uncertain work in a dangerous or unknown region, mining and other work-camps present massi ve markets for commodified sex. Riley cites the extreme case of Helena, Montana, where prostitution was the largest single source of paid employment for women for the key boom years (Riley 1988, 131-32). She reported that United States census data sometimes intentionally obscured the presence of prostitutes on the frontier by listing women who worked in this way as cooks, servants, or as unemployed (1988, 121-22). Many women followed the booms around the west, some had seasonal routes of migration. The openness with which Parauapebas women spoke of their profession and the legality of the work belie concerns that such work was under-reported in this survey. Under Brazil's 1988 constitution prostitution is not considered a crime, only those who make money off the prostitution of others are considered criminals.²

One historical study has taken up the question of mining and prostitution specifically and deserves a moment's attention. Marion Goldman carefully reconstructed the world of the "Bawdy District" in nineteenth-century silver mining towns of Nevada, tying the explosion in prostitution there with a series of factors which boosted demand (Goldman 1981, 34-37). The unbalanced sex ratio, the urban nature of silver mining towns, violence, lack of government presence, and lack of other distractions in the simple towns all sent men in search of commodified sex. She finds statements from the period to the effect that men's need for sex was "fueled by a widespread popular 'volcano theory' of sexuality which held that unless men had regular sexual contact, they would explode in orgies of adultery, rape, physical violence, or even homosexual embraces." (p. 35).

Daniels also notes for Australia's mining frontier that isolated locations led to an overwhelming demand for sex that drew aboriginal women into prostitution (Daniels 1984, 168-70, 252-56). Gramling and Brabant (personal communication) report prostitution in the Louisiana oil rig-supplying boomtowns mentioned above. The causes Goldman cites are all confirmed by these initial observations of prostitution in Parauapebas. Perhaps most intriguing, though, is Goldman's claim that exposure to risk at work leads solitary mining men to turn more frequently to prostitutes:

The danger and difficulty of work in the mines and the lesser perils of most other workingmen's occupations fueled the market for prostitution .... Discomfort and occasional terror in the mines probably produced immediate emotional strains which set many men in motion toward the bawdy districts after their shifts were over. (Goldman 1981,36).

Of course such psychological explanations are quite untestable for historical cases. However, noted sociologists have reflected on how social mores and "future orientation" have broken down in times of uncertain prospects (Liebow 1967,68-69, fn. 24).

Prostitution is quite common in all parts of Brazil and is badly neglected in the scholarly literature. In 1992, a best-selling book by reporter Gilberto Dimentein shocked Brazil recounting the prostitution-slavery of girls in the gold-mining towns across the Amazon region. He reports that the Brazilian Center for Infants and Children (CBIA) has recently estimated the number of girls in prostitution in Brazil at 500,000 (Dimentein 1992,11). Brothel owners in Parauapebas reported being in a bind because of the demand for young girls and the legal prohibition against their employment. The same law forces most prostitutes to abandon their own children because they are not allowed to live in the bordellos. Prostitution in Parauapebas has always involved a large number of illegal minors, and owners have developed a system of paying off the enforcement agents to keep from being prosecuted. One owner recounted:

There were a lot of young girls working here, and in 1986 the Judge of Marabá ruled they couldn't keep children here [in the brothels]. But he never came, just sent others. The [representative of the] Child-Labor Judge comes by on Sunday and we [each] pay him twenty to 100 Cruzeiros [tip]. Now they're calling for documentation of all the women. They don't really come to help the women, just to collect their tip.

The owners have responded to the pressure by meeting "every week" to decide on how, and how much, to tip the police. [end p. 102]

An important and less speculative conclusion from these interviews was that prostitution provided one survival or mobility strategy for women coming from desperate poverty in the country's northeast. All prostitutes interviewed came from sharecropper or farm laborer households; none was the child of a landowner of even the owner of a modest parcel. At least in one sense their material welfare had markedly improved, and this was what many other women noticed: they suddenly had more expensive clothing, jewelry and makeup. Of course much of their new accoutrements were not their own, but part of the owners' advancing them the tools of their work, and as a means to increase their debts (Roberts 1991).

Even with these new levels of laborforce participation in a "new town," we also documented elsewhere that most women remain burdened with traditional chores: performing the daily, time-consuming and exhausting housework involved in maintaining workers and reproducing the labor force generationally (Roberts 1991). In spite of women in Parauapebas being more likely than Brazilian women as a whole to be active participants in the labor force; in spite of men being in shorter supply in the town; and in spite of experiencing the rapid social change that Freudenberg (1981) and Guttentag and Secord (1983) suggested might broaden the range of work and other activities open to women in a boomtown, career mobility is extremely limited for women in Parauapebas. Women remain confined almost entirely to housework and to low pay, low status service jobs, with some in informal vending, smuggling, and prostitution. In short, women remain largely excluded from formal jobs which would provide them higher pay, established contracts, some job security and legal recourse against their employers. Government service, employing just fi ve percent of the women who were working for pay, appears to be virtually their only means of upward career mobility, jobs which are more common in state and national capitals than in frontier towns.

Population growth rates change depending on the economic dynamics of the frontier. We have shown here that with changes in population growth can come changes in the sex ratio. Continuing and dramatic population shifts in the region (documented in Godfrey 1992 and Roberts 1992) are likely to lead to further changes in sex ratios. In sum, this study suggests that we must view the relationship between population growth, sex ratios, and gender roles as a dynamic one, which present rapidly changing opportunities for women outside the home. Among other implications, understanding these dynamics would facilitate better planning of government services for women in resource boomtowns.

Better longitudinal data are clearly necessary to document changes in women's labor force participation in rapidly changing frontier towns. Such longitudinal data could considerably improve the reliability of statements about the relationship between population growth and gender role changes on frontiers. Such research should improve on the quality of data on frontier gender roles, bargaining power, remittances, housework, etc. Taking advantage of these natural experiments will require that social scientists reach the frontier soon after cities are founded and continue to return as the cities evolve. In addition to its vast natural wealth, the rapid growth and heterogeneity of the modern Amazon frontier provides a nearly untapped resource of natural experiments for social researchers to deepen our understanding of the relations between social change and gender roles.

ACKNOWLEGMENTS
We are grateful to Stephen Bunker, María Patricia Fernández-Kelly, Joel Devine, April Brayfield and Cindy Truelove for advice in the field and for comments on earlier drafts. We also would like to thank the Companhia Vale do Rio Doce, IBGE, Marabá and Belém, and the Prefeitura of Parauapebas. The field research upon which this article is based was funded by the Fulbright-Hays Commission, the Tinker/Johns Hopkins University Atlantic Studies Summer Research Grants Program, and two Mellon Foundation/Tulane Center for Latin American Studies travel grants. [end p. 103]

NOTES
¹ These correlations were statistically significant for 1940, 1950, 1970 and 1980. Only 1960, with the outlier point of Brasília, D. F, where massive construction teams of men were building the new national capital, was not. These correlations compare percent of all females who were listed as employed divided by the female population with the sex ratio. 1940: r = -.728, N=21, p<.001; 1950: r = -.409, N=24, p<.05; 1960: r = -.090, N=25, N.S. [with Brasília excluded, 1960: r = -.531, N=24, p<.01; 1970: r = -.363, N=25, p<.l; 1980: r = -.418, N=26, p<.05.

² As Rohner (1987: 54-55) pointed out: "In practice these laws aren't observed or are converted into their contradiction. The prostitute is prosecuted inside and outside 'the zone'. Only in rare cases are the madames or pimps repressed."

REFERENCES
Becker, Bertha K. 1990. Amazônia. (São Paulo: Editora Atica).

Benería, Lourdes and Marta Roldan. 1987. The Crossroads of Class and Gender: Industrial Homework, Subcontracting, and Household Dynamics in Mexico City. (Chicago: University of Chicago Press).

Blumberg, Rae Lesser. 1991. "Introduction" to Gender, Family and Economy: The Triple Overlap. (Newbury Park: Sage).

Boserup, Ester. 1970. Women's Role in Economic Development. (London: Allen and Unwin).

Browder, John and Brian 1. Godfrey. 1990. "Frontier Urbanization in the Brazilian Amazon: A Theoretical Framework for Urban Transition." Yearbook of Latin Americanist Geographers vol. 16: 56-66.

Cavalcante, Clovis. 1978. Viabilidade Do Setor Informal. (Recife, Brazil: Ministério da Educação e Cultura).

Daniels, Kay. 1984. So Much Hard Work: Women and Prostitution in Australian History. (Fontana: Collins Press).

Dimenstein, Gilberto. 1992. Meninas da Noite: A Prostituição de Meninas-Escravas no Brasil. (São Paulo: Editora Atica).

Fossett, Mark A. and K. Jill Kiecolt. 1991. "A Methodological Review of the Sex Ratio: Alternatives for Comparative Research." Journal of Marriage and the Family vol. 53: 941-957.

Freudenburg, William R. 1981. "Women and Men in an Energy Boomtown: Adjustment, Alienation, and Adaptation." Rural Sociology vol. 46(2): 220-44.

Gilmore, John S. 1976. "Boom Towns may Hinder Energy Resource Development." Science: vol. 191: 535-540.

Godfrey, Brian 1. 1992. "Migration to the Gold-mining Frontier in Brazilian Amazonia." Geographical Review vol. 82: 458-469.

Gutek, Barbara A. and Aaron Groff Cohen. 1987. "Sex Ratios, Sex Role Spillover, and Sex at Work: A Comparison of Men's and Women's Experiences." Human Relations vol. 40: 97-15.

Guttentag, Marcia and Paul F Secord. 1983. Too Many Women? The Sex Ratio Question. (Beverly Hills: Sage).

Hennessey, Alistair. 1978. The Frontier in Latin American History. (Albuquerque: University of New Mexico Press).

IEGE (lnstituto Brasileiro de Geografia e Estatística). 1981. Sinopse Estatística da Região Norte. Rio de Janeiro: lEGE.

_____ . 1986. Estatísticas Historicas do Brasil. (Rio de Janeiro: lEGE).

_____ .1989. Anuario Estatístico do Brasil 1989. (Rio de Janeiro: lEGE).

_____ . 1992. Censo Demográfico 1991: Resultados Preliminares. (Rio de Janeiro: lEGE).

Liebow, Elliot. 1967. Tally's Corner. (New York: Little, Brown and Company).

Medici, Andre Cezar. 1987. "Mulher Brasileira: Muito Prazer." Revista Brasileira de Estatística vol. 48 (189/90): 71-97.

Messner, Steven F and Robert J. Sampson. 1991. "The Sex Ratio, Family Disruption, and Rates of Violent Crime: The Paradox of Demographic Structure." Social Forces vol. 69: 693-713.

Meyerhoffer, M. Risin, and CVRD. 1986. "Diretrizes Para Integração dos Nucleo Urbanos do Rio Verde e Parauapebas." Unpublished paper.

O'Brien, Robert M. 1991. "Sex Ratios and Rape Rates: A Powe-Control Theory." Criminology vol. 29: 99-114.

Riley, Glenda. 1988. The Female Frontier: A Comparative View of Women on the Prairie and the Plains. (Lawrence, KS: University Press of Kansas).

Roberts, J. Timmons. 1991. Forging Development, Fragmenting Labor: Subcontracting and Local Response in an Amazon Boomtown. Ph. D. dissertation, Department of Sociology, Johns Hopkins University.

_____1992. "Squatters and Urban Growth in Amazonia." Geographical Review vol. 82: 441-457.

Rohner, Teodoro Helmut. 1987. Prostituição e Libertação da Mulher: Pastoral da Mulher Marginalizada: Subsídios para a Formação de Agentes. (Rio de Janeiro: Editora Vozes).

Sawyer, Donald. 1987. "Urbanização da Fronteira Agrícola no Brasil." in A Urbanização da Fronteira. (ed.) L. Lavinas (Rio de Janeiro: Publipur): 43-55.

SEPLAN-PR (Secretaria do Planejamento: Authors Meyerhoffer and Toledo). 1988. vol. XXIV: Parauapebas/Rio Verde. In Plano Diretor Do Estrada Ferroviaria Carajas. Infra-estrutura Urbana e Servicos. (Brasília: Secretaria de Planejamento).

South, Scott J. 1988. "Sex Ratios, Economic Power, and Women's Roles: a Theoretical Extension and Empirical Test." Journal of Marriage and the Family vol. 50: 19-31. [end p.104]

South, Scott 1. and Kim M. Lloyd. 1992. "Marriage Opportunities and Family Formation: Further Implications of Imbalanced Sex Ratios." Journal of Marriage and the Family vol. 54:440-451.

South, Scott J. and Steven F. Messner. 1986. "The Sex Ratio and Women's Involvement in Crime: a Cross-National Analysis." Sociological Quarterly vol. 28: 171-188.

South, Scott J. and Katherine Trent. 1988. "Sex Ratios and Women's Roles: a Cross-National Analysis." American Journal of Sociologyvol. 93: 1096-1115.

Summers, Gene F. and Kristi Branch. 1984. "Economic Development and Community Social Change."i> Annual Review of Sociology vol. 10: 141-66.

Walsh, M. 1981. The American Frontier Revisited. (Atlantic Highlands, NJ: Humanities Press).

Ward, Kathryn B. and Fred C. Pampel. 1985. "Structural Determinants of Female Labor Force Participation in Developed Nations, 1955-1975." Social Science Quarterly. vol. 66: 654-667.

Wood, Charles H., and Jose Alberto Magno de Carvalho. 1988. The Demography of Inequality in Brazil. (Cambridge: Cambridge University Press).

RESUMEN
Esta pesquisa utiliza o censo demográico do Brasil de 1991 e entrevistas feitas em 1990 no Estado do Pará para investigar algums hopóteses sobre a relação entre crecimiento da poblação (na maioria por causa da migração), proporções dos dois sexos (sex ratios), e participação das mulheres no trabalho remunerado. A evidecia e forte no caso da relação entre crescimento populacional e proporção dos sexos, mas existe muita variação ao nivel local. A relação entre proporção dos sexos e papeis ou funções dos sexos (sex roles), particularmente participação na força de trabalho e mobilidade ocupacional, é menos claro e aparece variável. [end p. 105]