Assistance I

22 Chapter 1 • Introduction

Exercise 1.3 Going Native

Going native:

POTENTIAL SOURCES OF OBSERVER BIAS iN POLICE OBSERVATIONAL DATA

Richard Spww'

Social Science Research, Vot . 34, pi'. 591-617,1urx 2004.

Abstract Much of our knowledge about police behavior "on the street" is based on in-person ob- servation of the police. However, little research has examined how the observers can potentially bias obser- vational data. In this paper, anecdotal accounts from a variety of field settings are categorized into four potential sources of observer bias. Secondary analyses of data from a large-scale observational study of po- lice are used to assess the impact of three out of the four types o£ observer bias (reactivity, going native, and burnout). There is evidence of reactivity for arrest, but not use of force behavior and little support for going native and burnout. More generally, observer bias is categorized as: (1) threats to the validity of observa- tional data collected at different stages of fieldwork; and (2) the potential for human error implicit in obser- vational data. The incorporation of a semi-structured field diary is suggested to more systematically document all four types of observer bias in future research.

I. INTRODUCTION

Since the discovery of police discretion, researchers have recognized the potential for bias, inequities, and prejudice in the application of the law by the police (Walker, 1992). Police researchers have documented the limitations of organizational controls on police behavior "on the street" as well as the factors that influence the behavior of beat officers (Reuss-Ianni, 1993). There is also a growing appreciation for the role of field research and observational data in the study of discretionary decision-making given the limitations of other sources of information on police behavior.

Police observational data have strengthened our understanding of the determinants of police be- havior by examining explanatory variables that encompass multiple units of analysis (e.g., officer and or- ganizational characteristics) (Mastrofski and Parks, 1990). For example, police researchers have observed officer behavior during encounters with citizens and found that police behavior is strongly in- fluenced by the structural characteristics of the situation or "situational factors" such as the nature of the problem, the characteristics and behavior of involved citizens, and the location of the encounter (Riksheim and Chermak, 1993).

In short, simply "being there" gives the researcher access to events, aspects of police behavior, and sit- uational dynamics not recorded in police reports and arrest records (Emerson, 2002a, p. 2; Ferrell, 1998, p. 27). However, some researchers categorize field research as a "marginal methodology" that produces data of questionable quality (benzin and Lincoln, 2000; Goodwin and Horowitz, 2002). The most serious challenge centers on the concept of observer bias. Since the observer is part of the context of observed behavior, ob- servers can potentially "bias" or contaminate observational data and undermine its reliability and validity (Tedlock,1991).

The purpose of this study is to identify potential sources of observer bias and to conduct a limited test of its impact on police observational data. Specific research questions will be developed utilizing concepts derived from the extant literature to operationalize three out of four types of observer bias (reactivity, cul- ture shock, and burnout) and assess its impact on POPN observational data.

2. FOUR TYPES OF OBSERVER BIAS

2.1. Reactivity

One common criticism of observational data is that research subjects may react to the presence of an ob- server by changing his or her behavior. [Note: This section has been omitted.]

2.2. Culture Shock

Immersion into an alien environment can be an intense and unsettling experience (Emerson, 2002b). Cul- ture shock refers to feelings of disorientation and anxiety that observers feel during the early stages of field research (see DeWalt and DeWalt, 2002, for a review). [Note. This section has been omitted.]

3Fax: 1-205-348-7178. E-mail address: [email protected] 0049-089X/$—see front matter. 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.ssresearch.2004.05.003

i

Chapter 1 • Introduction 23

2.3. Going Native

Gans (1982) argued that observers become sympathetic toward individuals under study for a number of ma-

sons. One reason is that observers and research subjects work together in close quarters over the course of

fieldwork. A natural consequence of the observer's task is a more sympathetic outlook toward individuals un- der observation. A second reason is that researchers feel a pull toward increased involvement while collecting observational data (see Adler and Adler, 1987, for an overview). Palmer and Emerson (2002) argued that re search subjects could use observers as an assistant and a resource while conducting their work. This level of involvement can range from innocuous requests for observers to act as a "go-fer" to participation in more crit- ical activities. For example, while studying drug dealers, Adler (1985) loaned drug dealers money, watched their children for extended periods of time, testified in court on their behalf, and allowed them to conduct drug deals in Aler's home. While observing police officers, Norris (1993) not only made tea for the shift, but also helped to chase and arrest suspects and administered first aid following suicide attempts. Since he did not look like a police officer, Skolnick (1966) aided police by walking into a bar to find a dangerous armed robber. Skolnick also helped some officers get past a lookout by driving a disguised truck up to a building.

A more general concern is that observing research subjects in their natural settings has the potential to re- socialize the researcher. This process of resocialization has been described as "overapport" and conceptualized as the problem of "going native," since observers lose sight of their role as a researcher and, in extreme cases, begin to think of themselves as members of the group under study. Observers may become involved as partici- pants and systematically influence the phenomenon under study (Manning, 1976; Pollner and Emerson, 2002). Reiss (1968, pp. 365-366) described one instance of observer socialivation that occurred during an observa- tional study of police. In aprecinct located in a high crime area populated predominantly by African-Americans, a common practice in the stationhouse was to walk down the row of cells and flush the toilets when an inmate asked for water. One evening, an inmate asked for water and an observer imitated the officers by walking down the cellblock and flushing the toilets. On his last shift, one officer asked the observer what he learned over the summer. The observer shocked a fellow observer by replying "I learned to hate niggers." In extreme cases, ob- servers could transfer their loyalty to research subjects, completely abandon data collection, and refuse to return from the field. Anthropologists who have gone native have stopped publishing material completely, rejected ac- ademia, and decided to live with indigenous people who were the topic of inquiry (Gronewold,1972).

2.4. Burnout

Burnout refers to inaccurate documentation of observational data at later stages of fieldwork due to the mentally and physically demanding nature of data collection (Fine, 1993; Lofland and Lofland, 1995). [Note: This section has been omitted.]

3. DATA

This study utih7es POPN observational data of patrol officers collected in Indianapolis and St. Petersburg dur- ing the summers of 1996 and 1997, respectively. Shifts were selected for observation based on time of day and day of the week, but busier days (Thursday through Saturday) were oversampled to maximize the number of police-citizen encounters. Observers rode with patrol officers in 12 of Indianapolis's 50 patrol beats and 12 of St. Petersburg's 48 patrol beats. Beats in both sites were selected to capture variation in service conditions within each jurisdiction. In addition, an attempt was made to match service conditions for beats selected from St. Petersburg and Indianapolis (Mastrofski et al., 1998). Thirty-seven observers accompanied patrol officers during 729 shifts and documented a total of 2488 encounters with suspects. Observers accompanied patrol of- ficers over the course of a shift and were trained to unobtrusively take brief field notes on their behavior.

4.1. Research Questions

4.1.2 GOING NATIVE This study will examine two research questions related to going native. First, are observers becoming socialized by officers and developing more positive attitudes toward the police over the course of fieldwork? Second, do these shifts in observer attitudes toward police (ATP) predict whether ob- servers help police officers and/or participate in police work?

5. RESULTS

5.2. Going Native

In Table 2, the impact of going native in POPN observational data is assessed by examining: (1) shifts in ob- server's attitudes toward police (ATP) over the course of fieldwork and (2) the probability that an observer will participate during fieldwork by helping with police work. Panel A in Table 2 addresses the first issue by examin- ing observer surveys (N = 32) that were completed before the start of fieldwork (Ti—see column one) and im- mediately following fieldwork (I2—see column two). Observer's ATP became more positive between Ti and T2 for the fast three out of four attitudinal measures, but this change is only statistically significant at the .05 level for one out of the three measures (see Panel A). For the fourth measure, there was no change in observer ATP

24 Chapter 1 • Introduction

TABLE 2.... Assessing the impact of observer.bias on police observational data: going native as a shift in observer's attitudes toward police and observer's level of participation impolite work (observer effects)

Before Fieldwork (TI) After Fieldwork (12)

(A) Observer's attitudes toward police (ATP): mean comparison Do police discriminate toward certain groups? (coded 0,1 where 1 = never or rarely) .13 .28 Would you approve if an officer struck a citizen who was attacking the officer with his/her fists? (coded 0,1 where 1 = yes) •84 •94 Would you approve if an officer struck a citizen attempting to escape custody? (coded 0,1 where 1 = yes) 59**,a •72 Would you approve if an officer struck a citizen who said vulgar or obscene things to the officer? (coded 0,1 where 1 = yes) .00 .00 N (number of observers who completed both Ti and T2 observer surveys) 32 32

Observer Effects

Observer Helps Police Observer Helps with OfficerC Police Workd

(B) The relationship between change in ATPb and observer effects: odds ratios from logistic regression Observers with more positive ATP (P2 >11) 50*,e .54 Observers with no change in ATP (12 = Ti) f _ Observers with more negative ATP (12 <T1) 1.25 .69 Initial ATP (before fieldwork) 2.07** 2.46** Model, 19.3*** N (number of shifts where observers completed both Ti and 12 observer surveys) 598 598

(A) Observer's attitudes toward police (ATP): mean comparison. °Two-tailed significance level indicated by a test of independence. bMeasures of change in ATP in Panel B were derived from an additive scale of the fast three observer survey questions from TI and T2 in Panel A. This more general measure of ATP at TI and T2 ranged from zero to three. To determine if the observer's ATP became more positive (T2 > TI), stayed the same (T2 = Ti), or more negative (T2 < TI), scores on this additive scale were compared as described above. `Dichotomous coding derived from descriptive data on reactivity (1 =rides where the observer helps patrol officer) (see Spano, 2003 for a more detailed description). `Level of statistical significance is derived from odds ratios from multivariate logistic regression. (Observers with no change in ATP is the comparison group. *p<.1 """p<.05 ***p<.01 i

Panel B in Table 2 addresses the second issue by investigating whether observers with more positive ATP are more likely to "participate" during fieldwork by helping police officers. An additive scale was devel- oped using the observer's first three attitudinal measures from Panel A from Ti and T2. Observer's ATP were categorized as more positive if this additive scale of ATP at T2 was greater than ATP at Ti. If there was no change in ATP, then T2 and Ti ATP were equal. Finally, observer's ATPs were categorized as more negative if T2 was less than Ti ATP.

Two measures of observer participation (or observer effects) were derived from a dichotomous coding scheme derived from descriptive and quantitative data on reactivity (see Spano, 2002, 2003, for a more detailed description). Column one in Panel B contains rides where observers helped police officers in any capacity. This measure includes more mundane and innocent forms of participation (e.g., the observer tells the officer that s/he left his/her lights on or offers advice when asked about whether to pull over a vehicle) to instances when ob- servers helped officers with police work (e.g., holding handcuffs or helping to fill out paperwork, arrest and/or detain suspects) (see column one in Panel B) (Spano, 2002). Column two in Panel B is a subset of cases from the q»aiitative coding in Column one where observers helped the officer with police work

Panel B examines shifts where observers completed both TI and T2 observer surveys (N= 598). Ob- servers with more positive ATP (T2 > Ti), no change in ATP (T2 = Ti), and more negative ATP (T2 <T1) were regressed on the two dichotomous measures of observer effects discussed above using multivariate lo- gistic regression after controlling for the observer's initial ATP before fieldwork. Observers with more pos- itive ATP are about 50% less likely to help police officers (see column one in Panel B). Although this

I I'

Chapter 1 • Introduction 25

relationship is marginally significant (p < .l), the direction of the effect is not in the expected direction. Ob- servers with more positive ATP are not more likely to help police officers or participate in police work, which would be consistent with going native. In fact, this finding suggests exactly the opposite.

6. IS RAPPORT A PROBLEM (GOING NATIVE) OR A SOLUTION (REACTIVITY)?

Another contradiction in the methodological field research literature relates to rapport. The development of rapport is essential to minimize reactivity to get research subjects to feel comfortable and trust observers, and to act naturally while under observation. Reactivity suggests that, for example, too little rapport be- tween police officers and observers can cause police officers to engage in atypical behavior and thus under- mine the validity of observational data collected during the early stages of fieldwork. In contrast, going native suggests the opposite: Too much rapport between observers and research subjects will affect the ac- curacy of data collected at later stages of fieldwork. Going native could be rationalized as a way of mitigat- ing reactivity by establishing a high level of rapport with those being observed. However, Miller (1952) observed that friendship is based on an all-accepting outlook that can limit the depth of an observer's inves- tigation into and criticism of the behavior of research subjects (see also Van Maanen, 1983a, 1983b). Thus, going native could limit the scope, the depth, and the accuracy of observational data.

1. SUMMARYAND CONCLUSION

Observers are a potential source of bias for two reasons. First, observers are part of the context of observed behavior. As a result, the research subject could react to the presence of the observer and behave atypically (reactivity). Second, observers are also data collection instruments. Thus, the quality of observational data could be undermined if observers become resocialized, contaminate observational data by participating in police work, and/or document inaccurate observational data (going native).

The relevance of these Endings is not limited to the field of criminology and criminal justice. Ethno- graphers, researchers who conduct face-to-face interviews, and child development researchers who "unob- trusively" study the interactions between parents and children (to name a few) all collect and utilize observational data to further our scientific understanding of various real-world phenomenon and use of force behavior, needed to be statistically significant. Second, the direction of the relationship between measures of observer bias and the outcome measures also needed to be consistent with hypothesized effects specified as research questions for reactivity, going native, and burnout.

The analyses of POPN descriptive and quantitative data focused on observer bias (going native) and revealed key findings. One key finding is the very limited support for observer bias in the form of going na- tive and burnout. On the surface, observers' ATP became more positive over the course of fieldwork in terms of absolute change, but the change in ATP was statistically significant for only one of four measures. In addition, findings from multivariate logistic regression were marginally significant (p < .1) and showed that observers whose ATP became more positive over the course of fieldwork were less likely to "partici- pate" in the form of helping officers for one out of two measures of observer effects. In contrast, the litera- ture and logic of going native would predict that these observers would be the most likely to participate in police work.

Future research should include additional control variables to rule out alternative explanations for patrol officer arrest and use of force behavior. To further explore the robustness of these findings, multi- variate analyses should be conducted to determine if observer bias is a significant predictor of, for exam- ple, arrest and use of force behavior net of a variety of control variables (e.g., officer work orientation, suspect characteristics, and situational factors). If the significant effect is suppressed in the "full" multi- variate model, then it is presumed that observer bias has not substantively affected the quality of police ob- servational data.

References

Adler, P.A., 1985. Wheeling and Dealing. Columbia University Press, New York. Adler, P.A., Adler, P, 1987. Membership Roles in Field Research. Sage Publications, Newbury Park, CA. Denzin, N.K., Lincoln, Y.S., 2000. The discipline and practice of qualitative research. In: Denzin, N.K , Lincoln,

Y.S. (Fels.), Handbook of Qualitative Research, second ed. Sage Publications, Thousand Oaks, CA, pp. 1-28. DeWalt, K.M., DeWalt, B.R., 2002. Participant Observation: A Guide for Fieldworkers. A1taMira Press,

New York. Emerson, R.M., 2002a. Introduction: the development of ethnographic field research. In: Emerson, R.M.

(Ed.), Contemporary Field Research: Perspectives and Formulations, second ed. Waveland Press, Prospect Heights, 1L, pp. 1-26.

Emerson, R.M., 2002b. Fieldwork Practice: issues in participant observation. In: Emerson, R.M. (Ed.), Contemporary Field Research: Perspectives and Formulations, second ed. Waveland Press, Prospect Heights, IL, pp. 113-151.

22 Chapter 1 • Introduction

Exercise 1.3 Going Native

Going native:

POTENTIAL SOURCES OF OBSERVER BIAS iN POLICE OBSERVATIONAL DATA

Richard Spww'

Social Science Research, Vot . 34, pi'. 591-617,1urx 2004.

Abstract Much of our knowledge about police behavior "on the street" is based on in-person ob- servation of the police. However, little research has examined how the observers can potentially bias obser- vational data. In this paper, anecdotal accounts from a variety of field settings are categorized into four potential sources of observer bias. Secondary analyses of data from a large-scale observational study of po- lice are used to assess the impact of three out of the four types o£ observer bias (reactivity, going native, and burnout). There is evidence of reactivity for arrest, but not use of force behavior and little support for going native and burnout. More generally, observer bias is categorized as: (1) threats to the validity of observa- tional data collected at different stages of fieldwork; and (2) the potential for human error implicit in obser- vational data. The incorporation of a semi-structured field diary is suggested to more systematically document all four types of observer bias in future research.

I. INTRODUCTION

Since the discovery of police discretion, researchers have recognized the potential for bias, inequities, and prejudice in the application of the law by the police (Walker, 1992). Police researchers have documented the limitations of organizational controls on police behavior "on the street" as well as the factors that influence the behavior of beat officers (Reuss-Ianni, 1993). There is also a growing appreciation for the role of field research and observational data in the study of discretionary decision-making given the limitations of other sources of information on police behavior.

Police observational data have strengthened our understanding of the determinants of police be- havior by examining explanatory variables that encompass multiple units of analysis (e.g., officer and or- ganizational characteristics) (Mastrofski and Parks, 1990). For example, police researchers have observed officer behavior during encounters with citizens and found that police behavior is strongly in- fluenced by the structural characteristics of the situation or "situational factors" such as the nature of the problem, the characteristics and behavior of involved citizens, and the location of the encounter (Riksheim and Chermak, 1993).

In short, simply "being there" gives the researcher access to events, aspects of police behavior, and sit- uational dynamics not recorded in police reports and arrest records (Emerson, 2002a, p. 2; Ferrell, 1998, p. 27). However, some researchers categorize field research as a "marginal methodology" that produces data of questionable quality (benzin and Lincoln, 2000; Goodwin and Horowitz, 2002). The most serious challenge centers on the concept of observer bias. Since the observer is part of the context of observed behavior, ob- servers can potentially "bias" or contaminate observational data and undermine its reliability and validity (Tedlock,1991).

The purpose of this study is to identify potential sources of observer bias and to conduct a limited test of its impact on police observational data. Specific research questions will be developed utilizing concepts derived from the extant literature to operationalize three out of four types of observer bias (reactivity, cul- ture shock, and burnout) and assess its impact on POPN observational data.

2. FOUR TYPES OF OBSERVER BIAS

2.1. Reactivity

One common criticism of observational data is that research subjects may react to the presence of an ob- server by changing his or her behavior. [Note: This section has been omitted.]

2.2. Culture Shock

Immersion into an alien environment can be an intense and unsettling experience (Emerson, 2002b). Cul- ture shock refers to feelings of disorientation and anxiety that observers feel during the early stages of field research (see DeWalt and DeWalt, 2002, for a review). [Note. This section has been omitted.]

3Fax: 1-205-348-7178. E-mail address: [email protected] 0049-089X/$—see front matter. 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.ssresearch.2004.05.003

i

Chapter 1 • Introduction 23

2.3. Going Native

Gans (1982) argued that observers become sympathetic toward individuals under study for a number of ma-

sons. One reason is that observers and research subjects work together in close quarters over the course of

fieldwork. A natural consequence of the observer's task is a more sympathetic outlook toward individuals un- der observation. A second reason is that researchers feel a pull toward increased involvement while collecting observational data (see Adler and Adler, 1987, for an overview). Palmer and Emerson (2002) argued that re search subjects could use observers as an assistant and a resource while conducting their work. This level of involvement can range from innocuous requests for observers to act as a "go-fer" to participation in more crit- ical activities. For example, while studying drug dealers, Adler (1985) loaned drug dealers money, watched their children for extended periods of time, testified in court on their behalf, and allowed them to conduct drug deals in Aler's home. While observing police officers, Norris (1993) not only made tea for the shift, but also helped to chase and arrest suspects and administered first aid following suicide attempts. Since he did not look like a police officer, Skolnick (1966) aided police by walking into a bar to find a dangerous armed robber. Skolnick also helped some officers get past a lookout by driving a disguised truck up to a building.

A more general concern is that observing research subjects in their natural settings has the potential to re- socialize the researcher. This process of resocialization has been described as "overapport" and conceptualized as the problem of "going native," since observers lose sight of their role as a researcher and, in extreme cases, begin to think of themselves as members of the group under study. Observers may become involved as partici- pants and systematically influence the phenomenon under study (Manning, 1976; Pollner and Emerson, 2002). Reiss (1968, pp. 365-366) described one instance of observer socialivation that occurred during an observa- tional study of police. In aprecinct located in a high crime area populated predominantly by African-Americans, a common practice in the stationhouse was to walk down the row of cells and flush the toilets when an inmate asked for water. One evening, an inmate asked for water and an observer imitated the officers by walking down the cellblock and flushing the toilets. On his last shift, one officer asked the observer what he learned over the summer. The observer shocked a fellow observer by replying "I learned to hate niggers." In extreme cases, ob- servers could transfer their loyalty to research subjects, completely abandon data collection, and refuse to return from the field. Anthropologists who have gone native have stopped publishing material completely, rejected ac- ademia, and decided to live with indigenous people who were the topic of inquiry (Gronewold,1972).

2.4. Burnout

Burnout refers to inaccurate documentation of observational data at later stages of fieldwork due to the mentally and physically demanding nature of data collection (Fine, 1993; Lofland and Lofland, 1995). [Note: This section has been omitted.]

3. DATA

This study utih7es POPN observational data of patrol officers collected in Indianapolis and St. Petersburg dur- ing the summers of 1996 and 1997, respectively. Shifts were selected for observation based on time of day and day of the week, but busier days (Thursday through Saturday) were oversampled to maximize the number of police-citizen encounters. Observers rode with patrol officers in 12 of Indianapolis's 50 patrol beats and 12 of St. Petersburg's 48 patrol beats. Beats in both sites were selected to capture variation in service conditions within each jurisdiction. In addition, an attempt was made to match service conditions for beats selected from St. Petersburg and Indianapolis (Mastrofski et al., 1998). Thirty-seven observers accompanied patrol officers during 729 shifts and documented a total of 2488 encounters with suspects. Observers accompanied patrol of- ficers over the course of a shift and were trained to unobtrusively take brief field notes on their behavior.

4.1. Research Questions

4.1.2 GOING NATIVE This study will examine two research questions related to going native. First, are observers becoming socialized by officers and developing more positive attitudes toward the police over the course of fieldwork? Second, do these shifts in observer attitudes toward police (ATP) predict whether ob- servers help police officers and/or participate in police work?

5. RESULTS

5.2. Going Native

In Table 2, the impact of going native in POPN observational data is assessed by examining: (1) shifts in ob- server's attitudes toward police (ATP) over the course of fieldwork and (2) the probability that an observer will participate during fieldwork by helping with police work. Panel A in Table 2 addresses the first issue by examin- ing observer surveys (N = 32) that were completed before the start of fieldwork (Ti—see column one) and im- mediately following fieldwork (I2—see column two). Observer's ATP became more positive between Ti and T2 for the fast three out of four attitudinal measures, but this change is only statistically significant at the .05 level for one out of the three measures (see Panel A). For the fourth measure, there was no change in observer ATP

24 Chapter 1 • Introduction

TABLE 2.... Assessing the impact of observer.bias on police observational data: going native as a shift in observer's attitudes toward police and observer's level of participation impolite work (observer effects)

Before Fieldwork (TI) After Fieldwork (12)

(A) Observer's attitudes toward police (ATP): mean comparison Do police discriminate toward certain groups? (coded 0,1 where 1 = never or rarely) .13 .28 Would you approve if an officer struck a citizen who was attacking the officer with his/her fists? (coded 0,1 where 1 = yes) •84 •94 Would you approve if an officer struck a citizen attempting to escape custody? (coded 0,1 where 1 = yes) 59**,a •72 Would you approve if an officer struck a citizen who said vulgar or obscene things to the officer? (coded 0,1 where 1 = yes) .00 .00 N (number of observers who completed both Ti and T2 observer surveys) 32 32

Observer Effects

Observer Helps Police Observer Helps with OfficerC Police Workd

(B) The relationship between change in ATPb and observer effects: odds ratios from logistic regression Observers with more positive ATP (P2 >11) 50*,e .54 Observers with no change in ATP (12 = Ti) f _ Observers with more negative ATP (12 <T1) 1.25 .69 Initial ATP (before fieldwork) 2.07** 2.46** Model, 19.3*** N (number of shifts where observers completed both Ti and 12 observer surveys) 598 598

(A) Observer's attitudes toward police (ATP): mean comparison. °Two-tailed significance level indicated by a test of independence. bMeasures of change in ATP in Panel B were derived from an additive scale of the fast three observer survey questions from TI and T2 in Panel A. This more general measure of ATP at TI and T2 ranged from zero to three. To determine if the observer's ATP became more positive (T2 > TI), stayed the same (T2 = Ti), or more negative (T2 < TI), scores on this additive scale were compared as described above. `Dichotomous coding derived from descriptive data on reactivity (1 =rides where the observer helps patrol officer) (see Spano, 2003 for a more detailed description). `Level of statistical significance is derived from odds ratios from multivariate logistic regression. (Observers with no change in ATP is the comparison group. *p<.1 """p<.05 ***p<.01 i

Panel B in Table 2 addresses the second issue by investigating whether observers with more positive ATP are more likely to "participate" during fieldwork by helping police officers. An additive scale was devel- oped using the observer's first three attitudinal measures from Panel A from Ti and T2. Observer's ATP were categorized as more positive if this additive scale of ATP at T2 was greater than ATP at Ti. If there was no change in ATP, then T2 and Ti ATP were equal. Finally, observer's ATPs were categorized as more negative if T2 was less than Ti ATP.

Two measures of observer participation (or observer effects) were derived from a dichotomous coding scheme derived from descriptive and quantitative data on reactivity (see Spano, 2002, 2003, for a more detailed description). Column one in Panel B contains rides where observers helped police officers in any capacity. This measure includes more mundane and innocent forms of participation (e.g., the observer tells the officer that s/he left his/her lights on or offers advice when asked about whether to pull over a vehicle) to instances when ob- servers helped officers with police work (e.g., holding handcuffs or helping to fill out paperwork, arrest and/or detain suspects) (see column one in Panel B) (Spano, 2002). Column two in Panel B is a subset of cases from the q»aiitative coding in Column one where observers helped the officer with police work

Panel B examines shifts where observers completed both TI and T2 observer surveys (N= 598). Ob- servers with more positive ATP (T2 > Ti), no change in ATP (T2 = Ti), and more negative ATP (T2 <T1) were regressed on the two dichotomous measures of observer effects discussed above using multivariate lo- gistic regression after controlling for the observer's initial ATP before fieldwork. Observers with more pos- itive ATP are about 50% less likely to help police officers (see column one in Panel B). Although this

I I'

Chapter 1 • Introduction 25

relationship is marginally significant (p < .l), the direction of the effect is not in the expected direction. Ob- servers with more positive ATP are not more likely to help police officers or participate in police work, which would be consistent with going native. In fact, this finding suggests exactly the opposite.

6. IS RAPPORT A PROBLEM (GOING NATIVE) OR A SOLUTION (REACTIVITY)?

Another contradiction in the methodological field research literature relates to rapport. The development of rapport is essential to minimize reactivity to get research subjects to feel comfortable and trust observers, and to act naturally while under observation. Reactivity suggests that, for example, too little rapport be- tween police officers and observers can cause police officers to engage in atypical behavior and thus under- mine the validity of observational data collected during the early stages of fieldwork. In contrast, going native suggests the opposite: Too much rapport between observers and research subjects will affect the ac- curacy of data collected at later stages of fieldwork. Going native could be rationalized as a way of mitigat- ing reactivity by establishing a high level of rapport with those being observed. However, Miller (1952) observed that friendship is based on an all-accepting outlook that can limit the depth of an observer's inves- tigation into and criticism of the behavior of research subjects (see also Van Maanen, 1983a, 1983b). Thus, going native could limit the scope, the depth, and the accuracy of observational data.

1. SUMMARYAND CONCLUSION

Observers are a potential source of bias for two reasons. First, observers are part of the context of observed behavior. As a result, the research subject could react to the presence of the observer and behave atypically (reactivity). Second, observers are also data collection instruments. Thus, the quality of observational data could be undermined if observers become resocialized, contaminate observational data by participating in police work, and/or document inaccurate observational data (going native).

The relevance of these Endings is not limited to the field of criminology and criminal justice. Ethno- graphers, researchers who conduct face-to-face interviews, and child development researchers who "unob- trusively" study the interactions between parents and children (to name a few) all collect and utilize observational data to further our scientific understanding of various real-world phenomenon and use of force behavior, needed to be statistically significant. Second, the direction of the relationship between measures of observer bias and the outcome measures also needed to be consistent with hypothesized effects specified as research questions for reactivity, going native, and burnout.

The analyses of POPN descriptive and quantitative data focused on observer bias (going native) and revealed key findings. One key finding is the very limited support for observer bias in the form of going na- tive and burnout. On the surface, observers' ATP became more positive over the course of fieldwork in terms of absolute change, but the change in ATP was statistically significant for only one of four measures. In addition, findings from multivariate logistic regression were marginally significant (p < .1) and showed that observers whose ATP became more positive over the course of fieldwork were less likely to "partici- pate" in the form of helping officers for one out of two measures of observer effects. In contrast, the litera- ture and logic of going native would predict that these observers would be the most likely to participate in police work.

Future research should include additional control variables to rule out alternative explanations for patrol officer arrest and use of force behavior. To further explore the robustness of these findings, multi- variate analyses should be conducted to determine if observer bias is a significant predictor of, for exam- ple, arrest and use of force behavior net of a variety of control variables (e.g., officer work orientation, suspect characteristics, and situational factors). If the significant effect is suppressed in the "full" multi- variate model, then it is presumed that observer bias has not substantively affected the quality of police ob- servational data.

References

Adler, P.A., 1985. Wheeling and Dealing. Columbia University Press, New York. Adler, P.A., Adler, P, 1987. Membership Roles in Field Research. Sage Publications, Newbury Park, CA. Denzin, N.K., Lincoln, Y.S., 2000. The discipline and practice of qualitative research. In: Denzin, N.K , Lincoln,

Y.S. (Fels.), Handbook of Qualitative Research, second ed. Sage Publications, Thousand Oaks, CA, pp. 1-28. DeWalt, K.M., DeWalt, B.R., 2002. Participant Observation: A Guide for Fieldworkers. A1taMira Press,

New York. Emerson, R.M., 2002a. Introduction: the development of ethnographic field research. In: Emerson, R.M.

(Ed.), Contemporary Field Research: Perspectives and Formulations, second ed. Waveland Press, Prospect Heights, 1L, pp. 1-26.

Emerson, R.M., 2002b. Fieldwork Practice: issues in participant observation. In: Emerson, R.M. (Ed.), Contemporary Field Research: Perspectives and Formulations, second ed. Waveland Press, Prospect Heights, IL, pp. 113-151.