Trust in State Authority and Non-State Actors

The proposed research builds upon our pilot study (MIT Subaward Agreement No. 5710003). The two aims of the pilot were: (i) Developing effective information treatments; and (ii) Identifying feasible and appropriate services and target populations. We have now successfully developed and tested information treatments. Results from over 200 representative urban and rural respondents show that specific (accurate) positive informational primes on state actors do lead to positively (negatively) updating towards the state (non-state) actor. These results hold across subjective survey and experimental game outcomes. In addition, we have identified dispute resolution as a service where non-state actors are playing increasingly important roles and pose direct challenges to state authority. This has helped us develop specific "exposure treatments" pertinent to dispute resolution, establish partnerships with relevant players, and identify individuals at initial stages of a dispute, thus allowing us to measure the impact of the interventions on actual engagement in addition to survey and experimental outcomes.

Specifically, we propose to introduce three types of "exposure treatments" about dispute resolution services at the individual level. These treatments will vary in intensity from low to medium to high (from simple information provision about changes in effectiveness of the actor to real experiences with changed services). We examine how moving from low to high intensity treatments will impact beliefs about state (non-state) actors and movement away from non-state (state) options for dispute resolution. We will also study whether actual citizen engagement varies with the strength of exposure. Within each treatment type we also hope to introduce further variants that will shed light on when and why movements to state/non-state actors occur. We begin by discussing service and partner identification, move to sample identification, elaborate on the specific treatments in the study and then outline our data and methodology.

We focus on the most common types of disputes within the broader dispute resolution context: land-related and criminal disputes. The state actors that redress these disputes are the local courts (either civil or land revenue courts) and the police. Non-state actors are local entities operating outside of the state to provide dispute resolution services.2 In addition, we will utilize new state initiatives being undertaken by important state actors (the police and the land revenue departments) to improve service delivery for citizens. These initiatives have recently begun to be implemented and provide credible and new information for citizens.

The first state initiative is the introduction of a crime reporting helpline system by the police that can be used during periods of emergency, threat or victimization. The main component of this intervention, the helpline, records all citizen complaints and directs designated police officers to respond to these complaints.3 In addition, citizens have access to a Citizen Feedback Cell that they can call to report on the quality of the service. This police intervention removes the discretion of the police station in registering criminal complaints, lowers the cost of access for citizens and institute citizen feedback as a monitoring mechanism for police officers.

The second initiative is the introduction of a computerized Land Records Management and Information System (LRMIS)4 in Punjab that puts in place a reliable and transparent system for maintaining land records. The system records all transactions and provides access to these records for the relevant population at designated service centers. There are currently 98 (out of 143) service centers operational across 36 districts in Punjab, and the service is on schedule to further expand in the upcoming year. This initiative should reduce transaction costs related to the acquisition of ownership records and the recording of transactions and mutations thereby facilitating dispute resolution, and also removing local revenue officers’ discretion on the issuance of ownership.

We have secured partnerships with departments involved in implementing these services. Our first collaboration is with the Office of the District Coordination Officer (DCO) in Okara. The DCO is the highest civilian officer of the district and is the head of the district revenue department that deals with land related matters including adjudication. The DCO is also the first court of appeal on land related matters. The second collaboration is with the Deputy Inspector General (DIG) of Police Investigations in Lahore. The DIG leads dispute investigations for the district police service and has significant autonomy in devising interventions. These collaborations will allow us to access data which will form our sampling frame (see following section for further details), develop our treatment materials and allow us access to administrative outcomes data. In addition to these state players,

Foundation Open Society Institute-Pakistan (FOSIP) will coordinate with academics at Lahore University of Management Sciences (LUMS) and lawyers at a leading local firm to spearhead efforts to produce a legal triage guidebook (as described below) and provide access to a wide network of law students and lawyers for the intervention. Letters of support for all partners are attached.

B. Target Population
The identification of a sampling frame was another challenge we identified in our pilot proposal. Sampling from the general population is not efficient since it is likely that most may not currently have need of dispute resolution services. As such, while one would be able to examine the impact of the exposure interventions on their perceptions, doing so for actual engagement choices (state vs. non-state) will be harder. Instead, we would ideally like to identify individuals who have just experienced a dispute or are highly likely to experience one in the near future and are still in the process of deciding between state or non-state options. Following such individuals over time will be more likely to reveal whether the interventions impact actual engagement. Through our pilot, we successfully identified two sources through which we could create a sampling frame of individuals likely to be in–or already involved in—disputes. These two sources derive from the services mentioned previously: LRMIS and the police helpline system.

Through the LRMIS database, we can identify land records that are currently in dispute by querying records with ‘stay-orders.’ Stay orders are the first step in the registration and resolution of land disputes, and issuing one blocks any issuance of `fard’ (record of the land history and ownership). 5 Even though land cases can last a long time, there are frequent decision points where individuals choose state or non-state actors for resolution.

The second source for our study population is the call data from the police emergency helpline system. When a caller places a call to the helpline system to register an incident, the phone number is automatically registered in the system. The caller must also provide her name and an address for the complaint. The helpline receives roughly 5,000 calls daily in Lahore alone, out of which about 500 calls are deemed appropriate for follow-up (i.e. not blank or inquiry calls) by call operators. These 500 are further divided into types of offenses, and followed up by local investigating officers. These ‘appropriate’ calls will constitute part of our study population.

We hope to construct our sampling frame from these sources across rural and urban areas in Central Punjab in coordination with state authorities. We will use contact information available in administrative databases to find and offer individuals a chance to participate in the study.

A possible concern of our targeting method may be selection into the data sources (i.e. individuals in the database already show a predisposition towards the state). For the LRMIS database this is not a concern since individuals are automatically entered into the system using land revenue official records. Moreover, issuing a stay order is a necessary first step in case of a land dispute regardless of which dispute resolution avenue is pursued. Selection concerns could exist for the police helpline, though it is becoming fairly common for citizens to use the service. Nevertheless, we are planning to do a survey exercise where we will identify households in dispute through rapid community-based identification exercises. While this is a costly and more time-consuming process and so cannot be used to construct our primary sample, it will allow us to examine whether there is potential selection into either of our databases.

C. Exposure Interventions
We introduce three types of "exposure interventions" that will be delivered at the individual-level to a population randomly selected from the police helpline and the LRMIS systems.

In the low-intensity treatment, we will provide information about the effectiveness of state or non-state actors to respondents. Information provided will be typically 3-5 sentences long using real details based on local news articles and secondary data. We will develop two variations: One information treatment will provide positive information about state actors, while the second will provide positive information about non-state actors. The information treatments will be similar to ones we developed in the pilot study. We anticipate refining our more successful primes and developing new land revenue-related primes as part of a surveying exercise with respondents.

In the medium-intensity treatment, we will move beyond delivering simple information to making it more credible and persuasive by detailing new services—namely the LRMIS and the police emergency helpline system—and educating respondents about accessing them. In contrast to the low-intensity treatment, this treatment seeks to engage the respondents via direct demonstration of new and positive services. In visits to respondents, we will explain step-by-step all processes associated with availing these services and how beneficial they have been to others. In the case of LRMIS, we will demonstrate the system on a tablet but will recommend respondents visit service centers to further engage with the system. We will vary the intensity of this exposure by providing some respondents with detailed information on both services and others with information on just one; we refer to these variations as “state-heavy” and “state-light” from here onwards. While in the low-intensity treatment we had a non-state positive treatment variant, this is less feasible

2014-12-01

2016-09-30

Our set of interventions - moving from light to heavy treatments - will help map out the nature and extent to which differing degrees of (positive) exposure to state (and non-state) actors impacts state (non-state) trust and engagement. For each intervention evaluating the impact on both state and non-state actors will allow us to detect direct and spillover effects. In the light treatment, separately using positive state and non-state information, will allow us to examine whether the effects are symmetric or not. In the medium (and heavy) treatments, we will use the “state-heavy” and “state-light” variation to see the elasticity of trust in state and non-state actors with respect to varying levels of exposure of state activities. Moreover, since we will examine state-service specific perceptions and outcomes as well, we will be able to examine whether providing a positive exposure to one state service leads to a positive movement on the other service as well (a "service spillover" effect).

With states like Pakistan increasingly concerned about citizen disengagement, it is important to understand what factors promote trust in and re-engagement with the state. Whether positive policy innovations can help and if so, whether these need to be directly experienced by the citizen or the knowledge that the state is improving its effectiveness is sufficient. Since the trend away from state authority is not simply a Pakistan-specific phenomenon, the lessons from this study are likely to be of interest to a wide audience of policymakers and academics.

Respondents will participate in a field-based experiment to examine the effects of information treatments, which provide fact-based information on the behavior and ideology of state actors, on the level of trust and respect citizens hold for these state actors. The intent of the games is to identify the influences on citizens’ trust in state and non-state actors. They are being developed based on common behavioral trust games that ask respondents to make a monetary allocation with a given amount to various state and non-state actors. The experiment measures the change in their allocations given more positive or negative information relating to various the actors. Surveys will also be designed to gather more specifics in terms of the reasons for the allocation decisions of the players and also of perceptions of service provision by state and non-state actors. A double-blind design will ensure that allocation decisions cannot be traced back to each individual and that state and non-state counterparts will not have any knowledge of the identity of the respondents.

The study will recruit individuals through door-to-door household visits, use of land records system and lists of callers into the police helpline system as part of the study population to be given one of the three treatments or assigned to the control group. These systems are further explained below.

Along with the information treatment, two other treatments will be introduced to reduce barriers to accessing newly developed state services for respondents. These treatments are added to understand how additional exposure, not just information, to the state affects perceptions of state and non-state actors.

The second treatment will move beyond delivering simple information to making it more credible and persuasive by detailing new services—namely a computerized land record system and the police emergency helpline system—and educating respondents about accessing them. The computerized land registration management information system (LRMIS) in Punjab puts in place a reliable, efficient, and transparent system for maintaining land records. The system records all transactions and provides access to these records for the relevant population at designated service centers. The crime reporting helpline system by the police can be used during periods of emergency, threat or victimization. This system has been prevalent in Western countries for several decades (i.e. “911” service in the U.S. and Canada). The main component of this intervention, the helpline, will record all citizen complaints and direct designated police officers to respond to these complaints. In visits to respondents, we will explain step-by-step all processes associated with availing these services and how beneficial they have been to others. The first treatment consists of purely information about state or non-state actors (as was done in the first stage of the intervention). In contrast to the first treatment, this treatment seeks to engage the respondents via direct demonstration of new and positive services. In the case of land record system, we will demonstrate the system on a tablet but will recommend respondents visit service centers to further engage with the system.

The third treatment will directly reduce the transaction costs of engaging with state dispute resolution services. Specifically, we will offer "legal triage" to respondents. Initial evidence through the police helpline system shows calls with requests for legal advice. Our discussions with lawyers also suggest that such services would be of help and reduce the costs of accessing often burdensome state services. Trained professionals will develop a guidebook that will discuss resolution options for common types of cases and law students and/or paralegals will be trained and made available to assist respondents with their cases. While we haven't finalized the delivery logistics, options being considered include vouchers for consultation hours or a series of household visits by professionals. With respondents in all three treatments and a control group, we will collect data on perception of state and non-state actors, results of behavioral game play and use of state and non-state services. Additional surveys will be developed to be used at baseline and endline of these interventions. These documents will be added as an amendment when they are finished being prepared.

Details of Game Format
Two games that will be primarily used to test changes in perceptions of trustworthiness of state and non-state actors when given information treatments include the fund dictator game and the investment game.

Fund Dictator Game: Individuals are told that a given state or non-state actor has put an unknown portion of a given amount in a sealed envelope. The game allows participating individuals to make a choice between a fixed amount and this envelope given by the state or non-state actor.

Investment Game:
This allows participants to anonymously and confidentially invest some, or, or none of a specified amount representing their day’s income in a hypothetical scheme run by a type of state or non-state actors, knowing that their return is based on a measure of the actors’ effectiveness.

We will be using a randomized controlled trial (RCT) design with each of the three proposed types of treatments offered to a randomly selected subset of individuals. We expect to estimate “intent-to-treat” effects for all of our treatments. Our general specification will be:
Y_it= ∝ + β_1∙〖Treat〗_i+ β_2 ∙〖Post〗_t+ β_3 ∙〖〖Post〗_t×Treat〗_i+ ε_i
Y_it is the outcome variable for individual i at time t ∈(0,1). 〖Treat〗_i is the treatment dummy, and 〖Post〗_t is the post-treatment period dummy. β_3 is our coefficient of interest giving us the average treatment effect of the intervention. CERP will manage all surveying activities.

We will focus on a range of outcome variables using a combination of survey, experimental games and administrative data. We are interested in capturing subjective outcomes such as perceptions of state or non-state effectiveness or preference towards state or non-state actors as well as real outcomes such as take-up of state services, frequency of usage of state-services or dispute status. We will administer baseline surveys to all respondents and conduct follow-up surveys to track take up and usage of state initiatives as well as the status of disputes over a 1-2 year period. We will access administrative data to verify certain self-reported outcomes on state service usage. In addition to outcome measures, we will collect basic information on households (age, income, etc.) to serve as control variables if necessary (e.g. in cases of imbalance in key variables from randomization).

In the pilot study, we observed effects of roughly 0.2 standard deviations (sd) for experimental game outcomes from a simple information provision treatment. To conduct power calculations for the full study, we take this effect size as the benchmark for the low-intensity treatment; as we increase the treatment intensity, we expect stronger effects (0.25 sd for the medium-intensity and 0.4 sd for high-intensity treatments). Based on these assumptions, power calculations reveal that we will require a total of 2,100 respondents, with 500 respondents for the low-intensity treatment, 450 for the medium-intensity treatment, and 500 for the high-intensity treatment (assuming 50 percent take up, though we expect it to be higher), and 650 respondents in the control group.

We will identify neighborhoods with low to middle income households and visit individual households to gather a sample population to participate in the games. Respondents will also be recruited using the land registry and police helpline call registry.

Respondents will be informed of the nature of their involvement in the study at the time of recruitment. Subjects will be randomly assigned by computer into treatment and control groups after they give consent based on the ID number on their survey number that represents a randomly generated order of numbers pre-assigned to treatment and control. The rules of the games, their compensation, the double-blind design assuring anonymity, and their ability to leave at any time will be explained before they play. The game design will be piloted with the first group of individuals and any lessons learned will be used to modify future games.

Randomization is at the individual level

No

0 clusters planned

2,100

500 respondents for the low-intensity treatment, 450 for the medium-intensity treatment, and 500 for the high-intensity treatment (assuming 50 percent take up, though we expect it to be higher), and 650 respondents in the control group. Total 2,100.

In the pilot study, we observed effects of roughly 0.2 standard deviations (sd) for experimental game outcomes from a simple information provision treatment. To conduct power calculations for the full study, we take this effect size as the benchmark for the low-intensity treatment; as we increase the treatment intensity, we expect stronger effects (0.25 sd for the medium-intensity and 0.4 sd for high-intensity treatments). Based on these assumptions, power calculations reveal that we will require a total of 2,100 respondents, with 500 respondents for the low-intensity treatment, 450 for the medium-intensity treatment, and 500 for the high-intensity treatment (assuming 50 percent take up, though we expect it to be higher), and 650 respondents in the control group.