00:00 Welcome and the floor is yours.
00:02 Thank you and thank you for,
00:03 for the opportunity to present this work.
00:05 Uh,
00:05 so today I'll be talking about local elites as state capacity,
00:08 how city chiefs use local information
00:10 to increase tax compliance in the DR Congo,
00:12 and this work is joined with Pablo and Augustine
00:14 who are at Harvard and John who's at LSE.
00:17 So before I begin,
00:18 I wanted to give you a brief overview of
00:20 how this work fits into my research agenda,
00:22 which contains two focuses.
00:24 First,
00:24 one on the role of fiscal capacity and development.
00:27 And this includes work on the tax tools
00:29 at the disposal of very low capacity governments
00:32 and how those tools set a ceiling on the government's ability to raise revenues.
00:36 It also includes work on a national scale looking at how tariff
00:40 rates,
00:40 um,
00:41 and domestic consumption tax bases interact and how,
00:44 um,
00:44 trade taxes affect domestic production networks in Rwanda.
00:48 Um,
00:48 and also a focus on the role of actors outside the state system
00:52 and what they can bring to fiscal capacity and state capacity more broadly
00:56 in the realm of taxation and redistribution.
00:58 My other focus is on constraints that firms in emerging markets face.
01:03 And includes work in particular on labor market frictions.
01:06 Um,
01:06 in India,
01:07 in which I partner with a large online jobs platform
01:09 to run experiments studying frictions in the recruitment process,
01:12 but also work on behavioral biases within firms that include,
01:16 uh,
01:16 peer norms and heuristics and
01:19 decision making,
01:20 uh,
01:20 specifically in the realm of taxation.
01:22 And it's this third project under the
01:24 fiscal capacity focus area that I'll be talking about today,
01:27 the role of local elites in building state capacity.
01:30 And the motivation for this work is first that fiscal capacity,
01:33 the ability to collect and spend revenues effectively,
01:36 is an important input to economic and political development
01:39 in poor countries.
01:40 However,
01:41 however,
01:41 lower-income countries often find themselves in low tax,
01:44 low capacity traps.
01:45 Um,
01:46 and so a question for them is how can they collect sufficient revenue.
01:49 To fund public goods
01:50 in a sustainable manner and thus convert themselves into tax states.
01:54 And often in these settings,
01:56 tax collection is still done on a manual basis
01:59 and so the government faces a choice in
02:01 how to go about administering manual tax collection,
02:03 specifically in the type of agent to enlist in collection.
02:07 And the government can choose to enlist
02:09 state agents or bureaucrats in tax collection
02:12 or to delegate this responsibility to outside actors,
02:15 outside of the state system,
02:16 um,
02:17 such as those that are local elites,
02:19 um,
02:19 or informal actors.
02:22 And these local elites can
02:24 um fill many roles uh
02:26 outside the states that are often um contributing towards
02:29 filling gaps in governance outcomes and they
02:31 can take the form of customary leaders,
02:33 leaders in religious spheres,
02:34 or even in the military space.
02:36 And what we observe,
02:37 observed throughout history,
02:38 um,
02:38 and even in many low-income countries today is that governments on the
02:42 lower end of the capacity spectrum often delegate
02:45 collection to such actors outside the state.
02:48 And in making this choice,
02:49 the government will encounter certain trade-offs when making a
02:53 decision over the agent in charge of tax collection.
02:56 In the advantages column for uh local elites,
02:59 we often think that they might possess better local information about taxpayers
03:03 that could be a key input to generating compliance,
03:05 especially in um settings of imperfect uh compliance that because
03:09 these elites um know community members well,
03:12 they're able to target solicitation or they're more effective
03:14 in the bargaining game over taxation given they can leverage
03:18 potentially private information about
03:20 citizen assets or citizen income.
03:23 It's also thought that um local actors
03:26 um also have simply lower administrative costs.
03:29 You could think about uh
03:30 the government's decision to either staff a tax office in every province or
03:33 to enlist someone who's a leader in that community already to collect taxes
03:37 or to deploy state agents to every area in
03:39 a region and thus incur substantial transportation costs.
03:44 In the disadvantages column,
03:45 however,
03:46 because these actors exist outside of the state system,
03:49 almost by definition though they will be harder to control,
03:51 um,
03:52 and monitor,
03:53 and this could give rise to greater leakage and corruption,
03:55 especially in tasks like tax collection where money is changing hands,
03:59 and this could manifest in ways that cause real economic harm.
04:03 And second,
04:04 uh,
04:05 a concern that may speak to more longer run concerns
04:07 about enlisting or empowering actors outside of the state,
04:11 um,
04:11 is how
04:12 raising up these actors could undermine the
04:15 legitimacy and authority of central government institutions
04:18 or the central government itself,
04:19 that by making the central government look weak and that it
04:22 needs to extend the power to collect taxes outside of itself,
04:26 that could undermine in the long run um how strong the government is
04:29 perceived by citizens and how effective it can be in other tasks.
04:33 So this project asks two questions.
04:35 First,
04:36 can low capacity governments increase tax revenue by delegating
04:39 this collection to local informal actors outside the state?
04:43 And if they are more effective,
04:44 is it because these actors possess superior information about citizens,
04:47 or is it due to other margins like,
04:50 um,
04:50 these agents being able to put more effort into tax collection
04:54 or conditional on
04:56 level,
04:56 levels of effort being the same?
04:57 Are they simply more persuasive by virtue of,
05:00 um,
05:00 their position outside the state?
05:03 And second,
05:03 on the other hand,
05:04 does engaging these actors backfire
05:06 in ways that damage government legitimacy.
05:08 Specifically,
05:09 does empowering local elites allow them to extract
05:12 more in the context of taxation but also outside of it,
05:15 that by giving them the power over tax collection,
05:18 do they therefore extract more in other informal
05:21 tasks they undertake in their normal duties?
05:24 And then another dimension that the government may care about,
05:27 thinking about the government as social planner and the administration of
05:30 taxation
05:31 is who bears the ultimate burden of tax.
05:33 And so a relevant question is how does endlisting these actors shift the
05:36 burden to be more regressive or progressive
05:39 relative to collection by state agents.
05:42 And I studied these questions in the context of a field experiment
05:46 that was embedded in the 2018 property tax campaign in the city of Kananga in the DRC.
05:51 This experiment was implemented by the provincial
05:53 government of Kasai Central and involved assigning
05:56 neighborhoods of which there are 356 covering the entire city,
06:00 uh,
06:00 to one of 3 treatment groups.
06:02 In the first treatment group,
06:03 it's state agents that are
06:05 tax collectors from the provincial tax ministry,
06:08 uh,
06:08 that are tasked with collection in the neighborhood,
06:10 and I'll refer to this as the central treatment arm.
06:13 The second treatment arm collection was done by city chiefs that live and work and
06:17 have other informal details in the neighborhoods
06:19 in which they'd be tasked with collection,
06:20 and I'll refer to this as the.
06:23 In the third treatment arm that I call central plus local information,
06:26 is a hybrid arm that was designed in collaboration with the government
06:29 to shed light on the potential information mechanism behind,
06:32 uh,
06:33 state,
06:33 uh,
06:33 city chiefs' greater effectiveness
06:35 that allows state collectors first to consult with the neighborhood chief in
06:38 charge of an area before they those state collectors would collect there.
06:44 And to preview the results,
06:45 in answer to the first question about
06:47 the ability of local elites to raise revenues,
06:50 I find that the answer here is yes,
06:51 at least in this context.
06:53 The local treatment arm increases tax compliance and total revenues
06:56 relative to collection by state agents significantly.
06:59 These amount to approximately a 43%
07:01 increase in tax collection relative to collection by state agents.
07:06 And the evidence is consistent with these effects being driven
07:09 by an informational advantage on the part of chiefs,
07:12 chiefs rather than other margins like effort or persuasive ability
07:16 conditional on effort across the treatment arms being the same.
07:19 Uh,
07:19 the main piece of evidence.
07:22 An arm I mentioned,
07:23 uh,
07:23 just before raises compliance relative to collection by state agents.
07:27 So armed with
07:28 the chief's information,
07:29 state agents perform better,
07:30 but I also find other evidence to rule out
07:32 these other channels of effort and persuasive ability.
07:36 Then on the other hand,
07:37 I do find some evidence that
07:39 delegating collection responsibilities increases corruption,
07:42 uh,
07:42 in the areas where chiefs collect taxes relative to state agents.
07:46 Bribes are higher in these areas,
07:47 uh,
07:48 to a small degree.
07:49 However,
07:49 I don't observe a more informal resource extraction in
07:53 the chief's other duties that now that they have,
07:55 um,
07:55 at least in the short run gained the power over taxation,
07:58 it doesn't appear that
07:59 In the administration of informal labor taxes,
08:01 which is their main other duty in the neighborhood,
08:03 they're extracting more or requiring
08:05 more contributions.
08:09 In looking at the distributional implications,
08:11 it does appear on face that chiefs
08:13 are more regressive in their enforcement of the tax.
08:16 Um,
08:16 when we look at house quality,
08:18 which is maybe the,
08:18 the most direct measure
08:20 that maps to the ultimate base of this tax,
08:22 which is property value,
08:23 we see that chiefs are bringing lower value properties into the tax net.
08:26 But when I look at measures that more directly relate to economic need like income,
08:30 liquidity,
08:31 and well-being,
08:32 I don't observe any differences across the treatment
08:34 arms among the complier set of taxpayers.
08:37 Um,
08:37 so I'll return to this question of what we may think should be the
08:40 ultimate base in,
08:41 uh,
08:41 settings like this of taxes like,
08:43 uh,
08:43 those on,
08:44 um,
08:44 assets,
08:45 uh,
08:45 like property.
08:48 And this evidence on the distributional um
08:50 aspects also speak to the information mechanism.
08:54 It appears that state collectors target
08:56 households by observable characteristics like the
08:58 quality of the roof and walls of structures on the property,
09:01 whereas chiefs,
09:01 though they also target on these characteristics,
09:04 additionally target on what I call unobservable characteristics like
09:07 liquidity,
09:08 the employment status of the,
09:09 of the property owner that are measures of ability to pay in addition to
09:13 house quality.
09:16 And then finally relating to these longerrun concerns about how
09:19 empowering actors outside the state may undermine the government's claims on
09:23 authority or legitimacy in the administration of taxation,
09:26 I don't find any evidence.
09:28 At least in the short run,
09:30 that erosion towards
09:31 the govern the uh views of the government erode um at
09:34 all levels of government from the provincial to the national level,
09:37 and indeed,
09:38 uh,
09:38 citizens appear to have more positive views of the
09:40 chief in areas where chiefs collected relative to state agents
09:43 and would prefer to have the,
09:45 the chief serve as tax collector in the future.
09:50 In terms of the literature that this work contributes to,
09:53 I see it as contributing to two main strands.
09:55 The first is the role of fiscal capacity in fragile states.
09:58 It's long been thought that,
09:59 um,
10:00 fiscal capacity is important for the development
10:03 of
10:03 economic,
10:04 um,
10:05 uh,
10:05 and political institutions as well as um important for stimulating.
10:09 Participation in political processes
10:10 and recent work has documented the role of the choice of tax instruments,
10:14 how to incentivize collectors
10:16 in manual tax collection,
10:17 and how third-party information can be leveraged.
10:20 And where I see this work contributing is by showing
10:22 that the choice of the type of agent in tax collection
10:25 can matter a lot for the government's ability to raise revenues.
10:28 And the second strand of literature to which I see this project contributing
10:32 is the role of local elites in governance and development more broadly,
10:35 which has included areas
10:37 um from land governance and law and order to the redistribution side of public goods
10:41 provision and targeting of welfare programs to
10:42 even the political space of brokering votes.
10:45 And here I see the
10:46 the evidence I'm able to provide on the mechanism of chief
10:49 superior information allowing them to be more effective in tax collection.
10:53 Um,
10:54 we,
10:54 uh,
10:54 provide a link to potentially other literatures to the extent to which
10:57 this mechanism can apply and is at work in those settings.
11:02 So a brief outline for today's talk.
11:04 First,
11:04 I'll describe the
11:06 experimental design in detail and then show you the main
11:08 treatment effects on tax as well as non-tax outcomes.
11:11 Then I'll spend much of my time um building the evidence on mechanisms
11:15 to convince you that it's really about how chiefs use their local information
11:18 that is driving their greater effectiveness.
11:20 Um,
11:21 then I'll talk about the distribution of tax burden,
11:23 tax visits as well as the ultimate tax
11:25 burden and conclude and discuss some next steps.
11:30 So,
11:31 uh,
11:32 this project is based in Kanga,
11:33 which is the 4th largest city
11:35 in the DRC.
11:36 It has a population of 1 million people,
11:38 and it's the capital of the Kasai Central
11:40 Province and seat of the provincial government.
11:42 This is a very high poverty setting.
11:45 Median income is only $1.5 US dollars per person
11:48 per day.
11:49 Um,
11:51 and the DRC is a weak state both at the national and provincial levels,
11:54 and as you might expect,
11:55 it has a very low formal tax compliance.
11:58 In Kasai Central,
11:59 the province in which,
12:00 uh,
12:00 this experiment is based,
12:01 uh,
12:02 there are 6 million people live there,
12:04 and on an annual basis,
12:05 the government only raises about 30 cents per person,
12:08 uh,
12:08 in total revenues.
12:10 And so recently over the last 4 years,
12:13 the government has been trying to raise revenues through a property tax,
12:16 and the reason for emphasizing this tax instrument relative to others
12:20 is first that the revenue stays local.
12:22 So um for taxes on things like interprovincial trade,
12:25 the government needs to share a portion
12:27 of those revenues with the national government,
12:29 whereas all of the money from property
12:30 taxation stays within the provincial government.
12:33 It's also thought to be a more efficient form of taxation relative to
12:37 other forms because it's levied on at very low rates across very broad
12:41 bases.
12:44 And the government first uh implemented a systematic property tax collection
12:49 campaign in 2016 that involved sending state agents door to door to collect taxes,
12:53 and this raised compliance relative to a system of voluntary compliance
12:57 in which citizens were expected to pay at the bank themselves,
13:00 um,
13:00 and less than 1% of citizens complied with the tax under that system.
13:05 And so believing that chiefs may be more effective collectors,
13:08 the government decided to involve
13:09 uh chiefs in the 2018 campaign,
13:12 um,
13:12 thinking that they could further expand the tax net.
13:17 And the details of the tax that's uh levied in the context of this campaign
13:21 uh are,
13:22 are that it's the first uh fixed annual fee,
13:25 and this fee nature of property taxes is common in lower-income countries where
13:29 property evaluation rules don't exist because they're costly
13:32 to construct as well as to keep updated.
13:34 Um,
13:34 in contrast to more developed countries.
13:36 Settings where we observe more complex and progressive formulas that apply um
13:41 uh algorithms based on the size of the land area,
13:44 the quality of the materials,
13:45 etc.
13:46 Here the government charges one rate
13:48 um in two broad bins.
13:50 So there's some effort at classifying properties,
13:52 but
13:52 they classify them into two categories.
13:55 And the lower value band captures 90%
13:58 of properties in this setting,
13:59 uh,
13:59 and these are properties that have
14:01 less developed land area.
14:02 The structures on the property are made of lower quality materials like mud,
14:05 brick,
14:06 or clay,
14:07 um,
14:08 and these are assigned an annual tax liability of about $2 US dollars,
14:12 um,
14:12 and the property value in this category is about $1000 US dollars.
14:16 The remaining 10% of properties falls into the high-value band category,
14:19 um,
14:20 and these are properties with more developed land area,
14:22 built structures are built out of higher quality
14:24 materials like concrete and sheet metal roofing,
14:27 and property value in this category is about,
14:29 um,
14:29 $8000 US dollars,
14:30 and they face an annual tax liability of $9.
14:33 So you may be struck by how low the overall levels of tax liability are here,
14:38 between $2.09 dollars US dollars.
14:41 But when we look in terms of the percentage of property value they represent,
14:44 they're not that different from what we observed.
14:46 In more developed country settings.
14:47 So because no valuation rule exists,
14:49 um,
14:49 I use a machine learning strategy to estimate property value in this context
14:53 and show that it's a point that the rates are about 0.32%
14:57 of total value,
14:58 um,
14:58 which is within the range in the US where property value,
15:02 property taxes,
15:03 uh,
15:03 tax rates range between 0.27% in places like Hawaii to 0.35%
15:08 in places like New Jersey.
15:11 For those that declined to pay the tax,
15:13 the government officially imposes a penalty for non-compliance that is
15:17 a fine of 1.5 times the tax amount plus
15:19 the original tax amount they were supposed to pay,
15:22 and this is communicated through a summons to appear
15:24 at court and either contest or pay the fine.
15:27 So while this is the penalty on the books,
15:28 in practice,
15:29 enforcement is extremely uncommon.
15:31 It usually only happens for um
15:33 firms that own property in the downtown area that have
15:35 large outstanding arrears where the government thinks it can make up
15:38 a lot of revenue by enforcing the tax there and that those firms are likely to pay.
15:43 But at baseline,
15:43 52% of citizens think that sanctions are likely for not
15:47 complying with the tax,
15:48 and this will be a channel that I'll test explicitly
15:51 um by studying how influencing
15:53 citizens enforcement beliefs through primes and reminders about the
15:55 penalty affect the compliance decision and whether that's different,
15:58 uh,
15:59 depending on the type of agent who's collecting taxes.
16:04 So what this campaign looked like in practice,
16:05 uh,
16:06 it involved two stages by a team of 2 collectors in each neighborhood.
16:10 In the first stage,
16:11 the collection team visited every property in the neighborhood to
16:13 construct the property register because no evaluation rule existed.
16:17 During this step,
16:18 they would assign a unique tax ID
16:21 and provide a tax letter to the property owner informing them about tax liability,
16:25 um,
16:26 and our enumerator accompanied them during this visit to record
16:29 information about the property as well as the tax ID,
16:31 uh,
16:31 so that it could be linked to surveys as well
16:34 as the administrative data on tax compliance later on.
16:37 During this visit,
16:38 they could solicit the text if needed or make appointments uh for a next revisit.
16:43 The second stage of the campaign involved the collection
16:45 team making door to door appeals at their discretion,
16:48 and the enumerator did not accompany them during this stage.
16:51 This stage approximately lasted 1 month in each neighborhood,
16:55 and collectors would approach households and request
16:57 that they pay,
16:58 um,
16:58 or make appointments to revisit them later on.
17:01 And when they found someone willing to pay,
17:02 they would use a handheld receipt printer to issue receipts,
17:05 and these receipt printers create an electronic
17:07 record of tax payments that forms the
17:10 administrative data on tax compliance,
17:12 and they're matched to the collections that
17:13 these collectors hand in to the tax ministry
17:16 on a weekly basis,
17:17 uh,
17:18 and they earn a piece rate compensation for the amount of tax they deposit,
17:21 about 30% of
17:23 what they collect,
17:23 they receive as a bonus.
17:26 And so the experiment here is to compare collection done by tax ministry agents
17:30 to collection done by city chiefs when they're assigned as tax collectors.
17:35 And I'll give you a little bit more background
17:37 about what these uh treatments look like in practice.
17:40 In the central treatment arm,
17:41 it's agents of the provincial tax ministry that collect,
17:43 and these are unsalaried contractors with the tax
17:45 ministry and other parts of the bureaucracy.
17:48 So they often work on short uh work on short-run
17:50 campaigns of this nature and rotate between different uh bureaucracies.
17:54 And so their incentive for working on the campaign
17:56 is the um percentage of the revenues that they get to take home as a bonus but also the
18:02 um possibility that the highest performing agents could
18:04 receive a future salary position in the government,
18:07 which is a highly sought after position in
18:08 the city because it provides a regular salary.
18:14 So the neighborhood is randomly assigned to the central treatment arm.
18:17 These,
18:18 uh,
18:18 state agents will travel from their home
18:20 on a daily basis to the tax ministry and go
18:22 and collect in the neighborhood and return to their home,
18:24 typically in a different part of the city at night.
18:26 By contrast,
18:27 in the local treatment arm,
18:28 it's the city chiefs that live and work in the neighborhoods
18:31 that they're tasked with collection,
18:32 and these city chiefs are local notables or leaders.
18:36 That have other informal responsibilities in the neighborhood.
18:39 They help resolve disputes over issues like property and their
18:42 main task is the administration of an informal labor tax
18:45 by which citizens contribute labor on a weekly
18:47 basis towards local public goods projects in the neighborhood
18:51 like repairing of roads,
18:52 and this labor is organized and directed by the chief.
18:56 They're officially approved in their position by the mayor,
18:58 uh,
18:59 um,
18:59 of the city,
19:00 but
19:00 in fact the mayor has little discretion over who's selected
19:04 as chief.
19:04 The main qualification is being a well-known,
19:07 respected,
19:07 longtime resident
19:09 of the neighborhood,
19:10 and these chiefs have on average pretty long tenure,
19:13 about 10 years,
19:14 so it can be much longer than that.
19:16 And it's even the position is even sometimes heritable within families
19:19 in about 1/5 of cases.
19:21 They don't receive any salary in their position as chief.
19:24 Indeed,
19:24 most have another job,
19:25 they own a business or they're a teacher or a pastor,
19:28 and so the compensation that they receive as chief is primarily the status
19:31 as chief and any rents that they could potentially extract through their position.
19:36 So you can think about these chiefs being not
19:37 customary chiefs in a traditional sense but perhaps neo-customary.
19:41 They have some of the trappings of more powerful rural chiefs,
19:44 uh,
19:44 in Africa,
19:45 but not quite as much um authority over the daily
19:47 lives of citizens because they operate in an urban setting.
19:50 However,
19:50 they're a common institution in cities in Francophone Africa and
19:53 importantly they're often involved in tasks like property tax collection.
19:57 Um,
19:58 so I'll just talk,
19:59 uh,
20:00 a little bit about,
20:00 um,
20:01 how we make these comparisons across the treatment before I pause,
20:04 um,
20:04 for questions after this slide.
20:05 So what is held constant across these treatments is
20:08 all the nuts and bolts of the campaign in each
20:10 treatment group.
20:11 This includes the training collectors receive,
20:13 the protocol they use in,
20:14 in tax collection,
20:15 and the technology that they use when they collect.
20:18 Incentives are identical,
20:19 the same piece rate compensation for
20:21 the tax of the deposit that's equal to about 30%
20:24 of collections,
20:25 and additionally,
20:26 the government held fixed expectations
20:28 about future tax collection responsibilities,
20:30 um,
20:31 to avoid biasing effort in the context of this campaign.
20:35 What varies then is everything that attaches to the identity
20:37 or type of agent in charge of collecting taxes.
20:40 This can include some of the potential mechanisms or advantages and
20:43 disadvantages I spoke about earlier like
20:45 differences in information about taxpayers,
20:48 perceived legitimacy on the part of citizens or
20:50 the tools of sanctioning that are at the disposal of either collector type.
20:54 It also comes with some demographic differences.
20:57 Chiefs are on average older,
20:58 less educated,
20:59 and less wealthy.
21:00 Um,
21:01 they're also more skeptical of the government and taxation in general.
21:03 These are differences we see as hard-baked
21:06 into the definition of the chief.
21:07 Um,
21:07 so now I'll pause in case there are,
21:09 uh,
21:09 any questions.
21:12 Um,
21:12 I see one hand,
21:13 David,
21:13 go ahead.
21:15 Yeah,
21:15 I just wanted to ask about your 0.3 here about the sort of time horizons.
21:19 It sort of seemed like the Chiefs,
21:21 you were saying they're longer term positions.
21:23 It would seem like their time horizons would
21:25 Be
21:26 different and and you know,
21:29 it might take time for
21:30 the economy to adjust to this new system and things,
21:33 but I,
21:34 I,
21:34 I guess,
21:35 what does it mean to hold constant
21:38 expectations when people have different
21:40 time and job,
21:41 different,
21:41 uh,
21:42 you know,
21:42 incentives to engage with the population on the long-term,
21:45 etc.
21:46 Yeah,
21:46 that's a great question.
21:47 So by holding constant expectations,
21:49 I mean how the government communicated
21:51 whether or not a collector type would be involved in future campaigns.
21:55 So they told them this is how taxation is going to happen
21:58 in this campaign and we'll,
22:00 um,
22:00 see how each collector type performs and
22:02 then make assessments about future campaigns.
22:04 So that might have implicitly,
22:05 implicitly raised some concerns about competition,
22:08 uh,
22:08 between the collector types in the short run context of this campaign.
22:11 So we exploit some strategies in the timing of
22:15 Where tax collection happens
22:16 across the city and its rollout to address that concern,
22:20 um,
22:20 but I,
22:20 I think you're right that these,
22:22 these agents have different long-run considerations,
22:24 especially when it comes to thinking about
22:26 their relationship to citizens in the neighborhood.
22:28 Chiefs will have to live and work and,
22:30 um,
22:31 interact with,
22:31 uh,
22:32 citizens in their neighborhood.
22:33 Over a much longer horizon than these state agent collectors,
22:37 and so those are partly differences we see as like
22:40 built into the definitions of the treatments that
22:42 governments are making this decision in other spheres
22:44 to outsource um uh or like delegate this
22:47 responsibility to other actors that have stronger connections to
22:51 The community that comes with some advantages potentially in information or
22:55 sanctioning power,
22:56 um,
22:56 but it may come with some costs
22:58 that I haven't enumerated yet like,
23:00 uh,
23:00 not just in terms of corruption but also how those
23:02 actors may be constrained by their longer-run relationships with these,
23:06 uh,
23:07 with citizens.
23:07 Um,
23:08 does that answer your question?
23:09 Yeah,
23:10 but I guess just on that,
23:11 what,
23:12 you know,
23:12 what does that mean for the right time horizon to look for effects as well?
23:16 Uh,
23:16 I see,
23:17 yes,
23:17 um,
23:18 so
23:19 as,
23:19 uh,
23:19 since this is only a short run one-off campaign,
23:22 I can only show the like short run effects of
23:25 engaging chiefs in
23:26 the only the second time that the government has collected
23:28 taxes in this context in a door to door fashion
23:31 and so thinking about how chiefs may be more effective in the future or if
23:35 They might have like um be playing some
23:37 dynamic game by which they're asking citizens to contribute
23:40 now so they can win the authority and be
23:41 more progressive and tax taxation in the future.
23:44 Um,
23:44 I can't speak to the online data we collect is
23:47 about a year after the end of the tax campaign
23:50 and there I look at the other duties the chief is undertaking like
23:53 in the administration of their normal duties,
23:55 specifically in the implement the administration of the informal labor tax,
23:59 and there I don't see any differences across.
24:02 The areas where chiefs collected versus did not in um their other duties,
24:06 um,
24:06 and a follow-up project that I and my co-authors are working on
24:08 studies chiefs' involvement in the targeting
24:10 of anti-poverty benefits by whether they
24:13 engage in taxation or not,
24:14 and there we don't see any differences in
24:16 how the chiefs are um going about administering this like follow-up,
24:19 um,
24:19 fiscal task even though it's on the other side of the fiscal equation.
24:22 So I know that's not a perfect answer,
24:24 um,
24:24 but definitely a concern that's in mind here.
24:28 Thanks,
24:28 Gabriel.
24:29 I see two more hands.
24:30 Biju,
24:31 you go first.
24:32 Yeah,
24:32 hi.
24:33 Um,
24:33 so just to,
24:34 I mean,
24:34 it's related to David's question,
24:35 but first a clarifying question.
24:37 These revenues that these chiefs are collecting,
24:40 uh,
24:40 do they go towards local budget or they just,
24:42 uh,
24:43 taken centrally?
24:45 Uh,
24:45 so they go towards the provincial government's budget,
24:47 which the provincial government can spend anywhere in the province.
24:50 OK,
24:51 so you've got basically
24:53 an enforcement mechanism where local
24:56 Elites who could range from good people to utter thugs
25:00 are being used by the government
25:03 to extract money from regular folks
25:07 with no guarantee that that money is going to come back and benefit
25:10 them in the form of public goods or whatever.
25:12 That's what you're randomizing on.
25:14 I mean,
25:15 and then you're looking only at short term impacts.
25:17 Do you worry about the ethics of what you're doing?
25:20 Um,
25:20 so I think there's two,
25:21 I'll separate this into two concerns.
25:23 I guess the first concern is whether,
25:25 um,
25:25 our involvement stimulated the government's choice to do this.
25:28 So,
25:29 uh,
25:29 from the beginning,
25:30 the government wanted to involve chiefs thinking they would be more
25:33 effective collectors and asked us to participate in
25:36 the randomization exercise,
25:37 but I think that doesn't fully answer your question about whether it's ethical
25:40 to even engage in studying this and
25:42 helping the government implement this effectively.
25:45 Um,
25:45 and,
25:46 uh,
25:46 I think part of the reason that we felt comfortable participating in this.
25:50 Uh,
25:51 in the evaluation of this experiment is that
25:53 we could collect data and show the government
25:55 how much the like risks on the negative
25:57 side would be by collecting data on corruption and
26:00 extraction and the chief's other duties,
26:02 and I think the only answer I can give you
26:04 here that may speak to the longer run concerns about
26:07 Um,
26:08 engaging local actors that the state is less
26:10 able to control and monitor effectively like chiefs
26:13 is the follow-up projects we're doing studying chiefs' involvement and
26:16 subsequent tasks like the targeting of anti-poverty benefits and there's,
26:20 we're able to track more over the long
26:22 run whether empowering the chiefs in this way can
26:25 um lead them to have more extractive power or
26:28 be more um I guess negative actors
26:31 because they collected taxes.
26:32 But what about the delegitimizing of government?
26:36 That's a long term effect that could come.
26:37 I mean,
26:38 you,
26:38 you're now dealing with these chiefs who are doing other functions.
26:41 They're not elected.
26:43 Uh,
26:43 we're all worried about democracy nowadays,
26:45 you know,
26:45 and what's,
26:46 what's the,
26:47 what's,
26:48 I mean,
26:48 is that,
26:49 do you worry about that?
26:50 Yes,
26:51 definitely.
26:51 So
26:52 I spoke a little bit earlier about how
26:53 that's a potential disadvantage in undermining the government's
26:56 uh perceived legitimacy potentially over the longer run
26:59 and so in the context of this study,
27:01 we collect short run,
27:02 uh,
27:03 uh,
27:03 uh,
27:03 measures of views of the government at all levels as well as the views of the chief
27:07 and are in the process of collecting that,
27:09 um,
27:09 2 years,
27:10 3 years after the close of taxation to see
27:12 how those potentially evolve over the longer run.
27:15 And we see if anything areas of positive views of
27:17 the government increase in places where chiefs collected taxes,
27:20 and I think this may partly be a function of
27:22 who the chiefs are in the city.
27:23 These are not like
27:25 very powerful rural chiefs in more traditional settings that
27:27 have more say over the daily lives of citizens.
27:29 They're more like
27:30 quasi-traditional actors that operate in an urban setting.
27:33 And fill some gaps in governance outcomes.
27:35 So maybe they have,
27:36 they can bring some value to the government's
27:38 tasks,
27:39 but they have like less ability to punish,
27:41 um,
27:42 citizens which may speak to some of the external validity of this to
27:45 like whether it could extend to more rural settings or other areas where these
27:49 actors have more power,
27:50 but I think those are valid concerns.
27:54 Thanks,
27:54 Gabriel.
27:54 Anya.
27:58 Hi,
27:58 Gabriel,
27:59 um,
27:59 just a really quick,
28:00 um,
28:01 simple question,
28:01 uh,
28:02 and,
28:02 and do differ if you,
28:04 if that's,
28:05 if you
28:06 address this later,
28:06 but
28:07 can you say a little more about the logistics,
28:09 similarities or differences between the two?
28:11 So how many city chiefs are there per area?
28:14 How many agents were there?
28:16 Do they have the same number of houses to visit?
28:19 Who organized the agents and where they went.
28:23 Uh,
28:24 yes,
28:24 uh,
28:24 so one thing that's not in this list is some of the more details about how
28:28 they work within a neighborhood.
28:30 Both,
28:30 um,
28:31 in both treatment groups,
28:32 the agents work in teams,
28:33 so the chiefs work with the assistant that they work with in their normal duties.
28:36 In the central treatment areas,
28:38 it's two state agents that are paired
28:40 to whole constant the size of the teams,
28:42 and we show balance,
28:43 uh,
28:43 in the paper on these other things like the
28:45 size of the neighborhoods in which they work.
28:46 Work,
28:47 um,
28:47 characteristics of the neighborhoods,
28:49 etc.
28:50 on that more like balanced concern dimension,
28:52 um,
28:53 and in terms of the
28:54 overlap in chiefs and who's enlisted in tax collection,
28:57 there is some like overlap of jurisdictions that led us to
29:00 have to figure out with the government who
29:01 to select to collect in a particular neighborhood where
29:04 chief jurisdictions overlap,
29:05 um,
29:05 and those like conflicts and jurisdictions and overlap are balanced across the
29:10 Treatment and control groups again so that it appears that we chose like the best
29:13 chief or the counterfactual chief who didn't
29:15 end up collecting in the neighborhood to compare
29:17 at the neighborhood level.
29:18 Um,
29:18 does that answer your question?
29:21 Yeah,
29:21 thank you.
29:24 Perfect.
29:25 Uh,
29:25 I see no more hands,
29:26 Gabriel,
29:26 so you can continue.
29:28 OK,
29:28 thank you.
29:30 Um,
29:31 so this table summarizes the key actors involved in
29:33 the campaign and how their activities generate data.
29:36 Um,
29:36 so as I said,
29:37 the collectors are doing a property registration stage and then
29:40 tax solicitation that lasts about a month in each neighborhood.
29:43 In neighborhoods of which there are 356 were randomly assigned to
29:46 one of 7 months in the course of a seven-month campaign,
29:50 and these neighborhoods uh encompass approximately 45,000 properties.
29:53 So this will be the ultimate base of potential taxpayers
29:55 that I'll evaluate my compliance and revenue outcomes against,
29:58 and that comprises the government's administrative tax data.
30:02 The enumeration team undertook 3 separate surveys.
30:05 First,
30:05 a baseline survey about 10 months before the start of taxation.
30:09 Um,
30:09 in each neighborhood,
30:10 randomly selecting 10 households or 12 households per neighborhood.
30:14 Uh,
30:14 a midline monitoring survey happened 4 to 6 weeks
30:16 after the end of taxation in each neighborhood.
30:18 This is a very short survey concentrated on short-run outcomes
30:22 like,
30:22 uh,
30:22 immediate views of the government after the close of taxation,
30:25 reported bribes by by citizens,
30:27 reported visits,
30:28 um,
30:28 by the collectors,
30:29 um,
30:29 as reported by citizens,
30:30 and self-reported payments of the tax.
30:32 And the enumerators tried to reach as many households as possible,
30:35 um,
30:36 and we're able to reach about 36,000 of them.
30:38 The online survey um involves collecting
30:41 measures um of baseline measures like perceptions of the tax,
30:44 um,
30:45 views of government,
30:46 trust in government,
30:47 some of those more detailed measures and involve revisiting
30:50 the baseline sample with some attrition though
30:51 it's balanced across the treatment groups.
30:55 So now I'll turn to talking about uh results.
30:57 The estimation strategy is straightforward here.
30:59 We're regressing individual outcomes
31:01 for an individual in a neighborhood randomization stratum uh at a particular time,
31:06 uh,
31:07 on an indicator at the neighborhood level for that
31:09 neighborhood being assigned to the chief tax collection or local treatment arm.
31:12 In the excluded category in all regressions I'll show
31:14 you unless otherwise specified is the central treatment arm.
31:17 Minimization,
31:18 strain and fixed effects are a combination of two variables,
31:21 geographic divisions of the city and whether or not
31:23 a chief previously collected taxes in a neighborhood.
31:26 Um,
31:27 in some previous instance of taxation,
31:29 which is the case in about 15% of cases.
31:32 I also in my preferred specification include time fixed effects
31:35 to address a concern arising from a
31:36 secular decline in compliance that's common across
31:39 treatment arms over the course of the campaign in the run-up to a contentious
31:42 um presidential election in the DRC
31:45 and so to address this um time.
31:48 Uh,
31:48 this change in compliance over time,
31:50 even though it's common and secular across the treatment groups,
31:53 due to the staggering of how the treatments were rolled out neighborhood
31:56 by neighborhood because the government needed to train collector types and then
31:59 move on to the next collector type to train.
32:01 I include these fixed effects to restrict to the periods where
32:04 the treatments.
32:04 Most closely overlap um and then account for some time variation by splitting the,
32:09 the time up into waves,
32:10 two-month waves of of the campaign.
32:12 So I can talk a little bit more about other strategies
32:15 and robustness I showed to the strategy,
32:17 but I think that this provides the best estimate of
32:19 the treatment effects by accounting for this time variation.
32:22 Finally,
32:22 I cluster standard errors
32:24 at the neighborhood level,
32:25 um,
32:25 and fortunately randomization is balanced
32:28 across the treatments for demographic
32:29 and economic characteristics both of property
32:31 owners but also characteristics like the neighborhood of the neighborhood like
32:34 whether what's the average um level of tax compliance in the previous tax campaign.
32:40 OK,
32:40 here's the main results.
32:42 This table shows estimates from a regression,
32:44 regressing an individual level indicator for whether or not a citizen paid the tax,
32:49 um,
32:49 on the indicator at the neighborhood level for whether,
32:51 um,
32:52 a neighborhood was assigned to the local treatment arm.
32:54 And I see in column one,
32:55 not including house tight fixed effects,
32:58 uh,
32:58 that there's a 3.2% point increase in tax compliance relative to a central need
33:03 of 6.3%.
33:05 And so you may be struck by how low the overall levels of
33:08 compliance here which may speak to some of BJ's concerns about how extractive
33:12 this tax campaign overall is,
33:14 um,
33:14 but it's not that different from what we observe in other lower-income country
33:17 settings in the context of property taxation
33:19 where compliance ranges between 5 and 21%
33:22 in places like Liberia,
33:23 Rwanda,
33:24 and Ghana.
33:25 And to benchmark this magnitude a little bit more,
33:27 I compare it to a standard intervention in the public finance literature
33:30 which is embedding enforcement messages on tax fliers.
33:34 Um,
33:34 and so,
33:34 uh,
33:35 in cross-randomizing,
33:36 I,
33:36 I cross-randomize this as part of the,
33:38 this campaign embedding these messages,
33:40 reminders about penalties
33:42 on the tax letters distributed by
33:43 collectors during the property registration step
33:46 and compare those to a control message that just
33:48 reminds citizens about the importance of paying the tax.
33:50 Um,
33:51 and I see that these have a significant effect on compliance.
33:53 It doesn't differ across the treatment arms,
33:55 but the treatment effect of local is 5
33:57 times larger than the standard public finance intervention.
34:01 In column 2 to restrict variation to comparisons within property value bands
34:05 that include house type fixed effects
34:06 and the coefficient remains essentially unchanged.
34:09 In column 4,
34:10 I restrict uh the sample to those properties
34:12 that are deemed non-exempt by tax collectors this determination
34:16 is made during the property registration step and
34:18 so I'm conditioning on an endogenous outcome that I'll
34:21 consider as,
34:21 as such in a few slides,
34:23 but you can see that this reflects the check,
34:24 the fact that chiefs are more likely to exempt properties,
34:27 those that are elderly and disabled,
34:28 which are the official,
34:29 official,
34:30 um.
34:31 Designations uh for properties that can receive exemptions
34:34 and according to enumerator observations they're more
34:37 likely to do this process accurately.
34:40 In the paper,
34:41 I show extensive robustness to including baseline controls,
34:44 estimating these effects on different samples,
34:46 um,
34:47 and including running a fully saturated
34:49 regression including interactions and dummies for
34:51 all the cross-randomized experiments as well as dealing for the time variation,
34:55 uh,
34:55 concern I talked about earlier,
34:56 uh,
34:57 such as using matching strategies,
34:58 redefining the fixed effects using smaller time period fixed effects,
35:01 and the results
35:02 for the central and local comparison remain essentially unchanged to
35:06 these alternative robustness strategies.
35:11 When I look at revenues,
35:12 it's a similar story.
35:13 I see,
35:13 uh,
35:13 according to my preferred specification,
35:15 uh,
35:15 that the local treatment increases revenues by
35:18 43% relative to collection by state agents,
35:21 and this is similar to the increase in revenues
35:23 we see from high-powered incentives for property tax collectors
35:26 in what is probably the most recent,
35:28 um,
35:29 uh,
35:29 piece of evidence on manual tax collection and incentives,
35:32 uh,
35:33 for agents involved in that process
35:34 done by Adnan Khan,
35:35 um,
35:35 Asim Kwaja,
35:36 and Ben Olkin in Pakistan.
35:40 So now I'll turn to the downsides of cheap tax collection
35:43 and uh
35:44 these,
35:44 the,
35:45 the two potential downsides can be grouped under
35:47 the categories of cheats being harder to control
35:50 which could give rise to greater leakage and corruption
35:52 or these longer run concerns of undermining government legitimacy.
35:58 And in the context of the campaign,
35:59 I observed that bribes are higher when chiefs collect
36:01 taxes relative to state agents by about 1.8% points.
36:05 However,
36:05 this is on a very low base of 1.4%
36:08 of citizens reporting paying a bribe to the tax collector
36:11 in the context of the campaign in the central treatment arm.
36:14 So a positive and statistically significant increase though on a very low base,
36:18 and these bribes almost universally replace taxation.
36:21 So it's not that people are paying the tax and the bribe,
36:23 they're paying the bribe to avoid the tax.
36:25 And these effects on bribes are consistent
36:27 across collecting measures both at midline and endline
36:29 and also using a measure of bribes built from
36:32 comparing self-reports by citizens of paying the tax to
36:35 um administrative data on tax compliance at the individual level.
36:40 In the main other duty the chiefs undertake,
36:42 which is the administration of informal labor taxes,
36:44 I observed no differences in the extensive or intensive margin of
36:48 citizen participation in this informal tax system.
36:51 So citizens are not participate,
36:53 participating more in these public goods
36:55 projects organized and demanded by the chief
36:57 or contributing more labor conditional on participating.
37:01 So I think this is the strongest piece of evidence that these chiefs
37:03 and their main other duties aren't extracting more after engaging in taxation,
37:06 at least in the short run,
37:08 but these effects are the same at midnight and at 9.
37:12 On the assessment margin as I mentioned,
37:13 which the government may also care about in determining who should be
37:16 bearing the burden of the tax by ruling out or in,
37:19 um,
37:20 potential taxpayers,
37:21 chiefs are more accurate and they're granting more exemptions,
37:24 uh,
37:24 relative to state collectors.
37:26 On the government legitimacy side,
37:28 according to measures of views of the government
37:30 at all levels from the provincial governments up to
37:32 the national government as well as specific government
37:35 tax ministries and views of the chiefs themselves,
37:37 I don't observe any differences across the treatment arms in
37:41 these views.
37:42 Um,
37:42 indeed,
37:42 if anything,
37:43 trust in the government is higher in areas where chiefs
37:45 collected taxes relative to where state agents collected taxes,
37:48 even controlling for baseline levels.
37:52 And I see some evidence
37:53 engaging chiefs in tax collection builds some of the more
37:56 voluntary motivations for complying with taxes that
37:59 fall under the umbrella of tax morale.
38:01 I see a higher perceived compliance that others
38:03 in the neighborhood um complied with taxes,
38:05 so this may simply go hand in hand with
38:06 the true treatment effect um in local treatment areas.
38:09 Um,
38:10 uh,
38:10 but,
38:10 uh,
38:10 surprisingly I see some additional evidence of
38:13 fiscal externalities on other formal tax bases.
38:15 Citizens report paying more market vendor fees
38:18 and income taxes in areas where chiefs collected taxes relative to state agents,
38:22 though these should be taken with a grain of salt.
38:24 They're built,
38:24 they're only drawn from citizen self-reports about formal tax impacts.
38:31 OK,
38:31 so I think now I'll pause here on the results
38:33 before I turn to mechanisms if there are any questions on
38:36 these results.
38:39 Uh,
38:39 great,
38:39 thanks.
38:40 Uh,
38:40 Tristan,
38:40 are you ready for your question or you want to wait?
38:48 Tristan,
38:49 I don't
38:50 think we can hear you.
38:53 I'll just go to Chaviv before that,
38:56 and we can come back to Tristan.
38:59 Charlie,
38:59 go ahead.
39:00 Yeah,
39:00 hi,
39:01 thanks.
39:01 Um,
39:02 so,
39:02 uh,
39:03 sorry,
39:03 uh,
39:03 you know,
39:04 I,
39:04 I,
39:04 I missed Vis and,
39:06 uh,
39:06 so,
39:06 so the previous question,
39:07 so I don't know whether this was asked,
39:08 but,
39:09 but basically I,
39:10 I wonder whether,
39:11 you know,
39:12 uh,
39:12 the,
39:12 the chiefs,
39:13 um,
39:14 are basically able to provide better services,
39:17 uh,
39:17 to,
39:18 to these,
39:18 um.
39:19 You know,
39:20 to the,
39:20 to the people that they,
39:21 they collect taxes from.
39:22 And so,
39:23 you know,
39:24 whether,
39:25 whether the provision of these services,
39:27 then,
39:27 you know,
39:28 changes their,
39:28 their perceptions about,
39:30 you know,
39:30 the,
39:30 the,
39:30 the benefits of paying taxes and whatnot.
39:34 Yeah,
39:34 that's a great question.
39:35 So I'll get to this when I speak about mechanisms that even conditional on visiting
39:39 a household,
39:40 like setting aside how,
39:41 who they approach for taxation across collector types,
39:43 whether that differs,
39:44 Chiefs may be more affected by virtue of what they do for citizens in the community
39:48 because remember this is the DRC where
39:49 the both the national and provincial government provide
39:52 no like funding for public.
39:53 Like education or health services,
39:55 most of that is filled by private,
39:56 fill that space is filled by private NGOs or other actors.
40:00 So when thinking from the citizen side,
40:01 the chief is administering local public goods
40:03 through this informal labor tax institution.
40:05 The provincial government doesn't provide much for citizens,
40:08 um,
40:09 and within a particular like randomly selected neighborhood
40:13 compared to the chief,
40:13 so that may impact the compliance decision,
40:15 and that's something I'll,
40:16 um,
40:17 examine explicitly when we get to mechanisms.
40:20 Right,
40:20 so,
40:20 so just,
40:21 sorry,
40:21 just a,
40:21 a quick follow up.
40:22 So,
40:23 so,
40:23 you know,
40:24 well,
40:24 the case of the state,
40:25 so that these guys are only collecting taxes,
40:27 they're not providing services,
40:28 whereas the chief is doing both collecting and providing services.
40:31 So,
40:33 OK,
40:33 so yeah,
40:33 so that's,
40:34 that's a bit of an important difference.
40:35 OK,
40:36 thanks.
40:38 Yeah,
40:38 uh,
40:39 this is Tristan.
40:40 Can you hear me?
40:40 Yes,
40:41 yeah,
40:41 um,
40:42 uh,
40:42 so I guess my question was very similar,
40:44 and I just wanted to
40:45 sort of note that there's a,
40:46 there's a job market candidate at Berkeley,
40:49 Benjamin Kraus,
40:50 who does
40:51 also tax compliance experiments,
40:52 but then he has,
40:54 you know,
40:54 one variation of actually
40:55 Um,
40:56 the amount of services which are given,
40:58 and he finds that
40:59 if you don't,
41:00 uh,
41:00 deliver services,
41:01 and you enforce tax compliance,
41:03 tax revenue can actually go down,
41:05 whereas people actually give money only when they're
41:07 receiving services.
41:08 So I think,
41:09 you know,
41:09 to these points and also Javis that it would be good to see some,
41:12 I guess variation across individuals in terms of
41:15 the quality of of services they're receiving and,
41:18 and,
41:19 you know,
41:19 whether they get those from the state or the gov the chief.
41:22 Overall,
41:23 sure,
41:23 um,
41:24 yeah,
41:24 I'll just briefly preview the like set of evidence on that margin so I can look,
41:28 I look at heterogeneity and chiefs like
41:30 how effective they are according to citizens,
41:32 um,
41:32 at baseline and just look at heterogeneity and their performance by that,
41:36 but also I have some,
41:37 uh,
41:38 some of that cross-randomized variation in the tax prime experiments are,
41:41 I also include.
41:42 primes about
41:43 um motivations for paying the tax that included trust in the government or
41:47 expecting public goods in return and I can look at
41:49 like differences in the effect of those messages across the treatment arms
41:52 that would at least get about like is there
41:54 an effect of priming citizens on these margins differentially across
41:57 the collector type um and we can look at that evidence at,
42:00 at the end of the mechanism section.
42:05 You too.
42:08 Yeah,
42:09 thanks.
42:10 The,
42:10 the,
42:11 these responses on bribes and trust in government and so on,
42:14 they're from the household survey,
42:15 right?
42:17 Uh,
42:17 yes,
42:18 the part of the bribes measure is built from comparison,
42:21 comparing citizen self-reports of payments to the administrative data,
42:24 but everything else is built from,
42:26 uh,
42:26 citizen
42:27 surveys.
42:27 OK,
42:27 so do you,
42:27 do you worry about,
42:29 uh,
42:29 biased responses and particularly bias with
42:33 I mean,
42:34 particular attributes of the chiefs,
42:36 I suppose you could test for that,
42:37 but,
42:37 you know,
42:38 if I was scared of a particular chief,
42:39 I'm not going to tell you whether I'm bribing him or not or her.
42:44 Um,
42:45 so you're worried that because in areas where chiefs collected taxes,
42:48 citizens are less likely to say that they
42:50 need a chief.
42:51 I,
42:51 I'm,
42:51 I'm not sure which direction it's going to go.
42:53 I'm just worried that,
42:54 that,
42:54 you know,
42:54 if,
42:55 if I'm scared of retribution from a chief because they learned that I'm sort of
42:59 telling the interviewer bad things about them,
43:02 they're going to not be nice to me.
43:04 Yeah,
43:04 that's certainly a possibility.
43:05 I think the best response I can give here
43:07 is we look at and control for baseline measures
43:10 of these views before the chief engaged in taxation and we don't see like movement
43:14 in either treatment group controlling for baseline measures and
43:17 changes over time relative to these baseline measures.
43:19 I think
43:20 that's my best sense of before taxation took place and we
43:22 raised up the chief and the ability to collect taxes,
43:25 it doesn't look like views of the government
43:26 have changed in the treatment of control areas
43:29 over time.
43:32 Thanks,
43:32 Gabriel,
43:33 um,
43:33 Chuan,
43:34 and then we'll continue.
43:38 Yeah,
43:39 so I,
43:39 I think the results are quite
43:41 surprising in that
43:43 it's not clear why.
43:46 Which is we want to,
43:47 to,
43:47 to pay the tax to the tax authority itself.
43:50 My,
43:50 my question is,
43:51 is,
43:52 so you,
43:52 you create some tests to try to,
43:54 to kind of dismiss other stories of
43:57 long-term bribing,
43:58 displacement of tax to other,
44:00 other things,
44:00 but
44:01 are you,
44:01 are you
44:02 powered enough to be able to rule those out convincingly?
44:07 Or have you in designing that,
44:09 that,
44:09 that experiment,
44:10 did,
44:10 did you think about
44:12 how you would go and test
44:14 the fact that they could enter long-term contracts and,
44:16 and,
44:17 and the number of brides would increase or,
44:20 or the payment to on taxes that would go back
44:23 to the community more directly would increase as opposed to
44:26 tax to the provincial government.
44:27 So a lot of things can happen,
44:29 uh,
44:29 uh,
44:29 you know,
44:30 since the,
44:30 the chief is,
44:31 is present in very,
44:33 in many aspects of,
44:34 of citizen's life.
44:35 So,
44:36 have you thought about trying to,
44:37 to detect those or is that some kind of expo thing that you've done?
44:41 And,
44:42 and might not have the power to,
44:43 to actually test.
44:45 Um,
44:46 yeah,
44:46 that's a great question.
44:47 So in the context of this campaign,
44:48 we look at the short-run measures,
44:50 um,
44:50 like 4 to 6 weeks after taxation finished in the neighborhood,
44:53 what are the differences in these more corruption like
44:57 pure bribe taking but also extraction on these,
44:59 in these other chief duties,
45:00 and then the online data is collected between 10 and 12 months afterwards,
45:04 so that's the longest
45:05 period out in this study that we can see.
45:07 Um,
45:08 and then in the follow-up study I mentioned earlier looking at
45:10 Chief's performance in
45:11 the targeting of anti-poverty benefits,
45:13 um,
45:14 leveraging variation in whether those chiefs collected taxes or not,
45:17 in which all chiefs were asked to target these benefits.
45:19 We can again collect measures on differences in
45:22 not bribe taking during the tax campaign,
45:25 but these other margins like administration of the informal labor tax and
45:28 then additionally study how these chiefs
45:29 perform based on whether they collected taxes
45:31 in the targeting of these benefits
45:33 and we observed that the chiefs.
45:35 tax,
45:35 um,
45:36 or if anything better at targeting needier citizens
45:38 and are less nepotistic in doing so.
45:41 Um,
45:41 so that's the furthest out I can go.
45:42 The government wants to,
45:43 um,
45:44 conduct a new tax campaign
45:46 after COVID,
45:46 and we expect that they'll want to engage chiefs because they are more effective
45:50 and,
45:50 um,
45:50 we're currently thinking about ways that we
45:52 can get out these longer run considerations about
45:54 empowering the chief in this way.
45:55 Does it have
45:56 potential costs or even potential benefits as you alluded to
45:59 over the long run.
46:05 OK,
46:06 you can continue,
46:06 Gabriel.
46:07 OK,
46:08 thank you.
46:10 Um,
46:10 so now I'll talk about mechanisms and
46:11 the explanations for the chief's greater effectiveness,
46:15 like,
46:15 uh,
46:15 attempt to group into three main families of explanations.
46:18 And so the first is that chiefs are simply making more tax
46:21 visits because they live in the neighborhoods in which they collect.
46:25 The second is that
46:27 chiefs are better able to target tax solicitation
46:29 because they possess local information about citizens,
46:31 about their propensity to pay.
46:33 And the third,
46:34 um,
46:35 is that chiefs conditional on visiting a household are more persuasive,
46:39 uh,
46:39 and convincing citizens to pay in the tax bargaining process,
46:42 and this can be grouped into some of the tax morale,
46:44 uh,
46:45 explanations around the voluntary decision to comply,
46:47 and these can be linked to chiefs greater trust with
46:49 citizens or some of the explanations that were brought up,
46:52 uh,
46:52 in the questions about chiefs,
46:54 um,
46:55 having a,
46:55 a stronger claim on the reciprocity.
46:57 Of taxation and public goods spending that because
47:00 they're they actually provide goods in the neighborhood
47:02 they may be better able to activate
47:04 um or signal
47:05 the importance of taxes um in
47:08 um
47:09 to pay
47:10 and then simply cheese could be more effective because they have
47:12 other tools at their disposal for punishing non-compliers relative to.
47:18 So the intuition behind this first channel is that um it's less
47:21 costly for chiefs to make visits because they live where they work.
47:23 Estate agents are traveling from their home,
47:25 typically in a different part of their city
47:27 to collect in,
47:27 in the city,
47:28 to collect in the neighborhood on a daily basis
47:31 during collection,
47:32 um,
47:32 so chiefs can spend more time in the neighborhood,
47:34 potentially make more visits or repeat visits.
47:36 This could increase compliance if,
47:37 for example,
47:38 time-bearing liquidity constraints bind.
47:40 Chiefs can catch people on days when they're more likely to
47:42 have cash on hand if they're visiting them more often and if
47:45 liquidity is an important input to compliance,
47:47 this may matter.
47:48 It could also simply increase the perception that
47:51 necessary collector is not going away and
47:53 you comply with the tax.
47:56 Raise compliance.
47:57 So the first test I apply here is to look
47:58 at differences reported visits by citizens in the midline monitoring survey
48:03 on the extensive,
48:04 uh,
48:04 margin of receiving a visit at all,
48:06 and here I'm talking about visits after the registration step in which
48:09 all households are visited that are made at the collection team's discretion,
48:13 as well as the intensive margin of number of visits
48:15 received by a specific household.
48:17 And here I see no significant differences along either margin.
48:21 Uh,
48:21 and I'm relatively well-powered here because this draws on the midline
48:25 monitoring survey where there's about 19,500
48:27 observations between these two treatments.
48:29 So the null effects have a relatively high degree of precision.
48:32 But an additional piece of evidence is to look at interactions that
48:35 take place outside of the official tax visits that collectors conduct,
48:39 and I asked citizens about interactions that happen
48:42 outside of those visits.
48:43 So you can imagine that the chief in doing his or her normal duties is
48:46 reminding citizens that I'm gonna visit you tonight and ask that you pay the tax,
48:49 so you better.
48:49 Have your money together
48:51 and to see whether that would be an additional
48:52 way that cheese can put pressure on citizens,
48:54 but I don't observe any differences in these
48:56 unofficial discussions across the treatment areas with collectors.
48:59 So it appears that chiefs are making more effort outside of
49:02 the explicit tax visits,
49:04 uh,
49:04 to ask citizens to pay the tax.
49:09 The intuition behind the second mechanism is that chiefs know citizens well,
49:12 they interact with them on a daily basis
49:14 in the administration of their normal duties,
49:16 and so they have a potentially a better
49:18 signal about households payment propensity for the tax,
49:21 and this could make them more efficient collectors if they
49:23 can shift the composition of households towards higher propensity payers.
49:29 And the first test I'll apply here is to
49:31 see what happens when you share chief's information with a team of state collectors
49:34 and test whether that can increase compliance.
49:37 And I do this by evaluating the hybrid treatment arm,
49:39 the Central plus Local Information or CLI arm,
49:42 and the way this worked is that,
49:44 um,
49:44 the state collection team first
49:46 did the property registration stuff on their own,
49:48 and then
49:49 having constructed the property register for a neighborhood,
49:51 they met with
49:53 Of that neighborhood and discuss the ability and
49:55 willingness of each property owner to pay.
49:57 Then armed with this information,
49:58 they go to collect
49:59 in the neighborhood
50:00 and we check after the fact to make sure that
50:03 it doesn't appear that now that they've met one another,
50:05 they collaborate in taxation
50:06 um through both collector surveys and surveys of citizens.
50:11 So what this looks like in practice is there's
50:12 a team of state collectors at top and bottom
50:14 and they have the property register in front of
50:16 them with the name of the property owner,
50:17 the tax ID and tax amount.
50:19 The chief is that right and he's for who's in charge of that neighborhood,
50:22 um,
50:22 and he's looking at a picture of the property
50:24 on the tablet which was taken by the enumerator
50:27 during the property registration step and the enumerator is prompting the chief
50:30 to separately rate the ability and willingness to pay of each property owner
50:34 on the tax roll and they go line by line through it.
50:36 It takes approximately 4 hours to do one of these consultations,
50:39 and the state agents are recording the information on the register
50:42 and they'll take this with them when they go to collect.
50:45 So first I check what happens to compliance and revenues.
50:48 The CLI treatment arm as you see in column one,
50:50 increases compliance by 2.4% points relative to the
50:53 state agent arm where no consultation takes place
50:56 and increases revenuees significantly.
50:58 Incomtu.
51:00 This is also the case conditional on visits.
51:02 So columns 3 and 4 show that there's no differences
51:04 in the extensive or intensive margin of visits respectively.
51:07 In columns 5,
51:08 I restrict the sample to those that report receiving a visit in the midline survey
51:13 after the registration step and see that compliance is higher
51:16 in the CLI areas,
51:17 um,
51:18 among the set of visitors.
51:19 So this is,
51:20 this is a marginally significant difference is
51:21 consistent with the compositional shift in taxpayers.
51:25 Um,
51:25 when state agents,
51:26 uh,
51:27 engage in consultation with the chief.
51:30 However,
51:31 these,
51:31 um,
51:32 collectors don't fully recover the gap with chiefs.
51:34 Um,
51:34 when I compare
51:35 local to the CLI treatment arm,
51:37 I see still a negative
51:38 effect of 2.6% points,
51:41 which is consistent with,
51:42 um,
51:42 uh,
51:43 local information being
51:44 somewhat to some extent non-codifiable or
51:46 not fully transferable to the state agents
51:48 in the context of a one-off consultation
51:50 that the chief can't fully communicate their
51:52 full set of knowledge about citizens.
51:55 However,
51:55 it's also consistent with some of the
51:57 potential mechanisms on the persuasive side that even
51:59 conditional on reaching a high payment propensity household,
52:02 chiefs are more effective at convincing that property owner to pay.
52:05 And I'll return to those explanations in a few slides.
52:11 But first I want to build the evidence that the CLI collectors are
52:14 more effective because the chiefs possess
52:17 information about citizens that's effective and informing
52:19 uh
52:20 whether they're likely to pay the tax or not.
52:22 So first I want to check that the
52:23 CLI collectors actually followed the chief's recommendations,
52:25 and I do this by correlating the measures.
52:28 Collected during the consultation,
52:29 which are 1 to 3 scales of
52:31 ease of payment and willingness to pay that the chief
52:34 provided their subjective evaluations of for each household and that was recorded
52:38 by the enumerator that I can then match to the records of
52:41 citizen reported visits and administrative tax compliance.
52:45 And I see that I see positive and significant correlations for both these measures
52:49 on the visits and compliance margin that
52:51 area that properties that are rated as having a higher
52:53 ease of payment receive more visits by the CLI collectors
52:56 and they're more likely to comply with the tax.
52:58 Of course,
52:58 there's concerns about the endogeneity of compliance here to
53:01 shifts in the visit strategy and I'll return to that
53:03 um
53:04 in the next table.
53:07 So then I want,
53:08 I ask how informative our chief's recommendations beyond what
53:11 the state agents would observe on their own.
53:13 So the property,
53:13 looking at the property,
53:14 you can gain information based on observing how
53:17 high quality the materials of the property are
53:19 even from talking to the property owner
53:21 potentially and what observable characteristics there may be
53:23 about
53:24 that person.
53:25 Um,
53:25 but the main factors here in the decision of state
53:28 collectors to target households appears to be observable house characteristics.
53:31 So I repeat these correlations controlling for
53:34 enumerator observations of the wall and roof quality of structures on the property
53:38 and whether the property is threatened by erosion,
53:39 and the correlations are somewhat weaker,
53:41 but they're still positive and significant
53:43 for both measures in columns 124,
53:45 and 5.
53:46 In columns 3 and 6,
53:48 I look at the,
53:48 again the outcome of administrative
53:51 record,
53:51 the administrative record of tax compliance.
53:53 And restrict the sample to properties that said they received a visit,
53:56 and I see that conditional on visits
53:58 compliance is higher in areas where chief said these properties have a
54:02 property owners have a higher ease or higher willingness of payment.
54:08 Then there's the question of whether chiefs were actually providing
54:11 the information that was most strongly correlated with
54:14 uh potential tax compliance and weren't,
54:16 weren't playing some other dynamic game of or short run game of
54:20 trying to punish
54:21 their enemies in the neighborhood by directing collectors there,
54:23 reward their friends by
54:25 moving collectors away from them.
54:26 And so I,
54:27 I try to study how chiefs themselves would target and whether they follow
54:31 similar recommendation formulas as the chiefs that serve as consultants.
54:34 So obviously this counterfactual doesn't exist in
54:37 areas where chiefs served as consultants,
54:39 state agents in the CLI aren't collected taxes,
54:41 but I use the microdata built from these chief consultations.
54:45 To create a prediction formula for both ease and willingness to pay
54:48 using variables from surveys,
54:50 so I create this,
54:51 use those observations
54:53 from surveys to construct this predictive measure
54:55 and then apply,
54:56 apply that prediction formula in the local treatment areas
54:58 and evaluate how chiefs perform when they collect in those
55:01 areas relative to this predicted measure of ease of payment
55:04 and willingness to pay.
55:06 And I see similar correlations in local
55:08 treatment areas even conditional on observable characteristics of
55:10 the household that these predictive measures of
55:12 ease and willingness to pay predict both visits
55:15 and compliance positively.
55:17 And so then an obvious question is what does
55:19 this look like in the central treatment arms,
55:21 and we do see in columns 2 and 4 applying the same prediction formula
55:25 that the predicted ease and willingness to
55:27 pay are positively correlated with compliance,
55:30 but in columns 1 and 3,
55:31 I see that they're not predictive,
55:32 conditional and observable characteristics of the
55:34 property with receiving a visit.
55:36 So it really doesn't appear that the state agents have access
55:38 to the content of information that the chief has conditional on
55:42 the household's observable qualities.
55:46 So then I finally ask if this is the case that chiefs information is important,
55:49 we should observe that more knowledgeable chiefs
55:51 are more effective when serving as consultants.
55:54 So I built a
55:55 measure of chief knowledge built from quizzes um given to chiefs about the name,
55:59 job,
55:59 and education level of citizens in their neighborhood to build
56:01 an index of chief information and see how they perform,
56:05 um,
56:05 as consultants in terms of whether they raise the
56:08 performance of the CLI agents more
56:10 post-consultation relative to less knowledgeable chiefs.
56:13 And I see that this is indeed the case in column 4
56:15 that more knowledgeable chiefs raise compliance
56:17 more when they serve as consultants,
56:19 um,
56:19 so the indicator at left in this table is just
56:21 an indicator for the chief knowledge being above median,
56:24 uh,
56:24 for all chiefs in the city.
56:26 As a placebo test,
56:27 I can look in the central treatment areas
56:29 and observe no correlation between chief knowledge and
56:32 central collector performance when no consultation took place
56:35 and in column 6 I see also that chiefs that are more
56:37 knowledgeable appear to be more effective collectors when they themselves lied.
56:43 Um,
56:43 so that summarizes the,
56:44 um,
56:45 the evidence on the information side,
56:47 and so if you recall there's still a gap between the CLI and local treatment arms,
56:50 and this could either be consistent with a
56:53 still like some form of information gap
56:55 that the consultation doesn't fully fill that gap
56:57 or consistent with these other
56:58 persuasion mechanisms which I'll turn to now.
57:02 The idea here is that Gabriel,
57:04 can I ask a question.
57:07 Um,
57:07 so what is the,
57:09 what is the main tool in the hands of
57:12 the central tax collector who consulted with the chief
57:16 in terms of getting
57:18 Higher,
57:19 you know,
57:19 higher revenues.
57:21 I don't want to say higher compliance because it's conditional on the visit.
57:24 So is it just the
57:25 increasing the number of visits and nagging?
57:28 Or is there actually
57:30 something else where they just kind of threatening more with,
57:34 you know,
57:34 penalties or priming somehow,
57:36 whatever,
57:37 and
57:38 related to that,
57:39 I guess,
57:39 do the citizens know that these people consulted with the chief?
57:44 Um,
57:44 yeah,
57:44 so in answer to your last question,
57:46 they don't know
57:47 that the consultation took place.
57:49 Um,
57:49 we asked the chiefs and the estate agents not to mention it.
57:51 It's still possible that they
57:53 knew the consultation.
57:55 But in the survey we don't see evidence of that.
57:57 Um,
57:58 I think the,
57:58 the evidence that we have,
58:00 uh,
58:00 that suggests how these,
58:01 uh,
58:02 CLI agents raise their effectiveness relative to when they collect
58:05 without a consultation is not that they're doing more visits.
58:07 The visit rates are not different.
58:09 If anything,
58:09 they're lower in the CLI treatment arm relative to the state agent central arm.
58:14 It rather appears that they're shifting the composition of taxpayers,
58:16 so this is something I'll return to
58:18 in the distribution implication section,
58:20 but they come to resemble the types
58:21 of households that chiefs target when they collect
58:24 rather than state agent collectors.
58:25 Um,
58:26 does that fully answer your question?
58:28 Yeah,
58:29 it's,
58:29 I mean,
58:29 I had missed because I didn't read the paper.
58:32 I missed that they don't get to
58:35 visit everyone or even maybe a potentially a substantial portion,
58:39 and they have to kind of strategize on where to go.
58:42 Yes,
58:43 that's right.
58:43 So in the second stage,
58:44 only 40% of households receive a visit on average,
58:47 um,
58:48 solicitation,
58:48 right.
58:49 Thanks.
58:50 Um,
58:50 I'm seeing two more hands.
58:52 If you don't mind,
58:52 let's take those quickly and then we'll let you
58:54 run through your rest of yours until 2:15 uninterrupted.
58:57 Is that OK?
58:58 Sure.
59:00 Chubby,
59:00 you were first.
59:02 Yeah,
59:02 so what's the ownership rate here?
59:05 So
59:05 you know if there's a lot of renters,
59:07 what happens to a visit?
59:08 I mean,
59:08 the guy should not pay that.
59:09 So,
59:10 you know,
59:10 is it part of the better information that these chiefs know who
59:14 actually that they're,
59:15 you know,
59:15 this is a house with the owner there versus a house that is rented?
59:19 Um,
59:20 and then,
59:21 you know,
59:21 uh so were these,
59:22 sorry,
59:23 it's a separate question,
59:23 but were these chiefs
59:25 rewarded somehow by,
59:26 by,
59:27 you know,
59:27 in the elicitation only,
59:29 uh,
59:29 arm where they,
59:30 you know,
59:31 they had to go,
59:32 you know,
59:32 picture by picture saying,
59:33 you know,
59:34 whether or not
59:35 These guys would be willing to,
59:37 uh,
59:37 would be willing to pay the tax and,
59:39 and could uh,
59:39 you know,
59:40 could a
59:40 sort of an incentive where you kinda,
59:42 you know,
59:42 you make it,
59:43 you know,
59:43 you make a payment conditional on actually eliciting a tax
59:47 payment,
59:48 uh,
59:48 would,
59:48 would that improve,
59:49 you know,
59:50 their,
59:50 their,
59:50 at least their effort and,
59:51 and,
59:52 you know,
59:52 in,
59:52 in providing,
59:53 you know,
59:54 good information and that,
59:55 that could also help you,
59:56 um,
59:57 you know,
59:57 I guess in the,
59:58 in the pursuit,
59:58 you know,
59:59 whether that's
1:00:00 You know,
1:00:00 it's part of the,
1:00:01 part of the impact of the,
1:00:02 of the chief is,
1:00:03 is persuasion or is actually just better information.
1:00:07 Um,
1:00:07 yeah,
1:00:08 that's a great question.
1:00:08 So to your first question about um
1:00:10 whether they have content about who should be bearing the burden
1:00:13 of the tax and thinking about like renters versus property owners,
1:00:16 in the initial registration step,
1:00:17 uh,
1:00:18 collectors were instructed to like rule out
1:00:20 the households that were,
1:00:21 had renters residing there and to identify the
1:00:24 property owner if they were within the city,
1:00:25 then
1:00:26 they could list that property owner's name and the renter
1:00:28 would be in charge of hassling the property owner.
1:00:30 Um,
1:00:30 but it doesn't appear like the rental rates are not very high in this context.
1:00:34 They're about,
1:00:35 um,
1:00:35 like official rental rates
1:00:37 and how the government construes it.
1:00:39 Um,
1:00:39 I think less than 10%.
1:00:40 Most people are residing within the property,
1:00:42 but formal property ownership is like around 10% as well.
1:00:45 So people have
1:00:46 like a sale deed for their land,
1:00:47 but
1:00:48 in from the government's perspective,
1:00:49 this doesn't officially recognize
1:00:51 that they own the property though according to the tax they're
1:00:54 still liable to pay the tax because they're residing there.
1:00:57 Um,
1:00:57 so we can look at that uh we look
1:00:58 at that on some of the exemption and classification margins
1:01:02 at the beginning and the only differences we see across the treatment
1:01:04 and control groups are that chiefs are more likely to accurately exempt.
1:01:07 Households,
1:01:08 uh,
1:01:08 to your,
1:01:08 uh,
1:01:09 second question about the incentives for the chiefs during the consultation,
1:01:13 um,
1:01:13 they were just given a lump sum payment at the,
1:01:15 at the end of the consultations.
1:01:17 They're giving something equivalent to,
1:01:19 I think about $5 US dollars,
1:01:20 um,
1:01:20 to compensate them for traveling because they travel
1:01:22 to an office to do the consultation,
1:01:25 um.
1:01:26 Uh,
1:01:26 I think you were suggesting potentially that
1:01:28 you could get more effective information if
1:01:30 you like align the incentives for for
1:01:31 for providing that information during the consultation
1:01:34 more um and we didn't do that since it was just a lump sum payment.
1:01:38 Um,
1:01:38 and something that I haven't looked at
1:01:39 explicitly is how like baseline chief characteristics of
1:01:43 the,
1:01:43 that service as consultants other than their knowledge,
1:01:45 um,
1:01:46 would influence outcomes during this process,
1:01:47 but I think that's something that could get us
1:01:49 some of the points that you raised,
1:01:50 um,
1:01:51 to look at like how powerful the chief is,
1:01:52 how like extractive they are in their other duties,
1:01:54 and
1:01:55 look at those that heterogeneity within
1:01:57 the consultation data.
1:02:00 He's shiny.
1:02:05 Thanks.
1:02:05 Um,
1:02:06 so,
1:02:07 um,
1:02:08 I'm just wondering if tribal or ethnic,
1:02:11 uh,
1:02:11 ethnic,
1:02:11 um,
1:02:12 identities are
1:02:13 a salient dimension here,
1:02:15 um,
1:02:16 and if you have any variation,
1:02:17 for example,
1:02:18 um,
1:02:19 In the ethnicity or or uh tribe of the collectors versus the chiefs,
1:02:25 and so is it,
1:02:25 you know,
1:02:25 for example,
1:02:26 that
1:02:26 the chiefs just tend to be of the same ethnicity or tribe as
1:02:31 the people they're collecting taxes from
1:02:33 and so that would sort of suggest
1:02:35 an alternative
1:02:37 mechanism and an alternative
1:02:39 way,
1:02:39 um,
1:02:41 potential alternative ways of collecting taxes
1:02:43 that would get around some of the concerns
1:02:45 that for example Bijo raised about legitimacy.
1:02:48 Yeah,
1:02:48 that's a great question.
1:02:49 So in within this,
1:02:50 uh,
1:02:51 province and within Kananga,
1:02:52 there's a dominant ethnicity,
1:02:54 uh,
1:02:55 that,
1:02:55 uh,
1:02:56 accounts for about like 68% of citizens within the city,
1:02:59 and the rates of like membership of this identity which is Lulua,
1:03:02 um,
1:03:03 among chiefs and state collectors is the same at baseline.
1:03:05 And so the match between the ethnicity of the property
1:03:08 owner and the ethnicity ethnicity of the tax collector,
1:03:10 um,
1:03:10 across the treatment groups is the same.
1:03:12 Um,
1:03:12 there is a question then of how like they go
1:03:14 about soliciting taxes by the ethnicity of the property owner.
1:03:18 And we see um no differences in the targeting by like whether you're,
1:03:21 you have a co-ethnicity match with the property owner and the tax collector
1:03:25 on the collector side.
1:03:26 There is some suggestive evidence that state agent collectors are more
1:03:29 likely to try and pursue co-ethnics more for tax collection,
1:03:32 so we see some positive increase in visits
1:03:34 for co-ethnics,
1:03:35 um,
1:03:36 relative to non-co-ethnics in the state collector side,
1:03:38 though.
1:03:38 No differences in the ultimate set of tax compliers
1:03:41 and this difference is only marginally significant and a little bit noisy,
1:03:44 um,
1:03:45 so
1:03:45 I think it surprised us a little bit given there's a substantial set of like
1:03:48 minority tribes within the city that at
1:03:50 least in tax collection it doesn't appear that
1:03:53 the agents are targeting
1:03:54 co-ethnics differentially or
1:03:56 appear to be targeting co-ethnics,
1:03:57 uh,
1:03:58 more strategically in the first place.
1:04:02 Uh,
1:04:02 thanks,
1:04:02 Gabriel.
1:04:03 So for time management,
1:04:04 you have a little less than 10 minutes and we'll try
1:04:07 not to interrupt you unless there's something urgent comes up.
1:04:10 OK,
1:04:10 great.
1:04:11 Um,
1:04:12 so I'll talk less about these persuasion
1:04:13 mechanisms before getting to the distribution implications.
1:04:16 Um,
1:04:16 and so the idea here is that the chief might have two
1:04:19 tools at their disposal.
1:04:20 The first is,
1:04:21 um,
1:04:22 uh,
1:04:23 the ability to activate tax morale because they're more trusted by
1:04:25 citizens or have a closer link to services as we discussed
1:04:29 in more detail earlier,
1:04:30 um,
1:04:30 given the government doesn't really provide much in,
1:04:32 in terms of public goods in this setting,
1:04:34 and this could engender greater voluntary motivation,
1:04:37 setting aside concerns about enforcement power to comply with the tax.
1:04:40 Then chiefs may also simply have more sanctioning tools at their disposal.
1:04:44 Well,
1:04:44 we might think that state collectors have a more credible
1:04:46 claim on enforcing the official legal fines for non-tax for non-compliance
1:04:51 because they're more closely linked with the tax ministry,
1:04:52 Chiefs may be able to,
1:04:54 uh,
1:04:54 threaten to withhold services,
1:04:56 to enforce more informal labor tax
1:04:58 or ostracize people in the community who don't pay,
1:05:01 and that can make them more effective.
1:05:03 Um,
1:05:03 and so I'll skip to the summary of evidence on this,
1:05:06 um,
1:05:07 mechanism,
1:05:08 and I think what,
1:05:09 um,
1:05:10 uh,
1:05:10 a piece of evidence that I find among the most convincing on this side is to look at
1:05:14 collection during property registration.
1:05:16 So this is the stage of the campaign in which every
1:05:18 household receives a visit in a linear property by property fashion
1:05:21 where they move from east to west across the neighborhood.
1:05:24 Our enumerator is present and is recording information,
1:05:27 uh,
1:05:27 that captures GPS GPS points so we can
1:05:30 validate the path of collectors to the neighborhood.
1:05:32 And this should really neutralize the chief's ability to target households by
1:05:36 payment propensity because every household is visited in a specific pattern.
1:05:39 And here I don't observe any differences across the
1:05:41 central and local treatment arms in tax compliance or revenues
1:05:45 when this targeting ability of the chief is neutralized.
1:05:47 So this seems to suggest that,
1:05:49 uh,
1:05:49 granted it's in the presence of enumerators
1:05:50 but chiefs are not more effective at convincing
1:05:53 property owners to pay the tax.
1:05:55 A second strategy is to look at cross-randomized
1:05:58 messages on tax letter treatments which included local public goods messages,
1:06:01 information about the government and why
1:06:04 you should trust the government and the provision of local
1:06:06 public goods to activate these more tax morale-like mechanisms.
1:06:09 And we don't see any significant differences.
1:06:11 I also in response to the enforcement messages
1:06:13 that are different across the treatment groups.
1:06:15 So you might think if these were more effective at priming citizens in a particular
1:06:19 reason to comply with the tax,
1:06:20 it might be more effective in local areas relative to central.
1:06:23 That doesn't appear to be the case.
1:06:24 I also don't see heterogeneity by baseline
1:06:26 chief characteristics that we think might be.
1:06:28 Correlated with persuasive ability,
1:06:29 chiefs that are more trusted or more powerful
1:06:31 or operate in more remote areas of the city
1:06:34 at baseline aren't more effective at collecting taxes.
1:06:36 It is the case that more active chiefs are more effective,
1:06:40 active as measured at baseline in terms
1:06:41 of the number of duties they're undertaking,
1:06:43 how often and regular they,
1:06:44 they administer the informal tax,
1:06:46 for example.
1:06:47 These more active chiefs at baseline are more effective at raising compliance,
1:06:51 um,
1:06:51 though this is also consistent with an information story,
1:06:53 more active chiefs will be better informed and knowledgeable
1:06:56 about citizens because they're interacting with them more frequently.
1:07:00 And finally I look at inline measures that are more direct measures of these um
1:07:05 uh potential sources of inputs to the um persuasion side of compliance
1:07:09 like
1:07:10 citizens perceived probability of punishment for non-compliance atline,
1:07:13 the,
1:07:14 their beliefs about the duty to pay the tax and how much of that
1:07:16 money they think will be spent on public goods,
1:07:18 and I don't see any significant differences
1:07:20 across the treatment and control room.
1:07:21 So all told,
1:07:22 the information on the persuasion,
1:07:23 uh,
1:07:24 mechanism doesn't appear to suggest that cheats are really more persuasive,
1:07:27 um,
1:07:28 conditional on visiting a household.
1:07:31 To now turn to the distributional impacts before concluding.
1:07:35 And the question here is what is the de facto incidence
1:07:37 of taxes when collected by chiefs relative to state agent collectors?
1:07:41 And there's reasons to think that taxation by
1:07:43 actors like chiefs might be regressive because in many
1:07:46 settings we observed the administration of informal taxes,
1:07:49 including in the DRC to be pretty regressive,
1:07:51 that lower,
1:07:52 um,
1:07:53 the poorer properties are bearing a higher burden of such taxes.
1:07:56 However,
1:07:56 chiefs might be more progressive in this context if they
1:07:58 possess superior local information that could allow for greater progressivity
1:08:02 in enforcement.
1:08:04 And so I'll study two dimensions here.
1:08:06 First,
1:08:06 how collectors
1:08:08 distribute tax visits and solicitation to households,
1:08:11 and then what the government may ultimately care
1:08:12 about most in the administration of taxation,
1:08:14 who ultimately pays the tax,
1:08:16 and how do they differ across
1:08:17 the treatment groups.
1:08:20 So in studying the visit margin,
1:08:21 I compare characteristics of households that are
1:08:23 revisited in the 2nd stage of the campaign
1:08:26 at the discretion of the collection teams
1:08:27 after registration across the treatment arms,
1:08:29 and I'll focus first on house quality
1:08:32 and I construct an index of house quality built for measures
1:08:34 like the quality of the walls of structures on the property.
1:08:37 And this table shows estimates from separate regressions,
1:08:40 regressing
1:08:41 um
1:08:42 local in blue and CLI in red,
1:08:44 um,
1:08:45 indicators for those treatments,
1:08:46 um,
1:08:47 on a measure of house quality,
1:08:48 this index where the comparison is the central treatment arm
1:08:51 and the sample is restricted to those who reported receiving a visit.
1:08:54 And so here we see to see that chiefs and CI CLI collectors are visiting lower quality
1:08:59 uh properties compared to central collectors or in
1:09:02 other words relying less on observable characteristics.
1:09:05 But it is still the case that the chief
1:09:06 NCLI collectors are targeting above median house quality,
1:09:10 uh,
1:09:10 properties.
1:09:10 So this figure,
1:09:11 uh,
1:09:11 above shows differences relative to central,
1:09:14 but I see positive differences,
1:09:15 um,
1:09:16 relative to lower quality households in terms of
1:09:17 the average house quality that collectors are targeting
1:09:20 in the local and CLI arms.
1:09:22 It's simply a less pronounced positive correlation.
1:09:24 So this suggests that chiefs and CI collectors might be targeting
1:09:27 other dimensions.
1:09:28 And these can capture unobservable characteristics
1:09:31 like the liquidity of the household,
1:09:32 past tax compliance,
1:09:34 and views of the government.
1:09:35 I also bring in the predicted ease of payment
1:09:37 measure I spoke about in the CLI treatment arm
1:09:40 and construct a measure of all these unobservable
1:09:42 margins that I call the payment propensity index.
1:09:45 And I repeat these comparisons to the central treatment arm conditioning
1:09:48 again to the set of households that report receiving a visit
1:09:51 and I see positive differences relative to central in the
1:09:54 local and CLI treatment arms which suggests that these
1:09:57 characteristics which we expect are positively correlated with compliance,
1:10:00 um,
1:10:01 are more likely to be used by chiefs and CLI collectors
1:10:03 when they go about soliciting um taxation by targeting visits.
1:10:09 And I can drill into this a little bit further by
1:10:11 looking at the probability of receiving a visit by different categories in the
1:10:15 combination of house quality and predicted ease of payment.
1:10:18 So the top two measures here,
1:10:20 um,
1:10:20 look at only households that we can rate as having above median house quality and
1:10:24 split it up within that category by having high or low predicted ease of payment.
1:10:28 And
1:10:28 I'm looking at the probability of receiving a
1:10:30 visit relative to the central treatment arm,
1:10:32 and we see that households that have high house quality but low predicted ease
1:10:35 of payment receive a visit less by chiefs as well as CI collectors.
1:10:38 And on the flip side,
1:10:39 in the bottom two measures,
1:10:41 properties with low house quality
1:10:43 but high predicted ease of payment are more likely to receive a visit.
1:10:46 So it appears that chiefs are targeting by ability to pay as well as house quality.
1:10:51 And I'll just mention that it doesn't appear that um
1:10:53 chiefs are targeting more their family and
1:10:55 friends or targeting them less relative to other
1:10:57 people um or targeting differentially people they have
1:11:00 a closer connection to in the neighborhood,
1:11:02 um,
1:11:02 like members of the same church or those
1:11:04 that they have a stronger social connection to.
1:11:08 So now finally the distribution of the tax burden.
1:11:10 I'll look first at how compliance varies
1:11:12 by the government's official property value gains,
1:11:14 but then look at these more direct measures
1:11:16 of economic need like income and liquidity.
1:11:19 So the first two columns rerun the original regression
1:11:22 estimate I showed you but separately by whether,
1:11:24 by which property value band of property is classified as belonging to,
1:11:27 and I see that the treatment effect
1:11:28 is coming through the lower-value property band,
1:11:30 that chiefs are bringing lower-value properties into the tax net.
1:11:35 This is reflected in column 3 by a lower average level of house quality
1:11:38 when I restrict the sample to the site of people who ultimately paid the tax.
1:11:42 However,
1:11:43 when I look at what are perhaps the most direct measures of need,
1:11:46 average monthly income and liquidity,
1:11:48 both from an index of measures related to this,
1:11:50 I don't see any differences across the compliers in terms of these measures.
1:11:54 Nor do I see your treatment effects on maybe the most direct measures
1:11:57 of need which are going without a meal um in the last week.
1:12:00 Um,
1:12:01 or not being able to pay school or health fees,
1:12:03 I don't see any differences across the treatment and control groups
1:12:06 even among the set of ultimate compliers of the tax.
1:12:11 So to take uh a stock,
1:12:13 we can place in the advantages column some increases in revenues,
1:12:17 uh,
1:12:17 as well as cost effectiveness.
1:12:18 The local treatment arm is in fact the only cost-effective treatment
1:12:21 as it doesn't incur substantial transportation costs on the government side.
1:12:24 I also see some evidence of trusting
1:12:26 government increasing when she collects taxes.
1:12:29 However,
1:12:29 on the disadvantaged side,
1:12:30 we have a small increase in bribes and some
1:12:33 increase in at least on-face regressivity by house quality,
1:12:36 though no ultimate differences in these most direct measures of economic need.
1:12:40 So to think about how the government can put together these
1:12:43 positives and negatives and make a decision over the collector type,
1:12:46 I assume first the government will trade off
1:12:48 the cost-effectiveness or net return on tax collection
1:12:51 but also potentially pay some social cost on bribes or even the
1:12:54 regressivity or tax burden beyond what they lose in terms of revenue.
1:12:58 So most of the bribes here are Replacing revenues,
1:13:00 as I said,
1:13:01 but the government may think that
1:13:02 increasing corruption by having chiefs involved in the process may
1:13:05 harm its ability to collect revenues in the future,
1:13:08 um,
1:13:08 and may place some value on that amount.
1:13:10 So in the paper,
1:13:10 I built a framework to
1:13:12 evaluate what,
1:13:13 how much the government would need to value the cost
1:13:15 of those additional bribes collected in local treatment arm.
1:13:18 To,
1:13:19 um,
1:13:19 prefer enlisting state agents over chiefs,
1:13:22 and I find that that amount would need to be 15 times higher
1:13:24 than the value of an additional $1 in net revenue in order to
1:13:28 prefer
1:13:28 uh employing state agents.
1:13:31 And a final consideration here is thinking
1:13:33 about citizen preferences for chief tax collection.
1:13:35 Uh,
1:13:36 and at online I see that citizens have more positive views of the
1:13:39 chief in areas where they collected taxes
1:13:41 relative to where state agents collected,
1:13:43 and they seem to appear to prefer having chiefs
1:13:45 as tax service tax collectors in the future.
1:13:47 Um,
1:13:47 so this is,
1:13:48 uh,
1:13:48 a question that I think we hope to collect data on the future
1:13:50 and see how this evolves over time as chiefs continue to be involved
1:13:54 as is expected in this process.
1:13:57 And so I'll just,
1:13:58 uh,
1:13:59 leave this,
1:14:00 uh,
1:14:00 slide up that,
1:14:02 that,
1:14:02 um,
1:14:02 the results here suggest that there's,
1:14:04 there's short-run benefits to collaborating like local,
1:14:07 with local elites like chiefs,
1:14:08 um,
1:14:09 in very low capacity settings like this,
1:14:11 but in the longer run,
1:14:12 state tax collection will likely produce higher revenues as the state
1:14:14 accumulates enforcement power and information about citizens that can be used,
1:14:18 uh,
1:14:18 in tax compliance.
1:14:19 But in the short run for state seeking to build rudimentary fiscal capacity,
1:14:22 engaging these actors can be a powerful tool.
1:14:25 Thank you.
1:14:28 Uh
1:14:30 Excellent.
1:14:31 Thank you,
1:14:31 Gabriel,
1:14:32 and thanks everyone.
1:14:34 Let me remind you that this is Gabriel.
- add-style
- lp-body-content