Honing the proper edge : CERP and the two-sided potential of military-led development in Afghanistan

Using a newer and expanded dataset as well as a survey of practitioner perceptions, this article adds to a recent body of literature on reconstruction and violence in Afghanistan. Data are taken from military-led development projects by way of the United States military’s Commander’s Emergency Response Program (CERP) and, to measure violence, from U.S. military Significant Activities (SIGACTs) reports. The results suggest that, at great cost, large-budget CERP efforts (those in excess of USD50,000 per project) may be associated with an increase in violence and thus counter-productive to military stability goals. In contrast, small projects (below USD50,000), which comprise a smaller proportion of total CERP allocations, are associated in statistically significant ways with reductions in violence. To explore why CERP projects may have these effects, the article also examines administrative modalities for CERP spending. The results suggest that timely, flexible expenditure of CERP funds are most effective at reducing violence. Influenced by the dubious outcome in Iraq and the ongoing difficulties in Afghanistan, a debate continues on the merits of counterinsurgency (COIN) and stability operations. Diplomatic and military operations are characterized by their decentralized and somewhat chaotic nature, which makes it difficult to produce general principles that practitioners might apply beyond Afghanistan and Iraq. The debate reflects this and has largely been based on personal anecdotes or historical interpretations. It has thus been difficult to capture objective “lessons learned” to improve policy design and implementation for conflict zones. Even with the end of major combat operations in Afghanistan, the importance of the topic has not waned for the international community. The need for evidence-based insight continues as weak states provide fertile ground for extremist groups such as the so-called Islamic State of Iraq and Syria (ISIS). According to instructions issued by the U.S. Department of Defense (DoD), all its military units must understand civilmilitary operations (CMO), the collective task of delivering on stability goals and facilitating nonmilitary aspects of national and international power. To support COIN and stability operations, the U.S. military developed the Commander’s Emergency Response Program (CERP), a CMO resource allowing military commanders to foster local development. To win the support of the local population, a key objective in stability operations and COIN, CMO adds to a commander’s ability to shape the combat environment, providing the option of offering “carrots” to accompany the “sticks” of military power. CMO is coordinated with other military operations such as raids or other offensive activities. While an offensive raid may result in a captured or destroyed enemy, it is often difficult to measure the effectiveness of the civil aspects of civilmilitary operations, such as the CERP program, which may also be a proxy for development aid at large. The question of whether military-led development aid reduces violence in the field is unsettled. In this article, I start to explore the question with an anecdotal response, namely, by asking a small group of Civil Affairs (CA) officers—the military’s experts in development—for their perceptions of CERP’s effectiveness. They report smaller CERP projects to be more effective than larger ones but are undecided on CERP’s overall impact on stability. I also use a new, previously unexploited dataset that covers 32 months of time and comprises data across 398 Afghan districts. I find that small CERP projects (<USD50,000) are associated with a statistically significant reduction in violence and that large CERP projects have the opposite effect. Further, to explore why some CERP projects might have been effective, I examine different ways to distribute CERP-related aid and find that projects conducted with cash on hand were more effective than those that required the clearing of administrative hurdles. The following sections summarize prior research, discusses THE ECONOMICS OF PEACE AND SECURITY JOURNAL ADAMS, Military-led development in Afghanistan p. 54 Vol. 10, No. 2 (2015) | doi:10.15355/epsj.10.2.53 The Economics of Peace and Security Journal — ISSN 1749-852X — http://www.epsjournal.org.uk © EPS Publishing, 2015. All rights reserved. For permissions, email: ManagingEditor@epsjournal.org.uk the CA officers’ survey, and describes the quantitative data, empirical methods, and findings. I conclude with some observations and recommendations for future work. The question remains: Does development work? Stereotypically reduced to, or otherwise cast in, terms of “grievance” versus “greed,” few frameworks of studying the relation between violent conflict and economic development are empirically nuanced and many findings are inconsistent in their predictions. Military doctrine adds to this literature, informed primarily by counter-revolutionary thinking and more recent, post-9/11 counterinsurgency work. While generally taken as a step forward for the military, the newest doctrine has also been criticized for offering simplified vignettes and a lack of actionable recommendations. Only recently has the topic received the attention of skilled empiricists. Grievance theories suggest that violence is due to political discord, ethnic differences, or other issues of identity and comparison, while greed perspectives suggest that aid can create an opportunity cost when joining a violent conflict and subsequently not receiving the aid. Both views have drawn criticism. Grievance models have suffered at the theoretical level for neglecting collective action problems when it comes to motivating violence and, empirically, lack of punch once accounting for relative income. Greed models, in contrast, have generally been easier to evaluate empirically, but rigorous tests have found little support of its constructs, especially when accounting for within-country trends. The U.S. military’s take on counterinsurgency, including CERP, has been labeled the “winning the hearts and minds” approach, formalized as an opportunity cost theory by Berman, Shapiro, and Felter (2011). The authors introduce the concept of conditionality, wherein aid is delivered subject to the provision of militarily relevant information to officials. Aid in exchange for information thus presumably allows improved targeting of insurgent forces. Indeed, the authors find that CERP spending in Iraq resulted in a statistically significant reduction in violence. Replicating this approach for the case of Afghanistan, neither Chou (2012) nor Child (2014) found effects significantly different from zero. Beath, Christia, and Enikolopov (2012; 2013) examine the impact of the Afghan National Solidarity Program (NSP) to see if Afghan government-led development reduces violence, but likewise do not find significant results. Chou (2012) also examines the NSP as well as USAID’s Local Governance and Community Development (LGCD) programs and again does not find results different from zero. Notably, Berman, Shapiro, and Felter (2011), Chou (2012), and Child (2014) all find a negative correlation between small CERP projects and violence, but only Berman, Shapiro, and Felter (2011) find it to be statistically significant. Conditionality is revisited by Berman, et al. (2013), and the additional concept of project expertise is introduced by way of Provincial Reconstruction Teams. Child (2014) questions conditionality as the mechanics of information sharing are not demonstrated. Berman, Shapiro, and Felter (2011) do use interviews in support of their conditionality concept, but no other work corroborates the empirical results by studying the perceptions of those who delivered the projects. In addition to standard quantitative work, this article therefore also reports results of a small survey of Civil Affairs officers. In the quantitative work I follow the model put forth by Berman, Shapiro, and Felter (2011), and revisited by Chou (2012) and Child (2014), but use a far more detailed dataset, explore the mode of CERP spending, and offer an interpretation from the perspective of military practitioners. I then go beyond the “Does development spending reduce violence?” question and add new insights as to “Why?” Data and methods Qualitative survey As mentioned, Civil Affairs officers are the U.S. military’s experts in civil-military operations. On 17 December 2013, I surveyed a small group of CA officers (n=9) who had utilized CERP funds. The survey was taken prior to obtaining my CERP and SIGACTs datasets so that any empirical results obtained would not influence the survey questions. Due to the small sample, the results cannot be said to be representative of the beliefs of all CA officers or military commanders that spent CERP funds in Afghanistan. The survey was not designed to focus on CERP per se and thus does not directly explore topics like conditionality directly. All interviewed CA officers were part of an active duty Army Special Operations unit. Most Civil Affairs units belong to the Reserve component of the U.S. military, so the CA officers I interviewed stand out for their experience across multiple theaters and the consistent use of CERP and similar tools. All have undergraduate degrees, and four hold master’s This article uses a new, previously unexploited dataset that covers 32 months of time and comprises data across some 398 Afghan districts. I find that small Commander’s Emergency Response Program (CERP) aid spending (<USD50,000) led to statistically significant reductions in violence and that large CERP had the opposite effect. Further, to explore why some CERP projects were effective, I examine different ways to distribute the aid and find that projects conducted with cash on hand were more effective than projects that needed to clear administrative hurdles. THE ECONOMICS OF PEACE AND SECURITY JOURNAL 


I
nfluenced by the dubious outcome in Iraq and the ongoing difficulties in Afghanistan, a debate continues on the merits of counterinsurgency (COIN) and stability operations. 1  Diplomatic and military operations are characterized by their decentralized and somewhat chaotic nature, which makes it difficult to produce general principles that practitioners might apply beyond Afghanistan and Iraq.The debate reflects this and has largely been based on personal anecdotes or historical interpretations.It has thus been difficult to capture objective "lessons learned" to improve policy design and implementation for conflict zones.Even with the end of major combat operations in Afghanistan, the importance of the topic has not waned for the international community.The need for evidence-based insight continues as weak states provide fertile ground for extremist groups such as the so-called Islamic State of Iraq and Syria (ISIS).
According to instructions issued by the U.S. Department of Defense (DoD), all its military units must understand civilmilitary operations (CMO), the collective task of delivering on stability goals and facilitating nonmilitary aspects of national and international power.To support COIN and stability operations, the U.S. military developed the Commander's Emergency Response Program (CERP), a CMO resource allowing military commanders to foster local development.To win the support of the local population, a key objective in stability operations and COIN, CMO adds to a commander's ability to shape the combat environment, providing the option of offering "carrots" to accompany the "sticks" of military power.CMO is coordinated with other military operations such as raids or other offensive activities.While an offensive raid may result in a captured or destroyed enemy, it is often difficult to measure the effectiveness of the civil aspects of civilmilitary operations, such as the CERP program, which may also be a proxy for development aid at large. 2  The question of whether military-led development aid reduces violence in the field is unsettled.In this article, I start to explore the question with an anecdotal response, namely, by asking a small group of Civil Affairs (CA) officers-the military's experts in development-for their perceptions of CERP's effectiveness.They report smaller CERP projects to be more effective than larger ones but are undecided on CERP's overall impact on stability.I also use a new, previously unexploited dataset that covers 32 months of time and comprises data across 398 Afghan districts.I find that small CERP projects (<USD50,000) are associated with a statistically significant reduction in violence and that large CERP projects have the opposite effect.Further, to explore why some CERP projects might have been effective, I examine different ways to distribute CERP-related aid and find that projects conducted with cash on hand were more effective than those that required the clearing of administrative hurdles. 3 The following sections summarize prior research, discusses the CA officers' survey, and describes the quantitative data, empirical methods, and findings.I conclude with some observations and recommendations for future work.

The question remains: Does development work?
Stereotypically reduced to, or otherwise cast in, terms of "grievance" versus "greed," few frameworks of studying the relation between violent conflict and economic development are empirically nuanced and many findings are inconsistent in their predictions.Military doctrine adds to this literature, informed primarily by counter-revolutionary thinking and more recent, post-9/11 counterinsurgency work.While generally taken as a step forward for the military, the newest doctrine has also been criticized for offering simplified vignettes and a lack of actionable recommendations.Only recently has the topic received the attention of skilled empiricists. 4 Grievance theories suggest that violence is due to political discord, ethnic differences, or other issues of identity and comparison, while greed perspectives suggest that aid can create an opportunity cost when joining a violent conflict and subsequently not receiving the aid.Both views have drawn criticism.Grievance models have suffered at the theoretical level for neglecting collective action problems when it comes to motivating violence and, empirically, lack of punch once accounting for relative income.Greed models, in contrast, have generally been easier to evaluate empirically, but rigorous tests have found little support of its constructs, especially when accounting for within-country trends. 5 The U.S. military's take on counterinsurgency, including CERP, has been labeled the "winning the hearts and minds" approach, formalized as an opportunity cost theory by Berman, Shapiro, and Felter (2011).The authors introduce the concept of conditionality, wherein aid is delivered subject to the provision of militarily relevant information to officials.Aid in exchange for information thus presumably allows improved targeting of insurgent forces.Indeed, the authors find that CERP spending in Iraq resulted in a statistically significant reduction in violence.Replicating this approach for the case of Afghanistan, neither Chou (2012) nor Child (2014) found effects significantly different from zero.Beath, Christia, and Enikolopov (2012;2013) examine the impact of the Afghan National Solidarity Program (NSP) to see if Afghan government-led development reduces violence, but likewise do not find significant results.Chou (2012) also examines the NSP as well as USAID's Local Governance and Community Development (LGCD) programs and again does not find results different from zero.Notably, Berman, Shapiro, and Felter (2011), Chou (2012), andChild (2014) all find a negative correlation between small CERP projects and violence, but only Berman, Shapiro, and Felter (2011) find it to be statistically significant. 6 Conditionality is revisited by Berman, et al. (2013), and the additional concept of project expertise is introduced by way of Provincial Reconstruction Teams.Child (2014) questions conditionality as the mechanics of information sharing are not demonstrated.Berman, Shapiro, and Felter (2011) do use interviews in support of their conditionality concept, but no other work corroborates the empirical results by studying the perceptions of those who delivered the projects.In addition to standard quantitative work, this article therefore also reports results of a small survey of Civil Affairs officers. 7 In the quantitative work I follow the model put forth by Berman, Shapiro, and Felter (2011), and revisited by Chou (2012) and Child (2014), but use a far more detailed dataset, explore the mode of CERP spending, and offer an interpretation from the perspective of military practitioners.I then go beyond the "Does development spending reduce violence?"question and add new insights as to "Why?"

Qualitative survey
As mentioned, Civil Affairs officers are the U.S. military's experts in civil-military operations.On 17 December 2013, I surveyed a small group of CA officers (n=9) who had utilized CERP funds.The survey was taken prior to obtaining my CERP and SIGACTs datasets so that any empirical results obtained would not influence the survey questions.Due to the small sample, the results cannot be said to be representative of the beliefs of all CA officers or military commanders that spent CERP funds in Afghanistan.The survey was not designed to focus on CERP per se and thus does not directly explore topics like conditionality directly.
All interviewed CA officers were part of an active duty Army Special Operations unit.Most Civil Affairs units belong to the Reserve component of the U.S. military, so the CA officers I interviewed stand out for their experience across multiple theaters and the consistent use of CERP and similar tools.All have undergraduate degrees, and four hold master's This article uses a new, previously unexploited dataset that covers 32 months of time and comprises data across some 398 Afghan districts.I find that small Commander's Emergency Response Program (CERP) aid spending (<USD50,000) led to statistically significant reductions in violence and that large CERP had the opposite effect.Further, to explore why some CERP projects were effective, I examine different ways to distribute the aid and find that projects conducted with cash on hand were more effective than projects that needed to clear administrative hurdles.
degrees.In addition to deployments to Afghanistan, seven of the nine had previously been deployed to Iraq and in Africa as well.To gain an understanding of how CERP was employed in the field, and for observations regarding the program's efficacy, these officers thus constituted a suitable group of interviewees.
Quantitative data and methods I use U.S. military data, built up from the Combined Information Data Network Exchange (CIDNE) and Significant Activities (SIGACTs) datasets from Afghanistan between 2011 and 2013.They are declassified, yet secure (not readily available), and assembled in a database called the International Distributed Unified Reporting Environment (INDURE). 8 SIGACT data is collected directly from the field of operations, which results in increased granularity.While the dataset does not cover the same time period as Chou (2012) or Child (2014), the scale of increased observations is striking: Chou (2012) reports more than 60,000 insurgent-initiated events over nearly eight years and Child (2014) observes 3,599 events over a four-year period in the Worldwide Incidents Tracking System (WITS).For the two-and-a-half-year period covered in this article, the cleaned dataset contains more than 107,350 insurgent-initiated events, consistent with numbers reported by the International Security Assistance Force-Afghanistan.As Child (2014) notes, WITS could comprise a subset of CIDNE data (which, in turn, is part of INDURE).The magnitude of the difference in captured events may be responsible for diluting the regression results toward zero in previous work.While the SIGACTs data used here therefore should not cause attenuation bias, they do come with limitations.For example, while SIGACTs tends to include them, and with finer granularity, it sometimes does not include nonmilitary violent events.Other databases also suffer from this problem.Neither does SIGACTs include the magnitude of violence, so that a large battle that kills many combatants or civilians cannot be distinguished from a report of shots fired without injuries.In line with previous research, I limit SIGACTs to enemy-initiated events, which include enemy attacks and explosive hazards.
CERP data was ported directly from CIDNE to INDURE.The database contains detailed project information, including project start and end dates, location, and total monetary outlays.For the period of January 2011 through August 2013, I look at 4,148 CERP projects, often coded with an additional CERP category-comprising 2,003 projects known as "bulk" CERP-to which this article pays special attention.Initially covering CERP spending in excess of USD30 million per month, at the beginning of 2011, this declined to less than USD5 million per month by mid-2013, which is congruent with SIGAR-reporting for this the time period.As shown in Table 1, two bulk CERP categories were approved at the battalion level.This could result in some measurement error for projects considered small and large in past research, as bulk spending at the small and the battalion levels share many attributes.It may even mean that bulk is a more useful distinction than small or large, as the mechanism of spending may have been more important than the magnitude of the outlay.The CERP dataset misses some opportunities to understand the mechanics of development spending in its structure.For instance, there are codes for categories of projects, which provide a nuanced view of whether an outlay was for agriculture or for temporary guards.However, this information is inconsistently coded in the database, especially for bulk CERP: A single bulk draw resulted in a single project code, yet there could have been twenty USD2,500 projects in a single USD50,000 bulk CERP entry in the database.One is thus unable to explore the efficacy of the agriculture or temporary guards projects. 9 The two INDURE datasets were merged, geographically Notes: For normal CERP, projects are identified and funds are requested.Details are provided here to give a sense of the bureaucracy involved in delivering CERP projects.The higher the monetary outlay, the less responsive the CERP project is likely to be due to required documents and signatures.There is also a "be prepared" case, called bulk CERP, where funds are drawn in advance of identifying needs and kept "on hand."The mechanism of spending changes as bureaucratic levels increase.At the small unit level, there is little contracting involved, and money goes directly to addressing local needs.As the level increases, contractors and bidders are required, which adds time to the allocation process and can create a zero-sum game for contract winners.
parsed to the district level, and adjusted to a per capita basis to support district fixed-effects analysis.I use the same model as previous researchers, with 398 districts and 34 provinces.As a balanced panel, there are accordingly 12,736 observations.CERP project spending is evenly distributed over a project period, consistent with the methods of Berman, Shapiro, and Felter (2011), Chou (2012), andChild (2014), as well as broken into administrative tranches of small and large based on whether projects exceeded a USD50,000 spending threshold.District population numbers are from 2010 and are used to normalize across districts by population.Events and spending are thus reported as SIGACTS per 1,000 and CERP spending per capita.Summary statistics are shown in Table 2. Whereas Berman, Shapiro, and Felter (2011) examine half-years, and Chou (2012) and Child (2014) use one-month lags, I employ a series of lags, ranging from one to three months.A larger lag period allows time for "hearts and minds" to be won over by CERP delivery and for a population to update its priors.Because my specification uses lagging to infer the effects of CERP, if any, it could be that statistically significant results occur randomly over time.I thus run quarterly checks as well to ensure that my results are robust.

Qualitative results
As a group, Civil Affairs officers are uncertain about CERP's impact.Of the nine officers who disbursed CERP funds in Afghanistan, four did not believe that this increased effectiveness, while the remainder believed that CERP was "effective."None said that it increased "stability."Evidently, even the officers who judged development outcomes as effective saw stability, as a military goal, as not achieved.Regardless, all officers believed that small projects (<USD25,000) could be more effective than large ones.This is consistent with Berman, Shapiro, and Felter's (2011) findings and, as it turns out, accords with my own results as well. 10 A significant critique of Berman, Shapiro, and Felter (2011) is that there is no direct evidence of information sharing.However, some survey results are telling.In response to, "How did you allocate your projects?" the CA officers responded with a range of answers that suggest negotiation and interaction with the local population: Clearly, some CERP projects were negotiated, and some information would presumably be shared.While this information may not have amounted to tips resulting in better insurgent targeting, it still represents a two-way information flow that could conceivably affect insurgent targeting as well as improve service delivery to Afghan communities.

Quantitative results
Table 3 shows the quantitative results of regressing SIGACTs on CERP, adding district-level fixed effects, accounting for previous violence trends, and adding a series of CERP lags to better understand the relation between CERP and SIGACTs.In running a basic OLS regression (col.1), small CERP are strongly and positively correlated with violence.From col. 2 onward, district-level fixed effects and seasonal controls are added, and both small and large CERP are lagged.Signs change with lags, and the results become statistically significant at the 1% level for most of the small CERP results with three lags (col.2).With a full series of lags, from 1 to 3 months, large CERP are positively associated with violence at the 1% level (col.3).Following Berman, Shapiro, and Felter (2011), a variable for previous violence trends is added in regressions (2), (4), (5), and (7).With the preceding month's level of violence thus taken to account, the coefficient on small CERP then becomes significant at the 5% level with two lags (col.4), and large CERP remain positively associated with violence after 3 lags.When violence outliers are removed in regressions ( 5), ( 6), and (7), and every variable except the prior month's violence trend is taken to account, the small CERP variable is significant at the 1% level and the large CERP variable at the 10% level (col.6).Adding violence trends, the coefficient on small CERP remains negative and significant at the 1% level and, for large CERP, is positively associated with violence at the 5% level (col.7).Quarterly analysis adds to the robustness of the results, as shown in Figure 1 (on p. 58). 12One way to view the monthly results is to say that a one dollar increase in small CERP spending per capita is associated with the reduction of about 8 violent events per 100,000 people (col.7).For context, recall from Table 2 that the mean level of violence is 13.4 violent events per 100,000 people.Thus, on average, each dollar of small CERP per capita spending is associated with a greater than 59 percent decrease in rebel-initiated violent events.In contrast, large CERP projects have a positive correlation with violence, an effect that is statistically significant at the 5% level, with a large number of controls and with violence outliers dropped (col.7).Here, a one dollar per capita increase in large CERP spending is associated with an increase in violence of nearly one violent event per 100,000 people.Since large CERP spending exceeded small CERP by a factor of four on a per capita basis-and at times by an order of magnitude-large CERP projects were nonproductive, or even counter-productive, and at great cost.
Notes: *** p<0.01, ** p<0.05, * p<0.1.Robust standard errors in parentheses.Seasonal controls are dummies created for each month to control for seasonal and monthly trends.Lag periods indicate months, where +1 indicates a control for violence trends in the previous month.
Dropping violence outliers restrict regressions to the 1st to 99th percentiles of violence incidents.Fixed effects are to the district level.
Other interactions?Berman, et al. (2013) suggested complementarity between security efforts and development, a case that is similar to this article.Inherently, there is some level of security provision in conjunction with CERP, due to the military's presence.Prior research also separated CERP into two categories, small and large, with USD50,000 being the threshold.Recall from Table 2 that small projects could be approved by a battalion commander, with reduced administrative hurdles to clear.In contrast, large projects required additional paperwork and approvals, and may have involved contracting.These details mean that small CERP projects better fit the model specified by Berman, Shapiro, and Felter (2011).Berman, Shapiro, and Felter (2011), Berman, et al. (2013), and Chou (2012) suggest that conditionality is important, and Child (2014) points out that the mechanics of conditionality need to be explored in more detail.The difference in the mechanics of bulk and non-bulk CERP offer one such opportunity.Table 4 shows a summary The effect of small or large CERP projects on incidents of violence (per 1,000 pop.), by quarter.The figure demonstrates that the results are robust to quarters.The scatter diagram on the left, for small CERP projects, shows a statistically significant and negative coefficient, similar to col. 5 in Table 3 but for a lag of one quarter.In contrast, large CERP projects, on the right, have a positive coefficient for the first quarter, although not statistically significant in this specification.Bulk is consistently on the right in these observations.The data was coded "bulk" or "non-bulk."Missing observations were dropped.I interpret the results to suggest that bulk CERP was more effective than planned (non-bulk) CERP projects at some levels, perhaps to the USD100,000 level (BN in the table ).There are only 19 projects for the regimental level for bulk, and so I do not focus on this result.
1.Both Col. Gian Gentile's (2013) and USAID employee Peter van Buren's (2011) books are examples of this debate.See also Fischerkeller (2011, p. 139)  comparison of CERP on SIGACTs regressions, focusing on the three-months lag, but otherwise with a specification similar to that of col.7 in Table 3. Non-bulk results are shown in the oddnumbered columns and bulk CERP results in the evennumbered ones.Rows reflect administrative levels corresponding to the bulk categories in Table 1.The findings suggest that bulk CERP projects were more effective as compared to non-bulk CERP projects (cols. 1 and 2).Spending above USD50,000 but below the USD100,000 approval threshold (BN level, col.4) also appears to have been effective.
The signs for all CERP are negative below the BN level, but once the BDE level is reached, non-bulk CERP has a negative coefficient, and bulk CERP has a positive coefficient, although neither is statistically different from zero.The Regimental results are discomforting-the coefficients are positive and statistically significant-but then there were few Regimental Bulk CERP projects in the database.In all, it appears that the subcategory of bulk CERP projects warrants further investigation.Since bulk funds are on hand when projects are allocated, this could mean that the speed with which projects were allocated is important to their success in reducing violence, although this hardly demonstrates conditionality as described by Berman, Shapiro, and Felter (2011).However, once one considers the interactions of CA officers with the local population and governments when allocating CERP, it is not a stretch to imagine that the ability to deliver projects quickly could make those CA officers more credible in the eyes of the local population or government, which could, in turn, make the opportunity cost of a project more tangible.

Conclusions
This article empirically tests the hypothesis that militarily-led development aid may have reduced violence in the Afghan context.I find that small CERP projects reduced violence, in line with the findings Berman, Shapiro, and Felter (2011) for Iraq.However, without a clear mechanism to explore the information-sharing theory, I focus-in the Afghan case-on the reduced administrative process required, and add the observations of CA officers for context and corroboration of the empirical results.It seems that the mechanism by which development money is spent is a critical factor, and that the timeliness of CERP delivery may have made a difference.Focus on small projects, as opposed to large-scale undertakings seems appropriate, at least when it comes to delivering reduced violence in the short term.More attention should be paid to the mechanism of CERP project spending and perhaps to development aid in general.Due to the structure of current datasets, this will be difficult to do.The U.S. military, other militaries, and aid organizations could add more detail to their datasets to facilitate empirical work.Moving away from paper and email record-keeping methods toward database records for the purpose of monitoring aid or CERP projects will be key to make this happen.Future qualitative surveys should explore the conditionality of aid, and more empirical work should focus on the degree of flexibility in allocated development projects.
Future research should also examine whether aid's "carrots" need to be paired with "sticks" of other military operations, or whether aid can be deployed on its own when attempting to reduce violence.Meanwhile, for policymakers as well as commanders who may be uncertain of CERP's utility, it appears that programs like CERP should continue to have a place in military operations, perhaps more so than hitherto.But this resource appears to be a double-edged sword, with the ability to both help or harm stability operations, depending on how the sword is wielded.Notes 4. Few frameworks: For a sample of this literature, see, e.g., Collier, et al. (2006); Blattman and Miguel (2010); Beath, Christia, and Enikolopov (2013); Berman, Shapiro, and Felter (2011);Chou (2012). Counter-revolutionary: Galula (1964).Counterinsurgency: Nagl (2002); U.S. Army Field Manual FM 3-24 (HQ, U.S. Army, 2006a).Empiricists: Blattman and Miguel (2010).
6. NSP: The specification in Beath, Christia, and Enikolopov (2012;2013) is based on proximity to development aid.The mid-line survey, published in 2012, suggests that NSP may have had a significant impact on reducing violence, but the end-line survey of 2013 does not confirm this.Any violencereducing impact of NSP may have worn off, or there may have been other exogenous effects.LGCD/NSP: Chou (2012) uses district-based specifications similar to those she uses for CERP, rather than the proximity-based specifications of Beath, Christia, and Enikolopov (2012;2013).Notably: In the Berman, Shapiro, and Felter (2011) specification, first differences are over 6 months periods, a specification I modify in my work as reported later on in this article.7. Nonetheless, the survey did not ask questions about conditionality or an increase in tips from locals that could lead to improved targeting of insurgent forces.8. INDURE includes Significant Activities (SIGACTs) as well as CERP spending information, all geolocated and timestamped.The data was cleaned using basic qualitative checks, code books, interviews and the sharing of project files with others who also have used these databases.See Lyall (2010).9. SIGAR-reporting: See SIGAR (2014).10.Survey details available from the author upon request.11.MISO: Military Information-Support Operations; formerly Psychological Operations.
12. Quarterly numerical results are available from the author upon request.

Response 1 :
Based on the deliberation of the valley-wide shura.[A shura is a de facto Afghan unit of governance and decisionmaking.]Response 2: Based on military objectives and local needs.What they needed wasn't always in line with what the local government would ask for.Civilian sampling and surveys with MISO personnel assisted in the decision. 11Response 3: In support of mobilizing district governance and development seminar.

Figure 1 :
Figure 1:The effect of small or large CERP projects on incidents of violence (per 1,000 pop.), by quarter.The figure demonstrates that the results are robust to quarters.The scatter diagram on the left, for small CERP projects, shows a statistically significant and negative coefficient, similar to col. 5 in Table3but for a lag of one quarter.In contrast, large CERP projects, on the right, have a positive coefficient for the first quarter, although not statistically significant in this specification.