Prosocial Project Management in Conflict Areas

2.1. World Bank Development Projects

The primary objective of the World Bank is to support the development of low- and middle-income countries by funding and supporting the design, preparation, implementation, supervision, and evaluation of development projects. The World Bank collaborates with recipient countries to implement projects that address development priorities. These projects typically last about 10 years, from identification to implementation and evaluation, and begin by having the borrower (the country’s government) and the World Bank jointly identify needs, beneficiaries, and priorities. With a draft of potential projects, the Bank and the borrowing government conduct feasibility studies to assess the social and environmental impacts of the projects, their expected performances, and the intended beneficiaries. After the Bank approves the government’s proposal and verifies that a project is consistent with the Bank’s guidelines, the government designates an implementing agency and starts the implementation.

When a project is completed, the World Bank’s Independent Evaluation Group (IEG) evaluates the project against its original objectives, aiming to provide comparable ratings across projects, time, and geography. This could be a challenging task. To obtain consistent and comparable ratings, the evaluation process follows procedures and frameworks consistent with international evaluation norms from the OECD and the UN (World Bank, 2013), and applies standardized evaluation criteria and definitions, in-depth product reviews, and semi-structured interviews. The evaluation includes a technical peer review and interviews with country stakeholders, World Bank staff, and donor agencies. After field visits and a thorough examination of the project’s documentation, the experts issue a project performance report that contains the project performance ratings. Although not perfect and subject to human bias, project evaluations offer indispensable, harmonized data to study project performance.

2.2. Adaptive Projects

The adaptable program loan (APL) is one of several lending instruments offered by the World Bank between 1998 and 2013. A lending instrument stipulates the financing and operational rules governing World Bank projects. The adaptable program loan, and later, the programmatic approach, were instruments born from the need to be adaptive and meet the client’s needs (The World Bank, 2017). Adaptive projects were designed to be implemented as a sequence of phases, as opposed to the rest of the projects: standalone projects with fixed sets of activities. After each phase, adaptive projects would be reassessed and adjusted to fit the local context and the new information encountered during their implementation. By allowing the borrowing country to incorporate corrections and adjustments, an adaptive project would allow “adaptive learning, and for the phases to adjust based on lessons from earlier and ongoing phases” (The World Bank, 2017: p. 14).

Examples of the adaptive approach. The adaptive nature of APLs comes from the opportunity for managers to modify project characteristics during implementation, providing them with flexibility to respond to changing circumstances based on results from previous stages and feedback from stakeholders. For illustration, Table 1 provides four examples of how adaptability helped managers adjust course while navigating political instability. From these examples, first, we observe the critical role of the government in the performance of World Bank projects. When political instability, social unrest, or armed conflict shocks a country, it will likely influence project performance via deterioration of the government’s capacity, priorities, or incentives. Second, projects in politically unstable countries can leverage adaptability in several ways, including, by changing project scope and focus, adjusting or canceling activities, selecting new implementing agencies or partners, simplifying the project’s work breakdown structure, prioritizing critical tasks, reassessing critical activities, extending completion times, suspending a project temporarily or indefinitely, extending the project’s due date, and relocating, suspending, or soliciting new funds.

However, the adaptable program was discontinued in 2013. In 2017, the program transitioned to what is now known as the multiphase programmatic approach (MPA)—we do not yet have evaluation data for projects under this approach. What motivated this transition? The adaptable program “worked well in terms of client engagement around a long-term perspective on sectoral reform;” (The World Bank, 2017: p. 8) however, its decline by 2010 was marked by excessive long processing times and high costs compared to other types of projects—an adaptable program loan, for example, required World Bank board-level approvals to make project adjustments and advance the project (Kenyon, 2019). By 2010, the program represented only 2% of the World Bank lending (The World Bank, 2017). MPAs, unlike APLs, do not use triggers (conditions that must be met before a subsequent phase of a project can be approved), require less paperwork, and feature a more streamlined process for approving subsequent phases and decision making (The World Bank, 2017). Despite APLs’ drawbacks, between 1998 and 2013, the program funded 429 projects and represented almost 9% of all lending instruments, as shown in Table A.2 (E-Companion).

3.1. Research Framework

This paper has both theoretical and practical motivations. It contributes to operations and project management literature by studying how adaptability influences project performance during armed conflicts. From a practitioner’s perspective, although adaptability is a well-known strategy in fluid and unstable contexts (Van Voorst and Hilhorst, 2017), it is unclear when and how to implement it, and what project characteristics reinforce or moderate its effects.

Following this motivation, we investigate four relationships, as shown in Figure 1—the definitions of all the variables introduced in this framework are presented in Table 2. First, we evaluate the effects of armed conflict on project performance, meaning by performance the extent to which a project reaches its original objectives (R1). Second, we analyze APLs (adaptive projects) and how their “adaptive” effect moderates the conflict-performance relationship (R2). Third, given APLs’ objective to “incorporate learning and adaptiveness” (The World Bank, 2017: p.9), we evaluate whether “experience” can reinforce APL effects (R2a), experience understood in terms of past projects implemented and the history of conflict in the focal country. Finally, we examine how project complexity, in terms of project scope, can challenge the implementation of APLs and the adaptive effect (R2b)—when the number of actors and development areas involved in APLs increases, project adjustments tend to be more complicated and resource-intensive, according to project evaluation reports.

OPEN IN VIEWER

Figure 1.

Research framework. Note. The figure shows the main variables and relationships (Rs) examined in this study.

OPEN IN VIEWER

Table 2.

Sample variables definitions, units, and measurement.

variable	definition	units	timing (project phase)	assessor (decision maker/source)
adaptive	(1) adaptive projects, (0) non-adaptive projects	binary	design	recipient
performance	the extent to which the projects’ objectives are achieved: (1) highly unsatisfactory, (2) unsatisfactory, (3) moderately unsatisfactory, (4) moderately satisfactory, (5) satisfactory, (6) highly satisfactory	ordinal	completion	IEG
experience	the number of projects implemented before the project’s start date in the focal country	ln	design	Bank
costOverruns	the difference between the project’s final costs and its initial budget; 0 in cost-saving cases	ln	design and completion	Bank
benefitsOverest	overestimation of project benefits, computed as the difference between the project’s initial economic rate of return and its final estimate; 0 in benefit-exceeding cases	ln	design and completion	IEG
conflict	the average annual conflict intensity during project implementation, where, each year, following Uppsala’s definitions, the conflict intensity is 0 if the country experiences between 0 and 24 battle-related deaths, 1 if the conflict results in at least 25 battle-related deaths, and 2 if the conflict leaves more than 1,000 deaths	conflict intensity	all phases	Uppsala
pastSuccessRate	the average past success rate for the focal country on previous projects, where a successful project has a satisfactory or highly satisfactory performance	percent	design	Bank
conflictExp	(1) the country experienced conflict five years before the project’s start date, (0) the country did not experience prior conflict	binary	design	Uppsala
designQual	quality of the identification, preparation, and appraisal of projects to achieve planned development outcomes using a six-point scale, from highly unsatisfactory to highly satisfactory	ordinal	completion	IEG
scope	the number of development areas that the project covers	areas	design	Bank
budget	the project’s total budget	ln	design	Bank
duration	the project’s duration	years	completion	Bank
monitoringQual	the project’s monitoring and evaluation technology assessment on a six-point scale rating, from highly unsatisfactory to highly satisfactory	ordinal	completion	IEG
noProjects	the number of projects occurring simultaneously in the country at the project’s start date	projects	design	Bank
evalDur	the project’s performance evaluation duration	years	completion	Bank
GDPc	the country’s annual gross domestic product per capita in constant prices	ln	design	Bank
education	the country’s annual government expenditure on education (% of GDP)	percentage	design	Bank
oilRents	the country’s annual difference between the value of crude oil production and total costs of production (% GDP)	percentage	design	Bank
agriEmpl	the country’s annual employment in agriculture (% of total employment)	percentage	design	Bank
politicalRisk	the country’s perceptions of the likelihood of political instability and/or politically-motivated violence, including terrorism	normal	design	Bank
regulatoryQual	the government’s ability to formulate and implement sound policies and regulations that permit and promote private sector development	normal	design	Bank
govEffectiveness	the government’s quality of public services, the quality of the civil service and the degree of its independence from political pressures, the quality of policy formulation and implementation, and the credibility of the government’s commitment to such policies	normal	design	Bank
corruptionCountry	the extent to which public power is exercised for private gain, including both petty and grand forms of corruption, and the degree to which the state is influenced by elites and private interests	normal	design	Bank

Notes. The table describes the variables employed in the main analyses of this paper. Normal refers to the standard normal distribution. Timing refers to the project phase in which the variable is obtained or calculated: during the design, implementation, or completion phases. The assessor indicates who is the appraiser or source of information for the calculated variables.

Although this study is exploratory, it contributes to the ongoing discussion on adaptive and uncertainty management in project management and operations management. To position our work, we explain next how our research framework aligns with, and extends, key literature in both fields.

3.3. Project Management

In project management, the notion of project adaptability is intertwined with the concept of agility. Although there is no single definition of agility, the Project Management Institute (PMI), in its Agile Practice Guide, suggests that agility is a project management approach aimed at addressing project uncertainty, an approach particularly useful in projects with unclear contingencies and unknowns. In project management, uncertainty emerges in the form of known unknowns and unknown unknowns—unlike known unknowns, unknown unknowns are unrecognized uncertainties that the project manager is unaware of (Ramasesh and Browning, 2014). This uncertainty framework creates the basis for project design methodologies and evaluation tools to uncover and mitigate known and unknown project risks (Herroelen, 2005; Ramasesh and Browning, 2014; Shenhar, 2001).

The agile method, in particular, consists of dividing a project into several iterations to find feasibility after continuous cycles of feedback, evaluation, and adaptation (PMI, 2017). The PMI’s PMBOK Guide suggests that agility enables continuous learning about the next project steps and continuous adaptation to new requirements. In the agile approach, less time is spent planning at the beginning of the project. In fact, project teams delay key decisions to observe new information and, in conjunction with customers, make changes during the course of the project. At the end of each iteration, the project team and the customer initiate changes based on what was learned from previous iterations. In theory, an agile approach works well in complicated and complex projects, whereas simple and stable projects are better suited for the linear method: the waterfall approach. Table 3 summarizes the key differences between the two approaches.

Agility may improve the effectiveness of disaster response projects. Ali Morad and Chandrashekhar (2022), for instance, evaluate project flexibility and agile project management for large public health projects during the COVID-19 crisis. The authors examine screening and vaccination projects and find the need of flexibility for adequate strategic planning, communication, and context-driven decision-making. Similarly, Baham et al. (2017) explore the agile methodology in disaster response scenarios and claim that a flexible and adaptive methodology is necessary for efficient disaster recovery. Furthermore, agility may attenuate the “progression fallacy,” a behavioral effect that consists of spending “too much time on early project phases at the expense of later ones” (Lieberum et al., 2022: p. 2294). The agile approach, with the method “Scrum,” addresses this point by leveraging the idea of iterations, phase-specific goals, time-boxed progressions (or sprints), continuous improvement, and division of projects into smaller components (Lieberum et al., 2022).

However, the notion of agility faces challenges in the prosocial sector, as such projects involve comprehensive collaboration among multilateral organizations, governments, and implementing agencies. Prosocial projects, while featuring high levels of uncertainty and coordination complexity, often cover broad areas, from relief operations to long-term sustainable development projects. These challenges motivated APLs to include (i) piloting to test assumptions and fit, (ii) heavy monitoring, feedback, and learning, and (iii) course correction in the life of a single project or throughout a series of projects (The World Bank, 2017). In practice, the adaptability inherent in APLs could have been slowed by bureaucracy and high processing complexity—an APL would require, for example, approvals from the World Bank board to finalize project adjustments, leading to long processing times and high costs (Kenyon, 2019).

Although it is unclear how effective adaptability is in areas of political instability and conflict, there have been attempts to tailor adaptive management to development aid contexts. Mahmoud Saleh and Karia (2022), for instance, develop an adaptive project management model for international development and aid projects that integrates notions from Total Quality Management and Core Humanitarian Standards. The model identifies challenges in the agile approach when implemented in the development sector and suggests highlighting the concepts of accountability, coordination, leadership, and effective monitoring and evaluation. In the same spirit, Saleh and Karia (2020) and Kelly et al. (2022) critically review the standard PMBOK guide to align it with the needs of International Development and Aid Projects, including an improved use of evaluation information and an adaptable monitoring framework.

The project management literature highlights the relevance of an enhanced agile approach featuring adaptiveness to deal with uncertainty and volatility (Baweja and Venugopalan, 2015). However, its effectiveness has yet to be tested for prosocial and nonprofit projects in contexts of political instability and conflict.

3.4. Operations Management

There are several differences between prosocial projects and other types of operations, such as service and manufacturing operations. Projects are unique, highly customizable, and less predictable than assembly lines and continuous processes. Yet, disruptions, such as armed conflict, can affect projects just as much as other types of operations, including humanitarian and commercial ones (Jola-Sanchez, 2022; Jola-Sanchez et al., 2016). In dealing with disruptions, and risk in general, a great deal of research in operations management focuses on risk management strategies using the concepts of incentives and learning (He et al., 2014), cooperation and supplier collaboration (Gopalakrishnan and Sankaranarayanan, 2023), risk monitoring strategies (Hoffmann et al., 2013), strategic resource allocation (Jola-Sanchez and Serpa, 2021; Zhuang and Bier, 2007), and vulnerability analysis (Kleindorfer and Saad, 2009; Knemeyer et al., 2009). Limited literature evaluates the role and effects of adaptive management in contexts of high uncertainty. Some literature, however, explores flexibility in the context of supply chain disruptions. Saghafian and Van Oyen (2016), for example, study flexibility and supply chain resiliency by analyzing a supply chain under dynamic disruption risk; the authors show that even a slight increase in supply chain flexibility increases supply chain resilience.

Our paper examines nonprofit operations; operations that, rather than aiming to enhance profitability, aim to address critical societal goals—long-term development goals. The body of knowledge in operations management related to commercial applications does not necessarily fit the context of nonprofit operations (Gupta and Starr, 2017). While profit-seeking organizations focus on reducing costs or increasing income within a rather precise universe of stakeholders and rules, nonprofit projects are society-centric and involve the collaboration of governments and the international community. Prosocial projects face not only the pressure from donors and society for results, but also the expectations, administrative burden, and limitations inherent in the public sector.

Given the unique context of prosocial projects, adaptive management cannot be evaluated through the lens of profit-seeking organizations. And although nonprofit sector research has grown significantly in the last decade (Berenguer et al., 2017) by studying phased and results-oriented project management approaches (Devalkar et al., 2017), fund disbursement processes for procurement (Gallien et al., 2017), revenue management (Tereyağoğlu et al., 2017), resource management (Lorca et al., 2017), and distribution processes (Natarajan and Swaminathan, 2017; Toyasaki et al., 2017), prosocial project management is still understudied. Thus, by studying how adaptive management influences the performance of prosocial projects funded by the multilateral system and implemented by public sector organizations, this article contributes to the nonprofit literature and the stream of research aimed at improving aid effectiveness.

4.1. Data

We combine four data sources. First, we obtain project-level data from the World Bank containing the universe of projects funded by the Bank between 1998 and 2013— the time span of the APL (adaptable program loan) pilot. This dataset has information for 4,834 projects, including 429 adaptive (APLs) and 4,405 non-adaptive projects (non-APLs), and includes project identification information, size, objective, and other financial data for each project. Second, we obtain project evaluation reports from the World Bank’s Independent Evaluation Group. Each report, spanning 10 to 20 pages, provides a holistic evaluation of each project’s performance, context, achievements, and challenges. Each report has a summary of the project’s objectives, antecedents, and outcome evaluations with detailed explanations for project failure or success and lessons for future interventions. Third, we complement this information with conflict data from Uppsala University’s Conflict Data Program (Pettersson et al., 2021), a dataset with armed conflict information by country for each conflict recorded from 1948 to 2020. The Conflict Data Program is a global conflict data collection program that uses standardized methodologies to ensure comparability across countries and years. Finally, using data from the World Bank Development Indicators, we collect contextual data at the country level on political risk, government effectiveness, corruption, and other social, economic, and institutional characteristics.

Descriptive statistics. Tables 4 and A.1 (E-Companion) provide descriptive statistics for the variables of interest, while Table 2 supplies their definitions, units, project phase in which a variable is calculated, and data sources. Our sample, featuring a median performance rate of (4) “moderately satisfactory,” includes 245 APLs and 2,043 non-APLs in non-conflict areas (conflict = 0) and 184 APLs and 2,362 non-APLs in conflict areas (conflict > 0). Figure 2 shows that the average levels of political instability and conflict is similar between the two groups during the study period—the average conflict intensity for non-APLs is only 4% higher relative to the conflict experienced by APLs. Furthermore, Figure A.1 (E-Companion) shows that the distribution of APLs and non-APLs by levels of conflict intensity is comparable.

OPEN IN VIEWER

Figure 2.

Adaptive projects, conflict intensity, and political instability during the study period. Note. The figure shows the average political instability and conflict intensity for adaptive (APL) and non-adaptive projects (non-APL).

OPEN IN VIEWER

Table 4.

Sample variables descriptive statistics.

	All sample					Adaptive					Non-adaptive
	N	Mean	Std. dev.	Min	Max	N	Mean	Std. dev.	Min	Max	N	Mean	Std. dev.	Min	Max
adaptive	4834	0.09	0.28	0.00	1.00	429	1.00	0.00	1.00	1.00	4405	0.00	0.00	0.00	0.00
performance	4834	4.02	1.02	1.00	6.00	429	3.91	1.06	1.00	6.00	4405	4.04	1.01	1.00	6.00
experience	4834	4.34	1.15	0.00	6.36	429	4.24	1.14	0.00	6.05	4405	4.35	1.15	0.00	6.36
costOverruns	4816	11.50	7.86	0.00	23.97	429	14.46	5.87	0.00	22.45	4387	11.21	7.97	0.00	23.97
benefitsOverest	593	0.93	1.17	0.00	4.54	72	1.14	1.27	0.00	4.11	521	0.90	1.15	0.00	4.54
conflict	4834	0.26	0.46	0.00	2.00	429	0.25	0.41	0.00	1.57	4405	0.26	0.46	0.00	2.00
pastSuccessRate	4795	0.54	0.14	0.00	1.00	425	0.52	0.14	0.00	1.00	4370	0.54	0.14	0.00	1.00
conflictExp	4834	0.32	0.47	0.00	1.00	429	0.36	0.48	0.00	1.00	4405	0.32	0.47	0.00	1.00
designQual	4773	3.99	1.07	1.00	6.00	427	3.80	1.12	1.00	6.00	4346	4.01	1.06	1.00	6.00
scope	4834	4.36	0.89	1.00	5.00	429	4.48	0.80	2.00	5.00	4405	4.35	0.89	1.00	5.00
budget	4834	15.39	5.99	0.00	21.82	429	16.67	3.63	0.00	20.17	4405	15.26	6.16	0.00	21.82
duration	4350	6.79	2.55	2.00	17.00	429	7.20	1.73	3.00	14.00	3921	6.75	2.62	2.00	17.00
monitoringQual	3758	3.14	0.92	1.00	5.00	357	3.08	0.94	1.00	5.00	3401	3.15	0.92	1.00	5.00
noProjects	4834	5.40	3.90	1.00	20.00	429	4.97	3.14	1.00	16.00	4405	5.45	3.97	1.00	20.00
evalDur	4834	1.83	1.50	0.00	14.00	429	1.75	1.59	0.00	14.00	4405	1.84	1.49	0.00	14.00
GDPc	4459	11.49	2.57	5.56	18.85	380	11.29	2.53	6.07	18.84	4079	11.51	2.58	5.56	18.85
education	3183	3.78	1.55	0.00	15.99	289	3.90	1.78	0.00	12.50	2894	3.77	1.53	0.00	15.99
oilRents	4468	2.43	6.79	0.00	66.54	383	1.84	4.73	0.00	39.68	4085	2.49	6.95	0.00	66.54
agriEmpl	4456	42.76	21.34	0.64	91.76	375	43.01	23.66	0.74	91.17	4081	42.74	21.12	0.64	91.76
politicalRisk	3876	0.59	0.82	−1.35	3.31	310	0.45	0.78	−1.17	2.57	3566	0.60	0.82	−1.35	3.31
regulatoryQual	3870	−0.37	0.54	−2.49	1.54	307	−0.28	0.49	−1.51	1.42	3563	−0.37	0.54	−2.49	1.54
govEffectiveness	3868	−0.43	0.50	−2.16	1.48	307	−0.37	0.45	−1.63	1.10	3561	−0.44	0.51	−2.16	1.48
corruptionCountry	3878	−0.55	0.50	−1.70	1.54	307	−0.45	0.48	−1.53	1.41	3571	−0.56	0.51	−1.70	1.54

Note. The table provides descriptive statistics for the variables of interest and by adaptive (APL) and non-adaptive projects (non-APL).

Although the samples are not perfectly balanced across all the variables of interest, the differences between APLs and non-APLs are small and less than half a standard deviation, suggesting an unbiased composition of APL and non-APL projects. It is worth highlighting that the World Bank imposed no constraints on the eligibility of APLs; governments could have requested funding for both APLs and non-APLs projects. In fact, all countries in our sample that implemented adaptive projects also undertook non-adaptive ones during the study period.

We use a mixed-method approach to investigate our research questions. First, exploiting project heterogeneity in the implementation of APLs across conflict, development sectors, and countries, we run ordinary least squares (OLS) and ordinal models to analyze our research questions. Our sample includes all adaptive projects (429 projects) and non-adaptive projects (4,405 projects) implemented during the time frame of the APL pilot—a sample covering 92 countries, and development resources worth about 55 billion dollars.

Dependent variable. We use performance as the main dependent variable—that is, the extent to which the projects’ objectives are achieved, which follows a six-point scale, ranging from highly unsatisfactory to highly satisfactory. Although our main model reports OLS estimates, we also compute specifications using ordinal models (probit and logit) and obtain consistent results, as shown in Tables 5 and A.8 (E-Companion). In practice, ordinal and OLS models yield similar marginal effects when the predicted probabilities are not at the extremes of the distribution (Angrist and Pischke, 2009).

Independent variables. The main variables of interest are the indicator variable, adaptive, the continuous variable, conflict, and their interaction effects.

Adaptive is an indicator variable that identifies APLs (1) from non-APLs (0). We study the adaptability inherent in APLs, which comes from the possibility for managers to modify project characteristics during implementation based on lessons from earlier and ongoing phases (The World Bank, 2017).

Conflict is the average conflict intensity during project implementation. We estimate this metric following Uppsala University’s conflict definition—that is, “a contested incompatibility that concerns government and/or territory” that results in the use of armed force between two parties, where one of the parties is the government of a state (Gleditsch et al., 2002: p. 618). We estimate conflict intensity for each project by computing the average annual conflict level during its implementation. Specifically, each year, Uppsala University reports low or no conflict (conflict=0) when a country experiences battle-related deaths of less than 25; a conflict level of 1 when armed conflict results in at least 25 but less than 1,000 battle-related deaths; and high-conflict intensity with a conflict level of 2 in cases of more than 1,000 battle-related deaths.

Although conflict is aggregated at the country level, it is a relevant metric for our analysis because it affects project implementation. Given that armed conflict involves the government of a state, which is also the funding recipient and the party responsible for implementing projects, the government’s operations and administrative capacity will likely be disrupted during conflict. Additionally, the conflict variable is calculated as an average over the project’s implementation phase, making it a reasonable proxy for the conflict environment observed throughout the project’s lifecycle.

Controlling for alternative explanations. We control for alternative factors explaining variations in performance by running alternative specifications, including a large set of covariates and fixed effects, and employing a comparable control group. The models include three types of “controls.” The first group controls for project-level characteristics affecting project performance including past success rates in the focal country (passSuccessRate), whether a country had conflict experience before the start date of the focal project (conflictExp), the quality of the project design (designQual), the scope, project size (budget), duration, project monitoring quality (monitoringQual), the number of concurrent projects in the focal country (noProjects), and project evaluation duration (evalDur), which controls for potential differences in how project performance is measured.

The second set of covariates includes contextual factors (context effects); that is, economic, institutional, governance structure, and social variables that control for country-level traits that potentially influence project performance. Specifically, we control for GDP per capita (GDPc), government expenditures on education (education), oil rents (% gross domestic product), employment in agriculture (% of total employment), the level of political risk (politicalRisk), the quality of the regulatory environment (regulatoryQual), the overall effectiveness of the host country in providing public services and its independence, credibility, and commitment (govEffectiveness), and the levels of corruption (corruptionCountry).

The third group of controls includes fixed and random effects, with effects for the project, year, region, financial conditions of the loan, sector, and “project evaluation.” Specifically, project effects avoid that other project-level factors not controlled for could drive the results. We estimate random effects because fixed-effects models in ordered models can lead to bias (Greene and Hensher, 2010). Year effects control for time events such as policy changes that may confound the results. Region effects includes fixed effects for countries in one of nine geographical areas, ranging from East Asia and the Pacific to Sub-Saharan Africa, and control for geographical clusters or spillover effects. Finance effects are indicator variables denoting a project’s funding mechanism and are included to control for project differences that influence performance due to the project’s financial characteristics. For example, a project can be funded using resources from a development funding program, a rehabilitation loan, or a program-for-results account, among other product lines. Sector effects control for sector-specific effects, which identify the main development sector to which a project belongs, from agriculture to urban development sectors—a total of 25 sectors. Finally, we control for project evaluation effects, which, by controlling for the project evaluation reporting methodology, account for potential differences in how projects were evaluated.

Our main model’s identification strategy is represented in Equation 1, where performance i denotes the performance of project i , conflict i represents the conflict intensity, adaptive i indicates whether the project is adaptive (1) or non-adaptive (0), X i is the vector of control variables, and ϵ i represents the error term. To ensure robustness, we provide specifications employing robust errors at the project level and errors clustered at the country level.

performance i = β 0 + β 1 conflict i + β 2 adaptive i + β 3 (conflict i × adaptive i ) + Δ X i ⏟ controls + ϵ i .

(1)

Qualitative analysis. To enrich our quantitative analysis, we analyze project evaluation reports. Each report provides a more holistic assessment of the project, its context, achievements, and challenges, providing context and depth to our statistical results. Specifically, using principles of cross-case synthesis (see Yin, 2017), we provide anecdotal evidence on how conflict and adaptability influence project performance and reasons for project failure—we analyze 128 adaptive projects rated as moderately unsatisfactory, unsatisfactory, and highly unsatisfactory. Table A.3 (E-Companion) describes the projects cited in the paper, referenced in the text with their “project ID” in parentheses next to the relevant finding, example, or quote.

5.1. The Conflict Effect

We start by examining the direct impact of armed conflict on project performance (relationship R1 in Figure 1) in the baseline scenario without adaptability. Consistent with expectations, we find that conflict significantly lowers project performance, as shown by the negative and statistically significant coefficient ( p < 0.05 ) of the variable conflict (-0.11)—this effect is estimated as E [ performance | adaptive = 0, conflict > 0] − E [ performance | adaptive = 0, conflict = 0]. The negative coefficient suggests that, without the adaptive feature, a higher conflict intensity will decrease project performance by up to 0.22 points on the rating scale, as conflict intensity can increase by up to two units. Although this represents a modest effect on performance, for non-adaptive projects, this effect is the third largest across all independent variables, after project design quality (designQual) and monitoring and evaluation quality (monitoringQual).

Project evaluation reports show that conflict can affect project performance at the project, government, and Bank levels. At the project level, armed conflict poses significant challenges, including logistical obstacles, interruptions of project activities, and negative effects on project personnel (P103189). At the government level, conflict can divert resources from project implementation to other priorities. This is a substantial challenge because the government is the recipient, beneficiary, and administrator of the Bank’s funds. For example, in a public sector reform project in Bolivia (P062790), “the government was largely immobilized by social and political conflict during much of the project implementation period and was unable to carry out the implementation[ ⋯ ]”. In more severe cases, conflict can trigger a full halt of government activities and disrupt the operations of the implementing agencies. In project P101206, for example, social unrest led to the resignation of the sitting president and to significant disruptions and project delays. At the Bank level, however, the concern is not only with project outcomes, but also with accountability. Funding can cease, leading to input shortages and project interruptions (P101206), if there are significant doubts about the government’s commitment to sound fiduciary management, as in the case of de facto governments (World Bank’s operational policy 7.30). Nevertheless, once a project is approved, the Bank remains committed to implementation and disbursements, provided a proper legal framework remains in place, and borrowers can fulfill their fiduciary obligations.

5.2. The Adaptability Effect

We now explore the effects of adaptability (relationship R2 in Figure 1). We find that adaptive projects outperform non-adaptive ones when implemented in conflict areas. Specifically, adaptive projects perform 0.29 points better when implemented in conflict areas, relative to non-adaptive ones, as evidenced by the positive and statistically significant coefficient of the interaction term, adaptive × conflict in Column VIII, Table 5. Moreover, adaptive projects in conflict areas see a net increase in performance of 0.18 points relative to adaptive projects implemented in non-conflict areas—this effect is estimated as E [ performance | adaptive = 1, conflict > 0] − E [ performance | adaptive = 1, conflict = 0]. The evidence suggests, therefore, that adaptive projects are better suited to thrive in armed conflict conditions, while non-adaptive projects are better suited to safer and more stable contexts.

5.3. The Adaptability Challenge

The main model in Table 5 shows that APLs implemented in low or no conflict areas (conflict = 0 ) exhibit lower performance relative to non-APLs—that is, E [ performance | adaptive = 1, conflict = 0] − E [ performance | adaptive = 0, conflict = 0] < 0 . Evaluation reports of adaptive projects with low-performance ratings, including projects with moderately unsatisfactory, unsatisfactory, and highly unsatisfactory ratings (128 projects), indicate design issues during their planning phases (the complete qualitative analysis is presented in Section 6).

These issues fall into three categories. First, inadequate risk mitigation: many projects failed to develop comprehensive risk mitigation plans and prepare for challenges posed by the political environment and project complexity, even when risks were identified during project formulation. Second, underestimation of project costs and complexity: project leaders often set overly optimistic assumptions about government and implementing agency capacity, project complexity, and the likelihood of successful collaboration among partners and donors. And third, over-optimistic benefit projections: projects covered too many development areas and made unrealistic assumptions about the chances of achieving desired outcomes.

These design deficiencies suggest that project design quality may mediate the negative relationship between adaptability and performance in areas with no conflict. To test this mechanism, we conduct an illustrative mediation analysis following Baron and Kenny (1986), examining how project design quality (designQual) explains performance. The results, shown in Table 7, suggest a strong mediation effect of designQual (M) when restricting the sample to low or non-conflict areas (conflict = 0 ). Column I reports a negative direct effect of adaptability (X) on project performance (Y); Column II shows that adaptability negatively influences designQual; Column III indicates a positive and strong effect of design quality on project performance; and Column IV, providing the combined effects of X and M on Y, shows that the negative and statistically significant effect of adaptability from Column I turns statistically insignificant when including the variable designQual in Column IV. Why, then, do adaptive projects underperform in non-conflict areas? The results suggest that adaptive projects have weaker designs than non-adaptive ones.

6.1. Common Failure Modes in APLs

Using evaluation reports from 128 adaptive projects rated as moderately unsatisfactory, unsatisfactory, or highly unsatisfactory (from a total of 429), we investigate why these projects failed to achieve expected results and the contexts in which they failed. We find recurring themes associated with insufficient risk mitigation, underestimation of costs and complexity, and over-optimistic benefit projections.

1. Inadequate risk mitigation. A recurring issue in project evaluations was the failure to develop robust, comprehensive risk mitigation plans. Many projects failed to anticipate and prepare for challenges posed by the political environment and project complexity (e.g., P105711, P082651), the institutional capacity of implementing agencies (P096058), and the complexities of donor coordination (P094917). This lack of foresight often led to “unrealistic” and inattentive project design as assessed by project evaluators (P082516). Projects failed to account for realities on the ground and conduct comprehensive risk assessments. For instance, a project evaluation report notes how “the risk assessment was not sufficiently critical [ ⋯ ] risks were underestimated [ ⋯ ] [and] the assessment was not comprehensive enough.” In Yemen (P064981), a project aiming to increase water resources in the Sana’a Basin shows how poor risk management may have been caused by overestimating “the political commitment of the Government” and the institutional capacity to undertake the project.

Even when risks were identified during project formulation, many projects failed to develop adequate mitigation measures. The shortcomings often stemmed from overly optimistic assumptions about the capacity of government and implementing agencies (e.g., P104995, P073689), project complexity and technical risks (e.g., P064981, P057394), and the likelihood of successful collaboration among partners (e.g., P082516, P096058) and donors (P094917). These assessment flaws frequently resulted in projects that did not adequately prepare for known risks, regulatory changes, political instability, or coordination challenges (e.g., P096058, P072080, P073689), especially in complex projects like this one implemented in Brazil:

Analysis underestimated risks and did not provide specific mitigation measures for known risks (e.g. political), and did not consider the high risk of geographic coverage (as it had too many parallel sub-project activities operating in different areas with little overlap) (P082651).

Thus, while risks in low or non-conflict areas were often well understood, adaptive projects frequently failed to develop comprehensive mitigation strategies. This evidence suggests that the flexibility inherent in adaptive approaches can lead to less rigorous risk mitigation strategies.

2. Underestimation of costs and complexity. A recurring problem found in project evaluations was the systematic underestimation of costs and project complexity. These issues were often rooted in poor project design (P089013), leading to activity cancellations (P059872), fund reallocations (P089013), and the rescheduling and reprioritization of project activities. In Djibouti (P086994), a project aimed at enhancing educational infrastructure exemplifies this pattern: “more money was needed to cover the underestimated construction costs of the middle schools, and so project activities such as improving the distribution of textbooks, providing on-site teacher training and evaluation [ ⋯ ] did not occur,” forcing managers to sacrifice crucial components and reassign scarce funds. Similarly, in Zambia, evaluators found that “costs [of a transport intervention] were underestimated,” leading to “dropping of some activities from this project” (P106596).

3. Overestimation of benefits. Adaptive projects exhibited a pattern of setting overly ambitious targets and objectives (e.g., P079633, P095240, P069491), covering too many development areas (P073526), creating overly optimistic reporting (P071985), and making unrealistic assumptions about the capacity of implementing agencies (P082452) and the likelihood of achieving desired outcomes (P072482). Actual benefits in many cases fell short of initial expectations. Moreover, many project managers, relying on the flexibility of the adaptive approach, underestimated the challenges of operating in post-conflict environments with weak institutional capacity (P082516), underestimated task complexity (P105711), or were overconfident in implementing novel projects (P117102). For example, a project in Sierra Leone (P073883) improving living conditions of rural communities shows how “there was a substantial gap between the theory behind the project design and practical considerations on the ground.” Similarly, in Tunisia (P064082), a project evaluator notes that “it is critical to set up a realistic and measurable objective in accordance with a practical and agreed time frame.”

6.2. Robustness Tests

7.1. Managerial Insights

When is adaptability recommended? Although adaptability facilitates agility and course corrections (The World Bank, 2017), continuous adjustments can be costly. Our study reveals that adaptability’s value is context-dependent. Adaptability offers the greatest value in areas of high political instability and armed conflict, where it enables critical adjustments to new realities. When instability is low, however, costs may outweigh potential benefits. Unlike other sectors, prosocial projects face heavy bureaucracy and complex stakeholder networks, making project adjustments overwhelming. We find that adaptive projects underperform non-adaptive ones in non-conflict areas. Therefore, in non-conflict areas, adopting adaptive management requires careful evaluation of benefits and costs and robust project design.

Aim for focused projects. Adaptability enables implementers to adjust project attributes as new information emerges; however, its positive effect diminishes when projects address too many development areas. As the number of agencies involved and complexity increase, prosocial projects require more administrative capacity and coordination to identify, approve, and execute project changes. Multilateral organizations, therefore, should prioritize focused interventions when funding adaptive projects.

Design and monitoring strategy matters. Project design quality partly explains the lower performance ratings of adaptive projects in non-conflict areas. Our results stress the importance of carefully designing and monitoring adaptive projects. Adaptability is a resource to address unforeseeable uncertainty. Known risks, however, should be identified early and mitigated through rigorous planning and efficient monitoring.

Enhance knowledge spillovers. We find that experience boosts the advantages of the adaptive approach. Relevant experiences provide critical knowledge to help managers implement project changes and revisions. To promote knowledge spillovers, we suggest tracking experiences early in the project, meticulously documenting progress, and sharing the lessons with peers and the project network.

Curb the enthusiasm. We provide evidence of planning fallacy bias in adaptive projects—a tendency to systematically overestimate project benefits and underestimate the resources needed for completion. This bias is exacerbated by adaptive approaches and can weaken project designs due to overconfidence and under-preparation. We recommend that both recipients and providers of multilateral funds be aware of this bias during project planning. Adaptive projects would benefit from doubling down on the identification of constraints and risks.

Adaptability in New Product Development (NPD) projects. Although prosocial projects are less definite than traditional NPD projects, both deliver customized, often unique solutions in contexts with high levels of uncertainty. Our results show that adaptability is particularly valuable in contexts of high political instability and armed conflict, and potentially less beneficial when political instability and armed conflict are low. Therefore, organizations executing NPD projects in volatile and highly uncertain contexts should consider implementing an adaptive approach.

7.2. Theoretical Contributions

Adaptability helps improve project performance in volatile environments, such as armed conflicts. Our findings contribute to nonprofit operations management by showing that adaptability yields the most value to projects facing high uncertainty but designed with focused scopes. We argue that adaptability helps managers address projects’ unknown–unknowns during implementation. However, adaptability can exacerbate coordination and administration issues when project scope is wide and covers too many development areas.

We also contribute to project management by revealing an unexpected problem of adaptability: planning fallacy bias. Although this bias has been documented in public works and business projects (Lorko et al., 2021), our research provides novel evidence of its presence in prosocial projects and multilateral operations. We argue that adaptability can inadvertently exacerbate this bias, as governments face funding needs and political pressure. This is an argument that aligns with existing theories suggesting that overoptimism bias and strategic misrepresentation are tied to political and organizational pressures (Flyvbjerg, 2008).

7.3. Limitations and Future Research

Prosocial project management in armed conflict contexts is an underdeveloped but promising line of work. Much remains to be done to help the international community understand how to use adaptability to safely and effectively scale up prosocial interventions. Thus, given the limitations of our study, we encourage further research on this topic. First, we investigated the World Bank’s Adaptable Program Loan, a program that was discontinued and later transformed into the MPA—a new approach featuring more flexibility, streamlined processes, and simplified documentation (The World Bank, 2017). We encourage researchers to evaluate the results of MPAs upon completion of evaluations of the first cohort of projects. Second, we studied the trade-offs of adaptability at the project level; however, more research is needed to understand its effects at the team and individual levels. Third, we evaluated adaptability through the lens of the APL pilot. However, adaptability in the prosocial sector might not look the same in other sectors. We advise caution when applying these results outside the prosocial sector. Finally, we analyzed adaptability under varying degrees of political instability and state-based conflicts. Future work should investigate adaptability in other conflict scenarios and under other types of uncertainty, including non-state violence and organized crime.