Dreams vs Reality: Urban and Rural Female Youth Aspirations Abstract:Youth’s aspirations are a key influence on their decision-making process but may perpetuate poverty traps among marginalised and socioeconomically disadvantaged groups. We aim to examine the complex formation of female youth aspirations and geographical demographic differences. We recruited 56 16-year-old Malaysian female students from one urban (n = 24) and two rural schools in Malaysia (n = 32). A mixed-design approach was utilised by (1) quantitatively assessing their aspirations from drawings of future selves and (2) qualitatively complementing the drawings with semi-structured interviews (n = 28, 14/14). Results showed that youth aspirations are socially produced, primarily by family members but also by internal factors (e.g., domain passion and self-efficacy). While there were similar rural-urban aspiration levels, rural youth demonstrated more self-blaming tendencies (resulting in them choosing careers of lower occupational prestige) and frequently mentioned resorting to part-time entrepreneurship to supplement employment their income. Our study has implications for interventions targeted at driving social mobility for rural and/or low-income young women and emphasises the importance of a multifaceted capacity-building approach while also addressing structural constraints.
Social expectations and government incentives in Malaysia's COVID-19 vaccine uptake Abstract:High vaccination rates are integral to reducing infection and severity rates of COVID-19 infections within a community. We examine the role of social expectations in COVID-19 vaccination take-ups and its interaction with potential government actions in Malaysia. We find that individuals’ expectations of others in their social groups towards vaccination predicts those individuals’ vaccination registrations. Using a vignette experiment, we examine the extent of normative expectations in normalizing pro-vaccination behavior beyond an individual’s reference group. We find that unless moderated by a high level of public trust, individuals prefer punitive policies as a way to increase vaccination rates in their communities.
Identity, trust, and the experiences of refugees during a COVID-19 lockdown Abstract:This paper examines the experiences of refugees in a developing country during its first COVID-19 lockdown by utilizing a two-stage qualitative data analysis of 39 interviews with refugees and asylum-seekers. We find that their experiences during the lockdown are shaped by identity, trauma and help from external parties–such as community leaders and local non-governmental organizations (NGOs). Experiences during the pandemic in turn moderate the relationship between policy changes and trust in domestic authority figures, which consequently affects attitudes towards and compliance with public health measures put in place to contain the pandemic. We then explore the role of identity in refugees’ pandemic experiences by comparing the differences between two refugee groups (Syrians and Rohingyas), validating them by utilizing comparative thematic analysis. Finally, the paper presents policy implications for crisis response in developing countries by suggesting improvements that can be made on the ground regarding the delivery of aid and assistance to vulnerable groups.
Perceptions of Entrepreneurship and Online Learning During the Coronavirus-2019 (COVID-19) Pandemic Abstract:Coronavirus-2019 (COVID-19) restrictions significantly influenced the learning and delivery of educational programs, especially traditionally hands-on educational programs. Entrepreneurship education and training (EET) studies on learners’ perceptions have so far focused on formal EET in university settings or Massive Open Online Courses (MOOCs). This paper explores youth perceptions of a non-formal, online EET program conducted during the pandemic. Perceptions matter since they tend to translate into attitudes, which in turn potentially translate into achieving learning outcomes (or not). Using hermeneutic content analysis (HCA), transcripts from 35 youth participants were analyzed, where the participants were categorized into four groups based on completion of the program and household income. Individual motivations were very important for all and the lack of social support was a concern for low-income youth. Almost a third of the dropouts who were part of the study did so to actually start their own business during the pandemic versus only one out of 18 of non-dropouts. The pandemic was disruptive to livelihoods and to their families, which sometimes made learning more difficult. Future online EET programs should relate learners’ motivations for learning entrepreneurship with learning outcomes by instilling social support structures and taking contextual influences into consideration.
Translating entrepreneurial intention to behaviour amongst micro and small entrepreneurs Abstract:Purpose Entrepreneurship studies have established various antecedents leading to eventual entrepreneurship by measuring entrepreneurial intention (EI). However, evidence has shown that intention does not necessarily translate into behaviour, especially for complex behaviours such as creating a business venture. Hence, this paper aims to examine how contextual and individual factors interact with one another to promote or inhibit one’s translation of EI into entrepreneurial action in an emerging economy. Design/methodology/approach The authors adopt a retrospective qualitative approach by interviewing 37 Malaysian micro and small business owners. Then, multidimensional scaling is used to examine the interactions between the identified factors. Findings The authors find that social networks are the main influence on an individual’s propensity to start a business – it provides financial and social capital, provides other means of support such as practical help and business opportunities and instils passion and drive. Furthermore, organisations such as schools, universities and employers play an important role in instilling the motivation for a career shift to entrepreneurship and by providing opportunities to upskill. In addition, the findings indicate that entrepreneurial traits such as proactiveness, resourcefulness and passion enable individuals to overcome entrepreneurial structural constraints, such as lack of resources and negative action-related emotions. By contrast, the role of macro-environmental factors such as governmental support play less prominent roles in the narratives of the entrepreneurs. Practical implications This study has important implications for governments and policymakers in implementing support for those transitioning from salaried employment to self-employment and for entrepreneurship interventions to adopt a holistic approach that encompasses building one’s entrepreneurial knowledge, skills and mindsets, alongside providing external incentives. Originality/value The authors provide a more holistic approach to exploring the EI–behaviour gap. In addition, this study explored facilitators and barriers to entrepreneurship specific to the context of an emerging economy such as Malaysia, which is highly dependent on small-scale self-employment.
A textual analysis of the effect of short-term volunteering on attitudes toward refugees in Malaysia Abstract:Does exposure to an unfamiliar migrant community change implicit biases? We conducted an experimental study by exposing Malaysians to a few hours of volunteering with Rohingya refugees, and we examined the effect of this treatment on their attitudes through a textual analysis. We measured changes in attitude through pure valence and multiple measures of implicit bias, including linguistic intergroup bias. We found that the volunteers became markedly more positive, and this change was statistically significant. Our results suggest that brief exposure to refugees may be a cost-effective policy lever for changing local perceptions of refugees.
The formation of youth entrepreneurial intention in an emerging economy: the interaction between psychological traits and socioeconomic factors Abstract:Youth entrepreneurship has been identified as a key driver in overcoming the economic crisis spurred by youth unemployment. However, the understanding of youth entrepreneurship is largely based on research in high-income countries. Furthermore, entrepreneurship studies to date are largely limited to the independent effects of individual traits on entrepreneurial intention (EI). Hence, this study aims to model the cognitive and social conditions, mediating processes and interactions to understand how youth EI can be formed and strengthened in an emerging economy.
Poverty and precarious employment: the case of Rohingya refugee construction workers in Peninsular Malaysia Abstract:How do refugees economically cope in host countries where they have no legal right to work? Most Southeast Asian countries have not ratified the 1951 United Nations Refugee Convention. This implies that many refugees in this region do not enjoy any protection from the law, do not possess the legal right to work, and often resort to participation in dirty, dangerous, and demeaning jobs in order to make ends meet. In this paper, we study Rohingya refugees working as construction workers in the informal economy in Peninsular Malaysia by employing a mixed methods methodology. Specifically, we utilize micro-level survey data collected from a representative population of 314 Rohingya refugee workers in the construction industry, as well as in-depth interviews conducted with a subset of 77 of the survey respondents. Using the survey data, we first provide an overview of the social and economic lives of our respondents by summarizing key variables such as demographics, integration measures, and healthcare access. We then utilize regression analysis to understand the relationships between these variables. The key quantitative finding is that Rohingya refugees in the construction industry earn significantly above minimum wage in Malaysia (albeit less than their legal counterparts), and significantly more than their earnings prior to arriving in Malaysia. This is true even after adjusting for purchasing power. We then conduct a thematic analysis on the qualitative data obtained through the interviews to understand the dimensions of employment for the respondents. We find that although the construction industry in Peninsular Malaysia has provided Rohingya refugees with the means to escape poverty, they still face a tremendous amount of precarity and uncertainty in their lives.
Issues Facing Refugees and Asylum-Seekers in Southeast Asia: Narrowing the Gaps Between Theory, Policy, and Reality. Abstract:Practitioners’ experiences and perspectives on social interventions with refugees are underexplored in Southeast Asia. This gap limits the ability to create impactful public policy in the region. In this report, we present findings from an interdisciplinary research workshop held in Kuala Lumpur in 2018. The workshop included sixty individuals from a diverse range of backgrounds – asylum-seekers, refugees, academics, NGO leaders and staffs, representatives from United Nation agencies, and government officials. Using thematic analysis, we extracted some issues considered to be the most pressing for refugees, as well as issues considered important yet understudied. Based on these workshop outcomes, we suggest a research process flowchart to aid researchers and practitioners in maximizing their impact through policy and advocacy, while at the same time partnering with refugee communities to better serve their needs.
Using classroom games to teach core concepts in market design, matching theory, and platform theory Abstract:Market design uses various economic tools, such as game theory and experimental analysis, to aid in the design, implementation, and the fixing of broken markets whenever needed. One application of market design is to study two-sided matching markets, such as marriage and labor markets. Market design and matching theory also relate to the study of platforms (i.e., profit-maximizing intermediaries who serve to connect two or more “sides” of people) such as Amazon, Netflix, and Airbnb. In this paper, we explore some of the core concepts in understanding market design, matching theory, and platforms, and we outline three classroom games with detailed instructions for instructors who may want to explore these topics in their own classes and curricula. The first game studies market thickness, market segmentation, and their use by platforms. The second game explores the ideas of preferences, differentiation, and search frictions to explore how positive assortative mating has had negative consequences for income inequality in the U.S. in recent history. The third game studies the function of algorithms in the design and operations of a platform and presents a game that teaches the Gale-Shapley algorithm, together with a supplementary handout for classroom activities, to further strengthen students’ understanding of this seminal algorithm. The games are flexible and can be implemented for a variety of levels of study, including an undergraduate introductory economics class, an upper-level undergraduate elective, or an MBA economics course covering topics relating to platform design and management.
Pricing schemes and market efficiency in private retirement systems Abstract:We study the effects of different pricing schemes on the overall surplus in a privately managed retirement system with multiple service providers and switching costs. We develop a theoretical model based on the Chilean retirement system and consider a repeated auction for monopoly rights over new enrollees. We consider a dynamic model solved by pension fund administrators and by consumers. We compare three different pricing schemes: (a) fees on contributions, (b) fees on returns, and (c) a two‐part tariff including an auction over a guaranteed rate of return and allowing the firm to keep a portion of returns generated above this guaranteed rate. We also consider heterogeneity across individuals where agents earn high or low wages and high‐wage customers have proportionally lower switching costs due to more cost‐effective access to financial planning services. We find that auction participants subsidize consumers. We also treat savings as a durable good. In this case, pricing over returns worsens the switching related inefficiencies just described relative to pricing over contributions, despite the better incentives it provides. These inefficiencies can be resolved by allowing firms to price discriminate.
Policy Papers
Norms, Trust, and Increasing Vaccine Registration in Malaysia Abstract:This report summarizes preliminary findings from a survey experiment with 1307 Malaysian respondents. We study 1) trust levels in institutions and norms surrounding the choice to get vaccinated in this country, 2) perceptions on state “carrots” versus “sticks” in incentivizing vaccinations, and 3) whether a cash lottery of RM 500 is an effective intervention to increase the number of registrations for vaccinations. Although our sample is not representative of the entire Malaysian society, they are suggestive and can provide some insights on incentivizing pro-social behaviour with regards to vaccination. Specifically, we found that incentives can encourage vaccine registration. When the cash lottery of RM 500 was offered to the treatment group in our small randomized controlled trial, almost all individuals in this group registered to get the vaccine (83.7%) compared to the control group (62.16%). This difference was statistically significant at p=0.1. Furthermore, through a randomized vignette analysis where respondents are asked to predict the behavior of someone like them in a hypothetical scenario, we found that a punishment of having to self- pay for PCR COVID-19 tests every two weeks was more effective at promoting vaccinations compared to an incentive of being enrolled in a lottery with one large cash prize of RM 1 million.
The COVID-19 Hardship Survey: An Evaluation of the Prihatin Rakyat Economic Stimulus Package Abstract:The COVID-19 pandemic has caused a global crisis, and, while Malaysia has controlled the infection more than most countries with early exposure to the virus, the Movement Control Order (MCO) has required major economic sacrifices. This, in addition to a COVID-19-caused global economic slowdown, threatens a budget crisis for Malaysian households. Malaysian households often carry a great deal of debt and have little savings; without a source of income, many households could quickly lose access to necessities like food or housing. In response, the Malaysian government has created a series of stimulus packages. This report uses a survey of Malaysian household income and expenditures conducted from the 20th to the 27th of March to analyze the effect of these stimulus packages on the cash-flow and solvency of Malaysian households. We find that they are likely to address most of the cash-flow issues brought on by the COVID-19 crisis, at least in the short term. While a substantial minority of M40 respondents and a nearly half of B40 respondents reported negative cash-flow due to the crisis, the stimulus policies are able to decrease negative cash-flow rates among our respondents to levels at or below those that persisted before the COVID-19 crisis for the month of April, assuming income and expenditures from March persist. Among the minority who still have negative cash-flow, most have enough savings to survive for more than three months. However, a small minority of respondents in our survey are still likely to run out of money in the next few months, especially among the B40. If policymakers wish to further protect households from budget crises after the one-time transfers of April, our analysis shows that transfers to lower-income households—families making RM 4,000 or less and single individuals making RM 2,000 or less—are dramatically more effective than transfers to higher income households. There are several caveats to our conclusions: our sample is not representative, and our cash-flow estimates are only approximate, so our results should only be viewed as suggestive. Additionally, our responses come from the beginning of the MCO, so they likely undercount the number of Malaysians who have lost their jobs or have decreased income. Finally, this analysis assumes benefits can be easily and quickly disbursed to distressed households. Elderly individuals or those in rural areas may not have the internet access or expertise to secure their benefits, and delays in disbursement could be catastrophic for households with no savings.
A COMPREHENSIVE STUDY OF ROHINGYA CONSTRUCTION WORKERS IN PENINSULAR MALAYSIA AND RECOMMENDATIONS FOR A FUTURE WORK PILOT PROGRAM Abstract:The is a comprehensive report based on qualitative and quantitative data from 288 Rohingya construction workers in Peninsular Malaysia, collected from May to August 2018. This report covers all aspects of refugee life and work in the construction industry – this includes data and ethnography on the demographics, language, family structure, and finances of these workers. It also describes the nature of construction work in Malaysia, and how the Rohingya work in this field. This includes career progression and specialization, income, hours of work, and job amenities. We focus on construction as it is the most common profession for Rohingya workers.
Abstract: I study a search model of online dating with nontransferable utility where agents are vertically differentiated, self-report quality, and must go on costly dates to verify a match’s quality. I show that these per-date costs induce some agents to over-report their type, consistent with the stylized facts of online dating platforms where users frequently over-report characteristics like height and income, a phenomenon known as catfishing. This make agents less picky by preventing high types from rejecting some low types, and since externalities in matching markets without transfers can make agents inefficiently picky, these costs can improve total market surplus. A monopolist platform owner may also have an incentive to increase per date costs in order to increase profits. Thus, inducing lying amongst users can actually be optimal for a platform.
Mirror Matches: Analyzing Matching Markets with N-Dimensional Preferences Abstract: This paper analyzes matching markets where agent types are points in R^n and agents prefer matches that are closer to them according to a distance metric over this space (horizontal preferences). Given a few additional assumptions (most notably a form of symmetry between the two sides of the market), we show that in the Gale-Shapley stable matching in this environment agents match to a linear function of their own type. We show this result holds both when the division of match surplus is fixed (nontransferable utility), and when agents can bargain over match surplus (transferable utility) given a convexity assumption on the utility function. We also show that the restriction to horizontal preferences is not as onerous as it may seem, as a rich variety of preference structures can be mapped into the horizontal framework, including vertical preferences where all agents agree on the preference ordering of a trait. This result can be interpreted as a generalization of Becker’s univariate assortative matching to multiple preference dimensions and more general preference structures.
Work in Progress:
AI Thematic Analysis: Methods and Challenges Description: Using GPT 3.5-Turbo, We develop a method to automate a version of the qualitative method of thematic analysis with a specific focus on coding passages for thematically relevant content. This method segments the textual dataset or corpus into paragraphs, assigning a code to each that identifies a theme relevant to the research question, if applicable. Using a team of expert human reviewers, we assess the quality of the output and find that it often, though not always, captures the source text's relevant themes, marking this as an important tool, especially as models improve. Our approach combines careful prompting, provision of contextual information about the document the passage came from, as well as more carefully analyzing a small subset of the passage, and, leveraging text embedding to identify similar passages, using similar members of the carefully coded subset as examples for few-shot prompting for each passage to be coded.
AI Thematic Analysis of Attiudes towards Refugees in Malaysian News Media Description: This project employs innovative AI thematic analysis methods (outlined in the paper above) to examine the evolution of attitudes towards refugees in Malaysian news media since 2017. Unlike traditional textual analysis, we are able to track narratives and discourses rather than one-dimensional sentiment or word frequencies, and unlike traditional thematic analysis, we are able to scale up our analysis to generate a large dataset amenable to statistical analysis. While this research is incomplete, we have found several interesting descriptive results:
During the COVID-19 crisis, discussion of refugee access to education—usually a major topic of discussion—was almost completely crowded out by discussion of COVID-19 related refugee issues.
The Malaysian government’s focus on Myanmar’s culpability in Malaysian refugee issues is highly periodic, primarily occurring in autumn.
Discourse around discrimination against the Rohingya has, over the past 6 years, shifted from a primary focus on individual citizens to the Malaysian government as the locus of discrimination.
AI Corpus Selection for Thematic Analysis Description: There are limitations to the AI coding described above and many researchers prefer human coding. This project focuses on using AI not to provide final codes for thematic analysis, but to select a highly productive corpus when the universe of possible documents to code is too large to be exhaustively human coded. Using a dataset of over 2,000 news articles on refugees and AI generated thematic codes for each article, we select a much more tractable corpus for human readers--less than 40 articles--that maximizes the number of unique AI codes among all subsets of the 2,000 articles. We then task expert human coders to code this corpus, as well as randomly selected articles for comparison. We find high correlation (0.82) between humand and AI code density (unique codes/1000 characters of text) and a far higher code density with the AI selected corpus the randomly selected articles, showing that this method can greatly speed up the process of achieving saturation in thematic analysis.
AI and Humans as Interviewers in the Social Sciences: an Experimental Approach. Description: In this project we build an GPT 3.5-Turbo based chatbot to conduct semi-structured interviews--interviews with a predetermined set of primary questions, but, unlike conventional surveys, flexibility to follow up on interesting topics the interviewee surfaces. This chatbot uses GPT as a foundation, along with sophisticated programmatic guardrails and background reflection functionality to ensure the chatbot fulfills their role as an active interviewer rather than a passive assistant. We conduct an experiment where interviewees are randomly assigned to be interviewed by a human or the AI via text messaging. Humans can recognize the AI as an AI, but also often believe the human interviewers are AI. Interviewer performance is assessed by expert reviewers and the interviewees themselves. Both methods yield slightly higher average ratings for the human interviewers, but with heavy overlap between the human and AI interviewers. We conclude the AI interviewer's performance is broadly comparable with a human expert, and AI-based chatbots are a viable method to scale up semi-structured interviewing without the massive headcount that usually requires.
Economic Geography of the Malaysian Roadside Economy Description: This project utilizes a novel computer vision pipeline to identify “hawkers” or roadside vendors in Google Street View (GSV) images using a geospatial dataset covering the entirety of Peninsular Malaysia and several hundred thousand panoramas over the years 2013-2021. Many prior papers using GSV images—such as the well-known (Gebru, Krause, and Fei-Fei, 2017)—focus on detecting standardized objects (models of cars in their case). Hawker stalls, varying from temporary to permanent structures often constructed ad-hoc, present unique challenges for object detection. To minimize measurement error we detect hawkers using the most advanced YOLO-family object detector, calculate the distance to the object via monocular (single image) depth estimation using MiDaS, and use the detections, their headings, distances, and that latitude and longitude of the GSV panoramas to attempt to identify each hawker in every adjacent panorama, allowing us to ensemble multiple detections from multiple angles for each hawker and improve detection accuracy. While this research is still in process, we have found a huge response to the Covid-19 pandemic in hawker frequency, but only among low capital intensity hawking modalities. Along with coauthors in Urban Studies and Political Science, we hope to explore how hawkers engage with the built environment and how municipal level political affiliation may influence enforcement against these often-informal vendors.
Matching with Single Peaked Preferences Abstract: This paper studies two-sided one-to-one matching in a frictionless nontransferable utility model where agents are characterized by a univariate type and have single-peaked preferences characterized by an ideal type (decoupled from own type) and greater preference for matches closer to that type. Given some modest distributional assumptions, we recover a closed form for the matching function. We also develop a generalization of the Eeckhout condition for uniqueness and show it applies to our model. Finally, we apply our results to a simple model of pre-market educational investments and marriage matching.
Quality Versus Fit: Market Design and Externalities on Multidimensional Matching Platforms Abstract: This paper studies externalities in one-to-one matching markets when agents have preferences over multidimensional types by utilizing a minimal search setting where agents are either high or low quality and have an idiosyncratic per-match ``fit" shock. Applications include dating and marriage, job search, and school choice. It identifies a novel source of externalities that does not exist in the one-dimensional models focused on in the previous literature, but is endemic in multidimensional settings, appearing in both search and frictionless matching models so long as nontransferabilities are present. Agents match too aggressively on traits where preferences are homogeneous across agents (quality), and too little on traits where preferences are heterogeneous across agents (fit). This effect is decomposed into an intermatch externality -- when you match to someone, you impose a cost on the rest of the market by removing them from it, and an intramatch externality -- you don't account for your partner's payoffs when choosing partners. Given these generic externalities, we provide a survey of instruments a matching platform could use to improve surplus, analyzing for each the efficiency properties of the solution and its ease of implementation under a variety of assumptions. These instruments include having the platform act as a middle-man to make the transfers that agents cannot make directly make, utilizing two-part tariffs, splitting the platform along quality, and censoring agents' choice sets (i.e. curating the set of partners they can see on the platform)
Competition with Local Network Effects Abstract:There is a significant literature on how network effects influence the structure of competition. However, most of this research assumes global network effects where agents have preferences over the total number of peers on the platform and everyone using a platform gets the same benefit. In practice, however, we’d expect to see local network effects, where different individuals receive different network effects based on their idiosyncratic network of peers or connections. In particular, social networks often display a high degree of assortation, where people tend to be similar to their peer network, and clustering, where two peers of a given agent tend to also be peers with each other. I study a duopoly model with Salop style horizontally differentiated platforms motivated by social networks like Facebook and Twitter and introduce a simple network structure that allows me to model networks ranging from global to local, making a rich model of competition tractable with local network effects. I find that local network effects allow consumers to retain more surplus and that, in many cases, prices (or ad intensity) may be held below the competitive level by the threat of unraveling when agents are priced out of the market, especially with local network effects. Because of assortation, the peers of agents that are priced out of the market are likely to be close to marginal as well, so losing a few peers on the platform will push them below indifference. Their friends are also likely to be close to marginal, and have likely lost both the first and second set of peers due to clustering. Thus, local network effects greatly increase the likelihood of unraveling. I also find that local network effects can reverse the dynamics of competition relative to global network effects. With local network effects, higher complementarity between platforms hinders their ability to extract surplus, while with global network effects complementarity helps platforms extract surplus. Complementarity may be considered a choice variable for a platform (e.g. the degree of integration between two social networks), so optimal firm behavior may vary greatly depending on the type of network effect.