- October 18, 2020
Econ 482 – Applied Economic Research
Anatomy of an Applied Research Project: The Literature Review
An important component of any research paper is placing in context of the relevant literature. This is typically done through the Literature Review or Context section of the paper. This section of the paper usually comes after the Introduction.
What is a Literature Review and Synthesis Matrix?
The goal of a literature review is to convincingly place your paper in context of the existing (and recent) literature. This means it is insightful, comprehensive but not too long. A good and compelling literature review does not simply summarize the papers in a sequential manner. Instead, it emphasizes the content and contribution of the papers as a body of literature.
You should be thinking about the following broad areas in developing your literature review:
- Policy context
- Theoretical context (if applicable)
- Empirical challenges
- Theoretical challenges (if applicable)
- Progression of literature
- Current state of literature
- Ideas about how to move literature forward.
A useful way to construct a good literature review is to first create a literature synthesis matrix. Please read the document titled “Writing a Literature Review and Using a Synthesis Matrix” for useful points about what a literature review should be and how to develop a synthesis matrix. Here is what this document says about creating a synthesis matrix:
“The synthesis matrix is a chart that allows a researcher to sort and categorize the different arguments presented on an issue. Across the top of the chart are the spaces to record sources, and along the side of the chart are the spaces to record the main points of argument on the topic at hand. As you examine your first source, you will work vertically in the column belonging to that source, recording as much information as possible about each significant idea presented in the work. Follow a similar pattern for your following sources. As you find information that relates to your already identified main points, put it in the pertaining row. In your new sources, you will also probably find new main ideas that you need to
add to your list at the left. You now have a completed matrix!”
Creating Your Synthesis Matrix
The first step to creating a synthesis matrix to create a chart where you will place the content of the papers your are including in your literature review.
Begin the chart by labeling the columns both vertically and horizontally. Here is an example:
|Source #1||Source #2||Source #3||Source #4|
|Main Idea A|
|Main Idea B|
Label the columns across the top of your chart with the author’s last name or with a few keywords from the title of the paper or both. You may want to include the year of the publication. Then label the sides of the chart with the main ideas that your papers discuss about your topic.
As you read each paper, make notes in the appropriate column about the information discussed in the paper.
Here is an example of a synthesis matrix that I started (but did not complete) for a paper I wrote with several co-authors in April. However, the two papers in the synthesis are the two papers most relevant to our research questions.
|Synthesis Matrix for Flu/Sports Project|
|Idea||Stoecker et al. (2016)||Adda (2016)|
|Viruses are a major threat to human health & impose costs on society||U.S fed gov’t spent $3 billion in direct response to 2009 H1N1 pandemic (p. 125); uncontained pandemics costly in terms of loss of life – 1918-19 pandemic estimated 500,000 deaths, 1957- 59 pandemic estimated 86,000 deaths (p. 125); costs are underestimates b/c do not include long-term costs of inutero exposure (p. 125); vaccination ismost effective preventive strategy but is often unavailable in the short term due to production constraints||Over last 100 years, viruses responsible for more deaths than all armed conflicts(p. 891); viruses impose costs on society through premature deaths, long-lasting morbidity, increased health care utilization; loss of schooling or hours of work (p. 892)|
|Role of social interaction, mixing & economic activity in the spread and containment of viruses.||Influenza is an infectious disease that spreads by airborne droplets w/ ap-proximate travel radius of 6 feet making human contact an important infection vector p. 126; altering human mixingrates & limited contact with vulnerable populations are strategies to reduce transmission (p.126); use team success (participation in Super Bowl) as a proxy for shifts in socializing behavior during periods of higher influenzatransmission – Jan & Feb(p. 126)||prevalence of disease can be linked to social network structures & have long=term effects on growth (p. 892)|
|Estimation strategy: overall ap-proach, DV, main variable of interest(VofI), specification (fixed effects, time trends), how standard errors areestimated, weighting||DiD; DV is flu mortality in county c at season s; VofI: indicator = 1 when teamin MSA in county c makes it to Super Bowl; focus on population over 65; county and season fixed effects; linear& quadratic time trends (in some specifiations); weight by population of indicated age cell; s.e. clustered at MSAlevel (p 129)|
|Identifying Variation/IdentificationStrategy||Consider inuenza transmission in context of repeated event – participation in Super Bowl (measure of team success) – with explicit treatment and control groups (p. 126)|
|Data Sources & Control Variables||MCD (Vital Stats) for flu deaths & other noncommunicable deaths; SEER (NCI) for population; dates & locations of Super Bowls from 1974 to 2009;Global Summary of the Day for humidity; Global Historical Climatology Network for precipitation; % population65+; % pop nonwhite (p. 130)|
|Potential Mechanisms||(1) increase in large gatherings; (2) changes in travel patterns to attend post-season games; (3) surge in local economic activity for hosting cities through increase tourism or increasedlocal expenditures (p. 127-128)|
|Main Results||Counties in MSAs w/ teams participating in Super Bowl see an increase in flu deaths of approx. 18% (p. 128); no in-creases in mortality in host cities – Super Bowl related travel does not pose significant transmission risk (p. 128,p. 132); larger impact in seasons where Super Bowl was closer to peak of flu season (p. 134); effect of making SuperBowl on flu deaths is greater in flu seson dominated by more virulent strains (p. 134); placebo tests – no impact for noncommunicable diseases or for flu transmission one to two years before successful year (p. 129; p. 137-138);suggestive evidence that making playoffs is associated with higher flu deaths but limited to teams that win playoff games (p. 139)|
|Evidence of Mechanisms||No impact of hosting Super Bowl on local influenza mortality so no evidenceof change in mixing patterns resulting from travel mechanism (p. 139); increased mortality in counties associatedwith teams that make it to Super Bowl- evidence of change in mixing patterns in teams’ home cities possibly due tomore large gatherings in bars, restaurants and private venues (p. 140)|
|Contributions||Successful NFL team can impose a negative externality on MSA’s elderly population; major contributor seems to belocal gatherings for watching games; policy implication – increase awareness of flu transmission vectors during times of sports-related gatherings (p. 140);look at more than one-time event such as Olympics or rock concert (p. 126)||high frequency, detailed data for France allow analysis of spatial & temporal evolution of flu in a developed economy (p. 892); use quasi-experimentalvariation to assess effectiveness of various events like school closures on transmission of viruses (p. 892); investigate long-run economic determinants of spread of viruses (p. 893); schoolclosures reduce incidence of flu with a stronger effect for children but increase incidence of disease for the elderly – suggests spillovers across age groups (p. 894); find decreases in transmission rates during public transportation strikes & increased transmission ratesfollowing extension of railway lines (p. 894).|
Example of Literature Review
I now provide the Context section of the paper that I started the synthesis matrix for.
Influenza is an infectious disease transmitted through airborne particles from an infected individual, generated by coughing, sneezing, or potentially talking or shouting, coming in contact with an uninfected individual. Another means of transmission occurs when an infected individual expels virus onto surfaces like railings, seats, doorknobs, and sink fixtures and an uninfected person touches this surface, gets u virus on her hands, and then touches her face and contracts the virus.
Uncontained influenza epidemics and pandemics impose sizeable economic costs, in terms of loss of life, health care costs, unmet health care needs, productivity losses, and reductions in human capital, generating large incentives to control and prevent their spread. While vaccination and herd immunity likely represent the most effective preventative strategies, they are not viable options in the short-run due to the time required to develop and produce an effective vaccine and to achieve an effective level of herd immunity. COVID-19 represents a new and especially virulent virus with no existing vaccine or herd immunity, amplifying the importance of preventative policies.
Feasible short-run strategies for reducing transmission include altering social interaction patterns through social distancing and limiting contact with vulnerable populations. Since the reporting of the initial novel coronavirus outbreak in China in December 2019, social distancing through “stay at home” directives, self-imposed and mandated isolation, and shutting down non-essential businesses has been the prevention strategy of the vast majority of states in the United States and countries around the world.
However, weighing the economic costs of these “stay at home” and lock down directives against the economic costs of the virus is an important and emotionally charged consideration for government and health officials. After two or more months of limited economic activity that rapidly threw many large and growing economies into recession, the pressure to restart economic activity, including the resumption of sporting events, through relaxation of COVID- 19 restrictions is high.
We investigate the impact of the relaxation of policies prohibiting large gatherings on local health outcomes through the lens of professional sporting events, in particular games played in four major professional sports leagues (MLB, NBA, NFL, and NHL) in the United States. We posit that professional sports games increase transmission of influenza, negatively impacting local health outcomes compared to an environment where no games occur. Increased transmission eventually reaches vulnerable populations that never attended or watched these game, increasing flu mortality.
The relationship between large numbers of spectators attending sporting events and flu transmission relates to social distancing because attending games, or gathering in confined areas like bars, restaurants, or homes, to watch games, significantly reduces social distance. Sporting events place large numbers of people in close quarters. These fans touch many surfaces and engage in substantial talking, yelling, face touching, and person-to-person contact like “high fivers.” People gathering in homes or public spaces to watch sporting events on media creates similar conditions. Maintaining the current recommended social distancing guidelines of small gatherings with at least 6 feet of space between individuals is challenging, if not impossible, in a crowded sports venue or bar.
Recent research exploits plausibly exogenous events to advance understanding of the role of social distancing and economic activity in the transmission of viruses. One strand of literature examines the impact of large sports-related gatherings on influenza infections and mortality including participants in the 2002 Winter Olympics (Gundlapalli et al., 2006), spectators at the 2006 FIFA World Cup (Williams et al., 2009), and residents of cities home to NFL teams playing in the Super Bowl (Stoecker et al., 2016). A related literature develops evidence linking negative cardiovascular and infant health outcomes to sporting events (Kloner et al., 2009, 2011; Leeka et al., 2010; Duncan et al., 2017).
A second strand of literature explores the impact of large non-sports related gatherings and events on the spread of infectious disease. These gatherings include music festivals (Guti_errez et al., 2009; Botelho-Nevers et al., 2010), the annual Hajj pilgrimage to Mecca (Balkhy et al., 2004; Al-Taw_q et al., 2016), college spring break Mangrum and Niekamp (2020), and school closures and transit strikes in France (Cauchemez et al., 2008; Adda, 2016). American Academy of Pediatrics (2006) and Adda (2016) analyzed the effect of travel-related shocks, in the form of the 9/11 related air travel shut down and transit strikes respectively, on infectious diseases.
Stoecker et al. (2016) and Adda (2016) are the two most closely related papers in the literature to this study. Stoecker et al. (2016) exploit temporal and geographic variation in annual NFL team success as a proxy for changes in social distancing patterns through either increased fan social interaction or travel. They compare u mortality for the population over age 65 in counties in MSAs with NFL teams that played in the Super Bowl to counties home to NFL teams that did not reach the Super Bowl in a given season. They find that counties in MSAs with NFL teams that played in the Super Bowl caused an 18 percent increase in iflnuenza deaths for the population over age 65 in those counties. They posit increases in large gatherings in MSAs home to these successful teams lead to increases in frequency of
human contact and an increased probability of local transmission.
Interestingly, they reported no evidence of increases in u mortality in cities that hosted the Super Bowl, suggesting that changes in travel patterns bringing large numbers of spectators to the host city play a small role in the process. This result does not completely rule out changes in travel patterns as a mechanism for transmission. It is possible that the mechanism works in the opposite direction: fans who travel to the host city for the Super Bowl become infected there and bring the virus back home with them resulting in an increase in infections and deaths in the sending city.
This explanation is consistent with anecdotal evidence for the quick and deadly spread of the coronavirus in Bergamo, Italy, in February 2020. Health officials speculate the large number of fans (about 40,000) who traveled to Milan to attend a football match between the team in Bergamo and a Spanish team from Valencia contracted the virus in or en-route to and from Milan, brought it back to Bergamo and subsequently infected many local elderly residents who did not attend the match Giuffrida, 2020).
Adda (2016) uses high frequency data from France to evaluate the impact of plausibly exogenous events (school closures, public transportation strikes, and expansion of transportation networks in the form of new high speed train lines) on the transmission of infectious diseases, including influenza, for different age groups. Adda (2016) finds a substantial impact of these events on the incidence of viral diseases in the expected direction; school closures and public transportation strikes significantly reduced disease prevalence while expansion of train lines increased viral spread.
Leagues and local policy makers suspended games due to concerns and uncertainties surrounding the safety and health of fans, coaches, and players due to the pandemic, in particular, concerns about the role of large gatherings of fans amplifying virus transmission. Our empirical analysis generates evidence about the impact of restarting games, based on past health outcomes in cities adding new teams in four major professional sports leagues over the last half a century. Analyzing the impact of the commencement of play in cities represents a closer analog to the decision facing policy makers today to restart league play compared to the NFL team success-based approach used by Stoecker et al. (2016) and extends the analysis of the impact of rail-based transmission mechanisms used by Adda (2016) to a
Note the following about the Context section:
- It is not too long – just about 3 pages. Literature reviews and context sections typically do not exceed 3 to 4 pages of a paper.
- The first two paragraphs state a little about flu transmission and why it is important in a broad economic sense.
- The last line of the 2nd paragraph connects flu transmission to COVID-19 – an important component of my research question.
- We then put our paper in context of “stay at home” orders in the next three paragraphs. Before talking about the relevant literature, we say in one concise paragraph what we are doing in the paper.
- We then talk about the relevant literature and focus mostly on Stoecker et al (2016) and Adda (2016) because they are the papers most closely related to ours.
- The last sentence of the section says how our paper extends/contributes to this literature in the context of Stoecker et al (2016) and Adda (2016).