Does economic policy uncertainty increase information search about condoms on the internet?
The intention of this paper is to explore whether economic policy uncertainty has implications on the information-seeking behavior of the mass public in the United States regarding condoms. This is a departure from prior work on economic policy uncertainty, which has mainly emphasized the behavior of firms in response to economic policy uncertainty (e.g. Colombo, 2013; Julio & Yook, 2014), and not the behaviors exhibited by the general public in response to economic policy uncertainty. Economic policy uncertainty having a role in the health-related decision-making of the mass public has not been explored in-depth.
Whether economic policy uncertainty increases search interest about condoms online is a topic then that is worthy of exploration. The reason for this is that the Internet is an avenue in which many people now elect to seek out health-related information (Ayers et al., 2012). Some contributing factors to this are that many members of the public who have had relatively limited access to actual healthcare often rely on health-related information disseminated on the Internet, and that there is also a potential stigma that exists to seeking out information publicly about certain health-related topics (Althouse et al., 2014).
Eysenbach (2011) suggests that the Internet might be the most relied upon resource in the general public for information pertaining to health. The stigma some might perceive in the usage of condoms (Kulczycki, 2004) can also compel people to seek out information about condoms on the Internet, as opposed to other information channels. The proposal of this paper is that an increase in economic policy uncertainty makes people more mindful of the potential costs involved in experiencing unintended pregnancies and/or sexually transmitted infections, and that this results in an increase in the search for information on the Internet about condoms. An increase in the search volume interest for specific items can be used as a leading indicator for the subsequent purchase of those items (Goel et al., 2010; Choi & Varian, 2012), so it is useful to see whether economic policy uncertainty has any bearing on search interest about condoms.
Economic Policy Uncertainty: A Brief Background
Economic policy uncertainty is a concept that represents the absence of clarity within a country about the current and prospective qualities of policies that can have significant ramifications on the general state of the economy (e.g. monetary, tax, spending, and regulatory policies). Baker, Bloom & Davis (2013) have developed a dynamic (changing over time) indicator of economic policy uncertainty in the United States. The same authors have used similar approaches to develop separate indicators for other nations, such as Spain, the United Kingdom, Canada, and India.
A priority behind the original development of this dynamic indicator of economic policy uncertainty in the United States was to explore whether the presence of economic policy uncertainty stymies recovery from economic downturns. Economic recovery can be stunted, since economic policy uncertainty can drive firms to become too risk-averse to engage in the level of investment, hiring, and consumption needed to spur economic growth. Heightening concerns about the makeup of the country's economic policy makes firms less willing to engage in behaviors that can stimulate economic growth.
In the empirical work of Baker, Bloom & Davis (2013) that analyzes their indicator of economic policy uncertainty, they find that a rise in economic policy uncertainty does produce risk-averse behaviors. Their work suggests that an increase in economic policy uncertainty results in reduced investment and hiring by firms, as well as a decline in industrial production. Additional work assessing economic policy uncertainty, be it either theoretical or empirical in nature, has also been focused on the macroeconomic consequences of, and at times, the macroeconomic causes of, surges in economic policy uncertainty.
For instance, Pastor & Veronesi's (2012) theoretical model suggests that when uncertainty about economic policy is high, stock prices will fall. Empirical research on the level of uncertainty held about election outcomes, and the potential ramifications of election outcomes on future economic policy, indicates that a reduction by U.S. firms in foreign direct investment to foreign affiliates occurs in election cycles (Julio & Yook, 2014). Economic policy uncertainty in the United States also appears to result in a statistically significant decrease in industrial production in Europe. In fact, the level of economic policy uncertainty in the United States has a greater impact on European industrial production than European-specific economic policy uncertainty (Colombo, 2013).
When economic policy uncertainty and firm-level uncertainty interact, firms can become more guarded about their investment plans (Kang, Lee & Ratti, 2014). Empirical research attempting to explore potential causes of economic policy uncertainty demonstrates that increases in the real price of oil that cannot be attributable to shifts in either global real demand for commodities, or in global oil production, can contribute to increases in the level of economic policy uncertainty (Kang & Ratti, 2013).
All of this work has helped to clarify how economic policy uncertainty can alter, or is affected by, economic conditions and the behavior of economic firms. A relatively untapped area of potential is research examining the response of the mass public to shifts in economic policy uncertainty. An aspect worthy of examination is whether an increase in economic policy uncertainty can result in changes in the sexual practices of the mass public. Since there are not many actual time-refined indicators in existence of the sexual practices of the mass public available to contrast with a monthly measure of economic policy uncertainty, an alternative, but related consideration is worth exploring. Whether heightened economic policy uncertainty produces an increase in the level of search interest within the public about condoms can demonstrate that people increasingly consider the consequences of unprotected sexual intercourse in the face of uncertainty about the current and prospective makeup of economic policy.
This project evaluates this topic through the examination of Google Trends search interest about condoms. As stated above, economic policy uncertainty can increase risk-averse behavior. Economic policy uncertainty can make the potential costs (be they health-based, financially-based, or both) of having children and/or contracting a disease by a lack of condom usage less acceptable. As a result, an increase in economic policy uncertainty should bring about an increase in interest about condoms. People attempt to reduce the level of uncertainty about potential outcomes by acquiring and evaluating information that is available (Taylor, 1974). In this case, people uncertain about the ramifications of the makeup of economic policy on their financial security and wellbeing will seek out information about condoms. An explanation for this is that economic policy uncertainty can produce concern about all of the ramifications of an unplanned pregnancy and/or contracting a sexually transmitted infection.
Potential Consequences of Uncertainty on an Information Search Regarding Condoms
When decision-makers face uncertainty, people will seek to minimize their personal losses (Fischhoff & Kadvany, 2011). A common response to uncertainty is to engage in an increased search for information (Lemieux & Peterson, 2011). In other words, the uncertainty about the consequences of economic policy on the financial health and status of individual households should produce an increased interest in information about topics or items that can help to lower perceived risks. Reducing perceived risks is a reaction from the mass public to a period where a broad swath of the civilian population, much like firms, has concern about the current and foreseeable makeup of economic policy.
Economic policy uncertainty induces concern about the current and future economic climate, such that we should expect people to be more mindful of the potential costs that are involved in having and raising a child, or to the potential costs of treating and living with a sexually transmitted infection. If there is an increase in concern about the potential costs of having a child and/or living with a sexually transmitted infection, we should see an increased level of interest in searching for information about condoms.
When an unintended pregnancy occurs, people often feel woefully unprepared financially to take on parental responsibilities (Henshaw, 1998); this perceived lack of financial preparation should only be exacerbated if economic policy uncertainty is high. Family planning decisions are tied to the economic health of a household, as many people see the prospect of having a child as too burdensome of a demand on both their time and resources when changes to family income are occurring, or are expected (Schultz, 1969). Condoms are seen as an option in terms of family planning that can prevent unintended pregnancies (Abma et al., 1997). An increased interest about condoms can result in the face of increased economic policy uncertainty, because the prospect of having a child in a hazy economic climate can only help to increase fears that having a child will limit a person's ability to support themselves financially, maintain a job, achieve their personal goals, or have the level of control they would prefer to have over their own life (Frost & Lindberg, 2013).
Condoms are also widely seen as way to lower the likelihood of contracting a disease (Mosher & Bachrach, 1996). A barrier to the expanded usage of condoms is the assumption that people are not risking anything when engaging in unprotected sexual intercourse (Tafuri et al., 2010). If economic policy uncertainty is high, and people can become more mindful of the potential risks that are present in the environment, people can potentially begin to be more mindful of the potential ramifications involved in risky behaviors, such as the practice of unprotected sexual intercourse.
As Tversky & Fox (1995, 269) state, when faced with making a decision under uncertainty, people assess the desirability of possible outcomes, as well as the likelihood those outcomes are to occur. Economic policy uncertainty, which can help to cast doubt about the current economic climate and future economic prospects, should make people more mindful of the consequences of engaging in unprotected sexual intercourse. Increased economic policy uncertainty should make people more mindful about the desirability of an unintended pregnancy and/or contracting a sexually transmitted infection. As a result, people should seek out information about condoms, items that can help to lower the likelihood of an unintended pregnancy and/or contracting a sexually transmitted infection.
Google Trends as an Indicator of Topic Interest within the Mass Public A measurable, time-refined means in which it is possible to gauge the level of information seeking about condoms is through publicly accessible information provided by Google (2014a) on the extent of search interest about specific topics or items. Google Trends, a website freely available to everyone (http:// www.google.com/trends/), allows people to get an indication of how often a specific search term is entered into the Google search engine at a specific point in time, and monitor the shift in search interest in that term across time.
People use the Google search engine to seek out information about subject matter of interest to them (Gunn & Lester, 2013). Internet search volume levels reflect what people are currently thinking about, and how they might potentially behave (Wilson & Brownstein, 2009). People can demonstrate their desires, interests, and concerns by the terms used in their Internet search behavior (Ettredge, Gerdes & Karuga, 2005). When an increase in Internet searches about a particular topic occurs, people are increasingly thinking about that topic (Ripberger, 2011). If enough members of the public in a specific region search for the same term during the same time period, that means there should be an increase in relative interest in that search term within that particular region.
There is a substantial body of scholarship that has come about in recent years contrasting Google Trends search interest levels with health, economic, and cultural indicators. Regarding health, epidemiologists use information from Google to track epidemics by contrasting the search volume dynamics of terms like 'food poisoning' and 'Salmonella' to confirmed infections reported by the CDC on the number of Salmonella Typhimurium infections (Brownstein, Freifeld & Madoff, 2009). Similar work has been done contrasting search interest and treatment for kidney stones (Willard & Nguyen, 2013), as well as chickenpox (Valdivia & Monge-Corella, 2010). Whether or not search volume trends help in predicting influenza epidemics is also an area of focus in this line of research (Eysenbach, 2006). Researchers have employed Google Trends as a tool to track Internet search activity levels about specific keywords to assist in the detection of a spread of an illness.
Contrasting Google search volume interest levels about mental health topics to the extent of self-harm and suicide incidences is a prominent area of research on human and computer interaction. There is evidence of individuals that suffer from mental health disorders being increasingly likely to search for health-related information on the Internet (Berger, Wagner & Baker, 2005; Lam-Po-Tang & McKay, 2010). Sueki (2011) finds that searches for the term 'depression' are positively related with the monthly suicide rate in Japan. In the United States, when breaking down search interest by state, an increase in Google search volume for the terms 'suicide prevention,' 'commit suicide,' and 'how to suicide' within a state are all positively related with the actual suicide rate within each respective state (Gunn & Lester, 2013). Internet searches in the United States for suicide-related terms are positively correlated with intentional self-injury and suicide amongst people aged between 15 and 25 (McCarthy, 2010). Within this age group, the search for information about suicide-related terms might actually help to facilitate acts of self-harm.
Whether economic conditions influence the search interest for health concerns has also been analyzed. Economic strain could bring about an increased search interest about physical health concerns. Between 2008 and 2011, a period in which a major economic recession occurred, Althouse et al. (2014) found an excess for searches on terms related to physical health concerns (e.g. 'stomach ulcer symptoms,' 'headache symptoms,' 'chest pain,' etc.). Tefft (2011) discovers that an increase in the unemployment rate is significantly associated with an increase in the Google Trends search index for 'depression.' Research from Frijters et al. (2013) indicates a rise in the unemployment rate is followed by an increase in search terms related to alcoholism ('alcohol,' 'alcoholic,' 'alcoholics,' 'alcoholism,' and 'AA'). Google search interest for terms related to psychological distress increased by 16%, for each 1% increase in the level of home foreclosures in the prior month (Ayers et al., 2012).
There is also evidence of a relationship between changes in economic conditions and Google search interest levels on topics pertaining to the economy. There is a positive and statistically significant association between search volume for keywords related to people searching about jobs and the unemployment rate (Ettredge, Gerdes & Karuga, 2005). When evaluating monthly indicators in the country of Germany, Google search interest trends pertaining to terms related to unemployment are correlated with the unemployment rate in that country (Askitas & Zimmerman, 2009).
Research on Google search interest and cultural topics also demonstrates that what the public searches for online helps to predict collective behavior in advance. Search interest levels for films can predict box office figures for those films, search interest levels for video games can predict initial month sales for those specific games, and search interest levels for songs can predict where those songs will rank on the music charts (Goel et al., 2010).
There appears to be an extensive body of scholarship providing evidence that Google Trends search volume about health, economic, and cultural topics are associated with actual statistics regarding health, the economy, and cultural items. As a result, this project attempts to explore whether the presence of economic policy uncertainty predicts change in information search about condoms, through the examination of Google Trends search volume dynamics about condoms. As previously stated, an increase in economic policy uncertainty should make many members of the mass public concerned that they do not have the capacity to properly handle either an unintended pregnancy, or a sexually transmitted infection. These aspects should be a widespread concern given the rising costs of health care in the United States (Popescu, 2014). Economic policy uncertainty should lead people to attempt to lower uncertainty about potentially undesirable outcomes by compelling them to search for information about an item that can help to lower the perceived risks of those outcomes. Due to this, the research hypothesis evaluated in this project is the following:
Research hypothesis--A positive change in economic policy uncertainty significantly and positively predicts change in Internet search volume interest about condoms.
2. Materials and Methods
There are two main dynamic variables that are studied in this project, economic policy uncertainty, and search interest for condoms in the United States. This section will cover several aspects about the research design of the project. First, details are given about how information regarding these two variables is collected. Second, a discussion about alternative variables that also need to be measured is offered, as there are other considerations that might influence the two primary variables of interest. Last, this section will chronicle the empirical techniques that are used to evaluate the research hypothesis, while still controlling for the alternative explanations for shifts in either economic policy uncertainty, or Internet search interest about condoms.
Measuring Economic Policy Uncertainty
The monthly indicator of economic policy uncertainty in the United States developed by Baker, Bloom & Davis (2013) is comprised of three components. The first component is the number of references to policy-related economic uncertainty discussed in major American newspapers. The second component measures the amount of federal tax code provisions that will potentially expire. The third component is the amount of disagreement amongst economic forecasters regarding government purchases and future inflation.
With the first component of the economic policy uncertainty index, the American newspaper content analysis measures the amount of uncertainty expressed along several fronts: uncertainty about who is making definitive economic policy decisions, uncertainty about the time period when specific economic policy actions will occur, uncertainty about the consequences of policy decisions, and uncertainty about the possibility of inaction. The content analysis work incorporates policy concerns that are not explicitly about the economy, but do have a bearing on the economy, such as homeland security. The content analysis is performed by searching for keywords that incorporate terms that relate to the economy, uncertainty, and policy in some way. The newspapers used in the content analysis work are the following: the New York Times, the Los Angeles Times, the Chicago Tribune, the Wall Street Journal, USA Today, the Washington Post, the San Francisco Chronicle, the Boston Globe, the Dallas Morning News, and the Houston Chronicle.
The second component of the index is the analysis of tax code expirations. This is evaluated by looking at the dollar value of potentially expiring tax provisions across time. The numerical amount of scheduled tax code expirations can increase uncertainty, since Congress will in many cases wait until very close to the expiration date of a tax code before actually deciding what action will be taken regarding the provision.
The third component of the index is comprised of information from the Federal Reserve Bank of Philadelphia's Survey of Professional Forecasters. The extent of the difference between individual forecasts in the survey responses of economists is a representation of uncertainty about future economic outcomes. The interquartile range of the inflation rate forecasts, as well as forecasts about government purchases at the local, state, and federal levels, provide the means in which disagreement amongst forecasters is evaluated.
With information collected about all three of these of components, Baker, Bloom & Davis (2013) normalize each component by its own standard deviation, and then average the weighted components each month. The end product of this is a monthly indicator of the amount of economic policy uncertainty, a measure that can be used in dynamic comparisons with the level of public search interest about condoms.
Measuring Information Search about Condoms
To collect a monthly indicator of public information search about condoms to contrast with the monthly indicator of economic policy uncertainty, information from Google Trends is used. The search was filtered to look for the volume of search interest in the United States between January 2004 and December 2013, as the earliest month of information available from Google is for January 2004. Google provides a query share for the search term that is entered over the Google Trends interface. In other words, Google does not provide the absolute total number of searches for a term. Only the search query share is reported.
The calculated search query share is first determined by dividing the total search volume for a specific term at a specific time by the total number of queries in the United States during the entire period studied. Next, this particular value is multiplied by 100. So for each month, a scaled value between 0 and 100 is calculated (Google, 2014b). A value of 100 represents a maximum query share, and designates a month where the highest relative search activity for a specific term occurred. A query share value of 100 reflects the highest volume of search for a specific term during the period studied within the United States. A query share value of 0 does not mean that absolutely no searches for a specific term occurred; rather it likely means the minimum threshold for query volume was not reached for that particular month (Google, 2014b). The scaled value provided by Google is a means to determine how likely a random user of Google will search for a specific term, within a specific period of time, and within a specific location.
When plotting the query share time series, a rising line indicates an increase in the specified search term's popularity, while a declining line indicates a decrease in query share. When entering 'condoms' into the Google Trends interface, the Google Trends report of search volume share for this term includes the level of interest in any phrase containing the term 'condoms,' such as 'condoms safety,' 'buying condoms,' or 'putting on condoms.' The benefit of using the term 'condoms' as the search topic for this research project is that the item itself can be easily represented by this single search term. The term 'condoms' is unique enough that we should feel comfortable that someone using the Google search engine who enters the word 'condoms' is seeking to learn something pertaining to condoms. The term 'condoms' then is likely to capture the substantial share of the searches conducted for this item/topic.
Such an advantage is not always evident when assessing Google Trends data. For instance, if someone searches for the term 'deficit,' it is not immediately clear if they are interested in the country's economic deficit, or they are interested in learning more about attention deficit hyperactivity disorder (ADHD). This concern requires refining the Google Trends information into one of its specific subcategories, such as "Health," "Finance," or "Business and Industrial." Since the term 'condoms' is very specific about a particular item/topic, there is no compelling reason to have to use a specific Google Trends subcategory, and instead we can rely on the general query share for the term. In other words, people engaged in a search about condoms are likely to use the term 'condoms' within their search, such that concerns about multiple interpretations of the keyword term are not warranted.
Another element of concern that needs to be addressed is whether Google Trends search volume information is a representative measure of public interest in learning more about condoms. This concern might be raised by those who believe Internet search trends will under-represent particular demographic groups that are unlikely to use the Internet. In order to assuage this concern, some research has validated Google Trends search interest levels about specific topics with public opinion survey information on specific topics. For example, in comparing Google Trends search interest levels about German political figure Paul Kirchhof to time series public opinion survey information gauging public interest about Paul Kirchhof, Scharkow & Vogelgesang (2011) find a statistically significant positive correlation of .49 exists.
In addition, there is a strong amount of evidence that Google is a heavily used tool to search for information. Google is far and away the most used search engine in the United States, as Experian Hitwise (2014) identified the Google search engine as enjoying over twice the number of visits than the next four top search engines (Yahoo! Search, Bing, Ask, and AOL Search) have combined. In terms of global usage, Netmarketshare (2014a) reports that in June 2014, Google had 68.75% of the desktop search engine market share, and Netmarketshare (2014b) reports Google had 91.02% of the mobile/ tablet search engine market share.
Google is the largest search engine in the world, both in terms of the number of web pages indexed by Google, and by the number of actual users (Dzielinski, 2012). Since at least 77% of the adult population uses the Internet (Ripberger, 2011: 251), and there is an overwhelming level of popularity in Google as a search engine both domestically and globally, any information Google Trends offers about search interest levels should be considered a viable indicator of the dynamic level of search interest about specific topics. An argument can be made that the dynamics of search interest on Google is representative of the level of interest the general public at large has about a topic (Dzielinski, 2012).
Since the measuring procedure for the two primary variables of interest have been described, it is worthwhile to visually see the dynamics of both of these variables for the time period studied. Figure 1 offers a visualization of both the economic policy uncertainty series and the Google Trends series reporting search interest volume about condoms. It is not immediately obvious based on the dynamics presented in Figure 1 how economic policy uncertainty and search interest about condoms relate to each other. It is essential for time series analyses to empirically determine whether changes in economic policy uncertainty can predict changes in search interest about condoms. Before discussing the necessary time series analyses, it is worthwhile to discuss alternative variables that need to be accounted for in any time series analyses assessing a potential relationship between economic policy uncertainty and search interest about condoms.
Exogenous Control Variables
To assess the potential role of other variables that can have a significant effect on economic policy uncertainty or search interest about condoms, several variables are examined as control variables in the project.
Baker, Bloom & Davis (2013) raise the possibility that the occurrence of events can contribute to economic policy uncertainty spikes. It is certainly possible these events can also result in a change in the level of search interest about condoms. The economic events accounted for in this project, with a separate dichotomous dummy variable indicator included for each, are the following: large interest rate cuts and stimulus (January 2008), Lehman and TARP (October 2008), the banking crisis (February 2009), the debt ceiling dispute (August 2011), and the government shutdown/debt ceiling dispute (October 2013). Since there is not a scholarly consensus about when the Great Recession began, and when it is possible to say the Great Recession ended (Dzielinski, 2012), the project decides to account for all the months the National Bureau of Economic Research (2014) designates as a recession period during the timeframe studied, December 2007 to June 2009. For all of these event variables, a score of "1" represents the occurrence of an event or period of interest within a specific month, while a score of "0" represents an event or period of interest did not occur in that month. Since the occurrence of elections can heighten uncertainty, every November with either a congressional and/or presidential election is accounted for with a score of "1" in a single variable measuring the occurrence of elections.
To account for the possibility that media attention to condoms can influence public attention to condoms, since mass media coverage can influence public attention to topics (McCombs, 2004), an indicator of media coverage of condoms is included. This attempts to account for the possibility that an increase in mass media coverage about condoms can increase the level of public search interest on information about condoms. As a measure of the level of news coverage about condoms in the mass media, the total number of evening news segments or news specials broadcast by ABC, CBS, NBC, CNN, or Fox News about condoms is collected. This information is found by performing a search of news abstracts that include the words 'condom' or 'condoms' through an advanced keyword search of the Vanderbilt Television News Archives (2014).
To account for dynamic indicators relating to general characteristics of the economy, variables representing unemployment and real disposable personal income are included. In terms of unemployment, the variable measured represents the monthly sum of initial claims in the unemployment insurance weekly claims report. Regarding real disposable personal income, the billions of dollars in chained 2009 dollars represents the level of income available for savings and spending purposes. Both of these measures come from information that is made available to download from the Federal Reserve Bank of St. Louis (2014).
With all of these control variables described, it is possible to now describe the analytical procedures used to empirically evaluate the relationship between economic policy uncertainty and search interest for condoms.
Time Series Methodology
For the purposes of this project, the procedures of vector autoregression and moving average representation are employed. The reason these time series techniques are used is because these techniques are appropriate when research wants to assess whether change in one variable can help to predict change in a second variable, while simultaneously accounting for the possibility that the second variable can predict change in the other variable. This means that these procedures allow for the evaluation of a multidirectional relationship between the variables of interest. We want to be confident that change in economic policy uncertainty predicts search interest level change about condoms, and that there is no evidence of the unlikely possibility that change in condom search interest levels predicts change in economic policy uncertainty levels.
Vector autoregression is a multivariate application of the Box-Jenkins causal model (Freeman, Williams & Lin, 1989). Vector autoregression (at time abbreviated as VAR) helps determine whether prior change in one variable predicts change in another variable, without imposing a theoretical restriction in the analysis as to which of the variables in the vector autoregression system are a priori exogenous. As Enders (1996, 106) states when describing the endogenous VAR system, "VAR treats all variables symmetrically, without making reference to the issue of dependence versus independence."
When performing a vector autoregression analysis, all the variables measured within the system are treated as endogenous variables. Each variable in the system is regressed on prior values of itself, as well as prior values of the other variables in the system. This technique makes it possible to get a sense as to the potential presence of a causal relationship between the variables within the system, while also accounting for the possibility of a multidirectional relationship. The two endogenous variables within the vector autoregression system for this project are economic policy uncertainty and the level of search interest about condoms. Remember, alternative explanations for changes in values of the variables within the endogenous system are accounted for by the usage of exogenous controls.
The performance of a vector autoregression calls for bringing the endogenous variables together into a single vector, with the vector represented as a linear function of its own lagged values, as well an error vector. The actual vector autoregression estimation is performed through the performance of separate regressions for each vector (Kennedy, 2003). Granger tests are conducted for the hypothesis on the restriction that the lags (prior values) of a variable do not significantly enter into the regression of a specific endogenous variable (Granger, 1969). The Granger causality tests essentially evaluate whether the lags of a variable can influence a particular endogenous variable. When the F-tests indicate a block of coefficients is statistically significant, that implies a Granger causal relationship exists between a variable and a specific endogenous variable in the system. In other words, the analyses assess whether prior values of one variable can predict current values of another variable.
In order to assess whether standard vector techniques are appropriate, it is essential to determine whether the variables are cointegrated, meaning they share a common trend across time. If the variables are cointegrated, standard vector autoregression cannot be used. To determine whether we should be worried about cointegration, the two variables in the endogenous system are evaluated for whether they are stationary. Stationary variables are those where the statistical properties of those variables (e.g. mean, variance, etc.) are constant over time, and do not follow any dynamic trends. Two variables that are stationary inherently cannot be cointegrated. The augmented Dickey-Fuller (Dickey & Fuller, 1979) and Phillips-Perron (Phillips & Perron, 1988) diagnostic tests reported in Table 1 indicate that we can reject the null of the presence of a unit root (non-stationarity) in each of the two variables in the system. As a result, performing standard vector autoregression techniques should be appropriate, as each variable is stationary. Engle-Granger (Engle & Granger, 1987) tests for cointegration (results not reported here for the sake of space, and to limit redundancy) reinforce the view that cointegration cannot exist in the system studied in this project.
While vector autoregression can help in determining whether change in one variable can predict change in another variable, this time series procedure does not help to clarify the polarity (positive or negative direction) and the magnitude of any relationship between the endogenous variables. The coefficient estimates performed by Granger causality tests are not particularly informative. The reason for this is that the usage of lags (prior values) produces concerns that coefficient estimates are not necessarily informative due to the presence of multicollinearity. A way in which to determine the polarity and magnitude of relationships between endogenous variables within a vector autoregression system is to perform a moving average representation analysis.
When conducting a moving average representation analysis, a simulated shock is induced on a single variable in the endogenous system, and then the consequences of that shock are assessed over an extended period of time (Wood, 2009: 172). Upon performing a shock to a single variable in the system, it is possible to see the response the other variable in the system has following the increase (shock) that is induced on the single variable in the system. The usage of moving average representation offers an indication of what happens to a specific variable when a change is induced on another variable in the system. To establish a more consistent interpretation of the amount of change a variable displays after another variable experiences a simulated shock, all of the variables in the system are standardized. When a variable is standardized, it is rescaled to have a mean of zero and a standard deviation of one.
3. Results and Discussion
The results of the time series analyses provide support for the research hypothesis of this project. The vector autoregression analysis suggests that prior economic policy uncertainty predicts the current search interest level about condoms. The moving average representation analysis indicates a positive increase in economic policy uncertainty results in a statistically significant increase in search interest levels about condoms that lasts for an entire year. These results are present, even though a multitude of alternative explanations are accounted for as exogenous controls. At least for the time period studied (January 2004 to December 2013), and in the specific country studied (the United States), there is evidence that an increase in economic policy uncertainty brings about an increase in search interest levels about condoms.
Vector Autoregression Analysis
Table 2 presents the results of the vector autoregression analysis. The analysis indicates that when the search interest level about condoms is the endogenous variable, the prior level of economic policy uncertainty Granger-causes current search interest levels about condoms (p = 0.00). That means the prior level of economic policy uncertainty can be used to significantly predict the current level of public search interest about condoms that is reported by Google Trends. Prior public search interest about condoms also Granger-causes current public search interest about condoms (p = 0.00), indicating that search interest levels in condoms are inertial. This means previous interest about condoms can help to predict current interest about condoms.
The prior level of search interest about condoms does not significantly predict the current level of economic policy uncertainty, when economic policy uncertainty is treated as the endogenous variable (p = 0.15). Economic policy uncertainty is a variable that is inertial though (p = 0.00). Prior search interest about condoms as reported by Google Trends does not predict the current value of the economic policy uncertainty index. These findings are the expected result, as a change in condom search level interest should not logically result in a change in economic policy uncertainty.
Thinking conceptually, a prior change in economic policy uncertainty predicting the current level of search interest about condoms makes sense, and not the other way around. A change in economic policy uncertainty should drive people to contemplate whether the potentially undesirable outcomes that can result from a lack of condom usage (such as an unintended pregnancy, or contracting a sexually transmitted infection), warrants a search for information about condoms.
In terms of exogenous controls, only two of the variables measured predict current levels of search interest about condoms, real disposable personal income and the recession period. With both of these variables, the direction of the relationship is negative. This could mean that as real disposable personal income increases, people are less apt to search for information about condoms, since they have more income available for spending and saving. As a result, there is less of a perceived need to search for information about condoms. Regarding the recession period, in the timeframe where a recession period is said to have occurred, search interest about condoms declines. This could be due to the possibility that the mass public, fully aware of the economic downturn given the characteristics of recession periods, are more likely to search for terms pertaining to specific economic relief. Future study will have to explore this possibility.
Regardless of the exogenous control results, change in economic policy uncertainty, at least for the time period studied in the United States, appears to be driving a significant level of interest in searching for information about condoms, even though multiple other explanations are accounted for. The vector autoregression results provide support for the proposal that economic policy uncertainty makes people mindful of the potential costs involved in experiencing unintended pregnancies and/or sexually transmitted infections, such that their level of search interest about condoms is significantly altered.
Whether or not economic policy uncertainty makes people more or less mindful though cannot be assessed with confidence through vector autoregression alone due to the multicollinearity concern. As a result, the usage of moving average representation is necessary.
Moving Average Representation Findings
Assessing the moving average representation findings helps to gain an understanding about the polarity and the magnitude of the relationship between economic policy uncertainty and search interest levels about condoms. As was also seen in the vector autoregression analysis, the results here offer support for the project's research hypothesis. An increase in the economic policy uncertainty index produces an increase in the search interest level about condoms at levels above the standardized mean. This finding persists for an entire year following the shock (increase) in economic policy uncertainty. The visual dynamics of the moving average representation analysis are presented in Figure 2.
Figure 2 illustrates the response to simulated shocks to each variable within the endogenous system. Shocking each variable in the endogenous system mathematically is performed to gauge the expected response the other variable in the system will display over time. The confidence intervals of the expected responses to the shocks are determined by Monte Carlo integration and the fractile method (Sims & Zha, 1999). The number of months after a positive shock to a variable is measured along the horizontal axis of each individual plot. The positive or negative shift from the standardized mean following the simulated shock is represented along the vertical axis of each plot. The variable given a simulated shock is the same within each unique column of Figure 2. This means that in the first column, the variable that receives a one standard deviation positive shock (increase) is the economic policy uncertainty index. In row one of column one, the response of the economic policy uncertainty index to a one standard deviation increase in itself is displayed. In terms of this current project, we are primarily interested in the dynamics of row two of column one, which illustrates the polarity and magnitude of the shift in search level interest about condoms following the positive shock to the economic policy uncertainty index.
With the one standard deviation increase in the economic policy uncertainty index, there is no immediate significant shift visible in search interest levels about condoms at the point of the actual shock (increase) to economic policy uncertainty. This is visible when looking at period zero of row two of column one; the 95% confidence interval is not clearly bounded away from the standardized mean of zero. This is made evident because the entire confidence interval is not shifted away from the horizontal axis at the point where the positive shock to economic policy uncertainty occurs. Despite the lack of an immediate response, we see that the level of condom search interest is bounded away and above the standardized mean of zero by the first month following the positive increase to economic policy uncertainty.
At one month after the positive shock to economic policy uncertainty, the level of search interest about condoms is about 0.125 standard deviations above the standardized mean. Two months following the shock to economic policy uncertainty, the level of search interest about condoms increases again to at about 0.15 standard deviations above the standardized mean. The level of search interest about condoms continues to be bounded away from the standardized mean for additional months, and then begins to show the signs of decay at around twelve months following the positive shock to economic policy uncertainty. For the entire twelve months though following the positive shock to economic policy uncertainty, the level of condom search interest is significantly above the standardized mean. An increase in economic policy uncertainty results in heightened search interest about condoms that persists for months.
The same cannot be said when looking at the response of economic policy uncertainty to an increase in the level of condom search interest. As seen in row one, column two of Figure 2, a shock (increase) in the level of search interest for condoms has no marked effect on the level of economic policy uncertainty across time that is clearly bounded away from the standardized mean. Since the 95% confidence interval is never significantly away from the standardized mean in any meaningful way, it is not possible to suggest during the time period studied that there is statistically significant evidence that an increase in search interest for condoms can produce a positive or negative shift in economic policy uncertainty. As stated before, this result is to be expected.
Changing levels of search interest about condoms should not be a meaningful signal to warrant shifts in economic policy uncertainty. It is not logical for the search query volume from the public about condoms to be followed closely when assessing the extent of economic policy uncertainty in the country. On the flip side, changing levels of economic policy uncertainty should be a meaningful signal that can help to bring about shifts in the level of search interest the public exhibits about condoms.
As work from scholars like Baker, Bloom & Davis (2013), Julio & Yook (2013), and Colombo (2013) indicate, firms are more risk-averse in terms of their behavior in the face of economic policy uncertainty. This current project is an initial effort at demonstrating that economic policy uncertainty can also make the general public more risk-averse about the economic climate, such that the uncertainty regarding the economic climate heightens interest in information about topics or items that can help to lower perceived risks. People want to minimize potential personal losses (Fischhoff & Kadvany, 2011) from engaging in unprotected sex. An initial step that can be taken to advance this is through seeking out information about condoms. This project is not in any way claiming there is evidence that economic policy uncertainty will increase condom purchases and/or condom usage. What is being suggested is that uncertainty about the consequences of economic policy on the personal situation of citizens increases the level of information seeking conducted about an item that can help to lower the perceived risks of engaging in unprotected sexual intercourse. The empirical results of the project provide an indication that such a proposal has some level of credibility, such that additional study in the future is warranted.
Baker, Bloom & Davis (2013) have provided a major service in devising a dynamic, empirically testable measure of economic policy uncertainty. While most of the research engaged with the concept of economic policy uncertainty is at the business firm level, this project explores the implications of economic policy uncertainty on the behavior of the mass public. The interest is in whether heightened economic policy uncertainty leads members of the public to be more mindful of the potential costs that can be incurred through engaging in unprotected sexual intercourse, such that we observe an increase in the search for information about condoms on the Internet. The empirical analyses performed here, using information about search volume interest reported by Google Trends, lend support to the hypothesis that an increase in economic policy uncertainty can result in an increased level of search interest about condoms.
It is essential to see whether this result is observable in other contexts. In other countries, will we see a similar relationship between economic policy uncertainty and Internet search interest levels on condoms? Unlike the economic policy uncertainty measure for the United States developed by Baker, Bloom & Davis (2013), the economic policy uncertainty measures developed for countries like France, Germany, Italy, Spain, India, Canada, and China are based only on the content analysis of newspaper coverage of economic policy uncertainty in those respective countries. Nonetheless, it will be worthwhile to contrast the economic policy uncertainty measure in these individual countries with the Google Trends information regarding condom search interest volume within these countries. The purpose of this is to show how consistently economic policy uncertainty can shape search interest levels about condoms.
Another stream of research can explore whether economic policy uncertainty alters the search interest volume for other topics/items pertaining to sexual health, mental health, and general physical health. If people use the Google search engine to find out information about topics of interest to them (Gunn & Lester, 2013), such that search interest levels are an indication of their concerns (Ettredge, Gerdes & Karuga, 2005), it will be interesting to see how extensively changes in economic policy uncertainty can alter search interest levels pertaining to health. If there is a lack of clarity regarding the current and prospective economic policy in the country, does this produce an increase in search interest in various health concerns? Addressing this question will help to make major headway in learning about how the mass public engages with the Internet as a result of particular conditions in the sociopolitical environment.
The author declares there are no competing interests.
Abma, J. C., Chandra, A., Mosher, W. D., Peterson, L. S., and Piccinino, L. J. (1997), "Fertility, Family Planning, and Women's Health: New Data from the 1995 National Survey of Family Growth," Vital and Health Statistics. Series 23, Data from the National Survey of Family Growth 19: 1-114.
Althouse, B. M., Allem, J.-P., Childers, M. A., Dredze, M., and Ayers, J. W. (2014), "Population Health Concerns during the United States Great Recession," American Journal of Preventative Medicine 46(2): 166-170.
Askitas, N., and Zimmerman, K. F. (2009), "Google Econometrics and Unemployment Forecasting," Applied Economics Quarterly 55(2): 107-120.
Ayers, J. W., Althouse, B. M., Allem, J.-P., Childers, M. A., Zafar, W., Latkin, C., Ribisl, K. M., and Brownstein, J. S. (2012), "Novel Surveillance of Psychological Distress during the Great Recession," Journal of Affective Disorders 142(1/3): 323-330.
Baker, S. R., Bloom, N., and Davis, S. J. (2013), "Measuring Economic Policy Uncertainty." Available at http://www.policyuncertainty.com/ (accessed April 10, 2014).
Berger, M., Wagner, T. H., and Baker, L. C. (2005), "Internet Use and Stigmatized Illness," Social Science and Medicine 61(8): 1821-1827.
Brownstein, J. S., Freifeld, C. C., and Madoff, L. C. (2009), "Digital Disease Detection--Harnessing the Web for Public Health Surveillance," New England Journal of Medicine 360(21): 2153-2157.
Choi, H., and Varian, H. (2012), "Predicting the Present with Google Trends," Economic Record 88(s1): 2-9.
Colombo, V. (2013), "Economic Policy Uncertainty in the U.S.: Does It Matter for the Euro Area?" Economics Letters 121(1): 39-42.
Dickey, D., and Fuller, W. A. (1979), "Distribution of the Estimate for Autoregressive Time Series with a Unit Root," Journal of the American Statistical Association 74(366a): 427-431.
Dzielinski, M. (2012), "Measuring Economic Uncertainty and Its Impact on the Stock Market," Finance Research Letters 9(3): 167-175.
Enders, W. (1996), RATS Handbook for Econometric Time Series. New York, NY: John Wiley & Sons.
Engle III, R. F., and Granger C. W. J. (1987), "Co-integration and Error Correction: Representation, Estimation and Testing," Econometrica 55(2): 251-276.
Ettredge, M., Gerdes, J., and Karuga, G. (2005), "Using Web-based Search Data to Predict Macroeconomic Statistics," Communications of the ACM 48(11): 87-92.
Experian Hitwise (2014), "Consumer Search Engine Trends." Available at http:// www.experian.com/marketing-services/online-trends-search-engine.html (accessed July 13, 2014).
Eysenbach, G. (2006), "Infodemiology: Tracking Flu-related Searches on the Web Syndrome Surveillance," AMIA Annual Symposium Proceedings/AMIA Symposium, 244-248.
Eysenbach G. (2011), "Infodemiology and Infoveillance Tracking Online Health Information and Cyberbehavior for Public Health," American Journal of Preventative Medicine 40(5S2): S154-S158.
Federal Reserve Bank of St. Louis (2014), "FRED Economic Data." Available at http://research.stlouisfed.org/fred2/ (accessed April 20, 2014).
Fischhoff, B., and Kadvany, J. (2011), Risk: A Very Short Introduction. Oxford: Oxford University Press.
Freeman, J. R., Williams, J. T., and Lin, T. (1989), "Vector Autoregression and the Study of Politics," American Journal of Political Science 33(4): 842-77.
Frijters, P., Johnston, D. W., Lordan, G., and Shields, M. A. (2013), "Exploring the Relationship between Macroeconomic Conditions and Problem Drinking as Captured by Google Searches in the US," Social Science & Medicine 84(May): 6168.
Frost, J. J., and Lindberg, L. D. (2013), "Reasons for Using Contraception: Perspectives of US Women Seeking Care at Specialized Family Planning Clinics," Contraception 87(4): 465-472.
Goel, S., Hofman, J. M., Lahaie, S., Pennock, D. M., and Watts, D. J. (2010), "Predicting Consumer Behavior with Web Search," Proceedings of the National Academy of Sciences 107(41): 17486-17490.
Google (2014a), "Google Trends." Available at https://www.google.com/trends/ explore (accessed June 23, 2014).
Google (2014b), "About Trends Graphs." Available at https://support.google.com/ trends/answer/4355164?hl=en&ref_topic=4365531 (accessed May 3, 2014).
Granger, C. W. J. (1969), "Investigating Causal Relations by Econometric Models and Cross-spectral Models," Econometrica 37(3): 424-438.
Gunn III, J. F., and Lester, D. (2013), "Using Google Searches on the Internet to Monitor Suicidal Behavior," Journal of Affective Disorders 148(2/3): 411-412.
Henshaw, S. K. (1998), "Unintended Pregnancy in the United States," Family Planning Perspectives 30(1): 24-46.
Julio, B., and Yook, Y. (2014), "Policy Uncertainty, Irreversibility, and Cross-border Flows of Capital." Available at http://ssrn.com/abstract=2024612 (accessed April 25, 2014).
Kang, W., Lee, K., and Ratti, R. A. (2014), "Economic Policy Uncertainty and Firm-level Investment," Journal of Macroeconomics 39(A):42-53.
Kang, W., and Ratti, R. A. (2013), "Structural Oil Price Shocks and Policy Uncertainty," Economic Modelling 35(September): 314-319.
Kennedy, P. (2003), A Guide to Econometrics. 5th edn. Cambridge, MA: MIT Press.
Kulczycki, A. (2004), "The Sociocultural Context of Condom Use within Marriage in Rural Lebanon," Studies in Family Planning 35(4): 246-260.
Lam-Po-Tang, J., and McKay, D. (2010), "Dr. Google, MD: A Survey of Mental Health-related Internet Use in a Private Practice Sample," Australasian Psychiatry 18(2): 130-133.
Lemieux, J., and Peterson, R. A. (2011), "Purchase Deadline as a Moderator of the Effects of Price Uncertainty on Search Duration," Journal of Economic Psychology 32(1): 33-44.
McCarthy, M. J. (2010), "Internet Monitoring of Suicide Risk in the Population," Journal of Affective Disorders 122(3): 277-279.
McCombs, M. (2004), Setting the Agenda: The Mass Media and Public Opinion. Cambridge: Polity Press.
Mosher, W. D., and Bachrach, C. A. (1996), "Understanding US Fertility: Continuity and Change in the National Survey of Family Growth, 1988-1995," Family Planning Perspectives 28(1):4-12.
National Bureau of Economic Research (2014), "U.S. Business Cycle Expansions and Contractions." Available at http://www.nber.org/cycles.html (accessed April 13, 2014).
Netmarkshare (2014a), "Desktop Search Engine Market Share." Available at http:// marketshare.hitslink.com/search-engine-marketshare.aspx?qprid=4&qpcustom d=0 &qpsp=185&qpnp=1&qptimeframe=M (accessed July 7, 2014).
Netmarketshare (2014b), "Mobile/Tablet Search Engine Market Share." Available at http://marketshare.hitslink.com/search-engine-market-share.aspx?qprid=4&qp sp=185&qpnp=1&qptimeframe=M&qpcustomd=1 (accessed July 7, 2014).
Pastor, L., and Veronesi, P. (2012), "Uncertainty about Government Policy and Stock Prices," The Journal of Finance 67(4): 1219-1264.
Phillips, P. C., and Perron, P. (1988), "Testing for a Unit Root in Time Series Regression," Biometrika 75(2):335-346.
Popescu, G. (2014), "Economic Aspects Influencing the Rising Costs of Health Care in the United States," American Journal of Medical Research 1(1): 47-52.
Ripberger, J. T. (2011), "Capturing Curiosity: Using Internet Search Trends to Measure Public Attentiveness," Policy Studies Journal 39(2): 239-259.
Scharkow, M., and Vogelgesang, J. (2011), "Measuring the Public Agenda Using Search Engine Queries," International Journal of Public Opinion Research 23(1): 104113.
Schultz, T. P. (1969), "An Economic Model of Family Planning and Fertility," Journal of Political Economy 77(2): 153-180.
Sims, C. A, and Zha, T. (1999), "Error Bands for Impulse Responses," Econometrica 67(5): 1113-1155.
Sueki, H. (2011), "Does the Volume of Internet Searches Using Suicide-related Search Terms Influence the Suicide Death Rate?" Psychiatry and Clinical Neurosciences 65(4): 392-394.
Tafuri, S., Martinelli, D., Germinario, C., and Prato, R. (2010), "Determining Factors for Condom Use: A Survey of Young Italian Adults," The European Journal of Contraception and Reproductive Health Care 15(1): 24-30.
Taylor, J. W. (1974), "The Role of Risk in Consumer Behavior," Journal of Marketing 38(2): 54-60.
Teftt, T. (2011), "Insights on Unemployment, Unemployment Insurance, and Mental Health," Journal of Health Economics 30(2): 258-264.
Tversky, A., and Fox, C. R. (1995), "Weighing Risk and Uncertainty," Psychological Review 102(2): 269-283.
Valdivia, A., and Monge-Corella, S. (2010), "Diseases Tracked by Using Google Trends, Spain," Emerging Infectious Diseases 16(1): 168.
Vanderbilt Television News Archive (2014), "TV-NewsSearch." Available at http:// tvnews.vanderbilt.edu/tvn-search-advanced.pl?SID=20140713885803222&code =(accessed April 24, 2014).
Willard, S. D., and Nguyen, M. M. (2013), "Internet Search Trends Analysis Tools Can Provide Real-time Data on Kidney Stone Disease in the United States," Urology 81(1): 37-42.
Wilson, K., and Brownstein, J. S. (2009), "Early Detection of Disease Outbreaks Using the Internet," CMAJ: Canadian Medical Association Journal (Journal de l'Association Medicale canadienne) 180(8): 829-831.
Wood, B. D. (2009), The Myth of Presidential Representation. Cambridge: Cambridge University Press.
Department of Government and International Affairs, University of South Florida
Caption: Figure 1 Dynamics of Economic Policy Uncertainty and Condoms Search Interest
Caption: Figure 2 Moving Average Representation Dynamics
Table 1 Diagnostic Tests to Assess Variable Stationarity Economic Economic Policy Policy Uncertainty Uncertainty (Augmented (Phillips- Dickey- Perron Fuller Test t-test with 0 with 0 lags) lags) Test Statistic -2.92 -2.94 Significance -2.88 -2.89 Level .05 Critical Value Significance -2.57 -2.58 Level Condom Condom Search Search Interest Interest Level Level (Augmented (Phillips- Dickey- Perron Test Fuller with 0 lags) t-test with 0 lags) Test Statistic -3.30 -3.33 Significance -2.88 -2.89 Level .05 Critical Value Significance -2.57 -2.58 Level .10 Critical Value Note: Tests are evaluated based on a null hypothesis of the presence of a unit root (non-stationarity). Lag length used is selected by finding value with minimum Bayesian Information Criterion (BIC) when choosing optimal lag length between 0 and 20 lags. Diagnostic tests performed using WinRats Version 8. Augmented Dickey-Fuller tests performed using @uradf command. Phillips-Perron tests performed using @ppunit command. Table 2 Granger Tests for Economic Policy Uncertainty and Condom Search System Independent Variable Dependent p-value Variable [F-Statistic] Economic Policy Economic 0.00 Uncertainty Index Policy [62.2234] [right arrow] Uncertainty Index Condoms Search 0.15 Interest Volume [2.1110] Exogenous Controls Unemployment Claims (+, p = 0.00) Real Disposable Personal Income (+, p = 0.00) News Coverage on Condoms (ns, p = 0.86) Elections (ns, p = 0.57) Large Interest Rate Cuts and Stimulus 2008 (+, p = 0.02) Lehman Brothers and TARP (+, p = 0.10) Banking Crisis 2009 (ns, p = 0.21) Debt Ceiling Dispute 2011 (+, p = 0.00) Government Shutdown and Debt Ceiling 2013 (+, p = 0.09) Recession (ns, p = 0.62) Economic Policy Condoms 0.00 Uncertainty Index Search [9.0476] [right arrow] Interest Volume Condoms Search Interest 0.00 Volume [right arrow] [132.3906] Exogenous Controls Unemployment Claims (ns, p = 0.95) Real Disposable Personal Income (-, p = 0.05) News Coverage on Condoms (ns, p = 0.98) Elections (ns, p = 0.61) Large Interest Rate Cuts and Stimulus 2008 (ns, p = 0.84) Lehman Brothers and TARP (ns, p = 0.47) Banking Crisis 2009 (ns, p = 0.46) Debt Ceiling Dispute 2011 (ns, p = 0.36) Government Shutdown and Debt Ceiling 2013 (ns, p = 0.49) Recession (-, p = 0.05) Note: The arrows indicate Granger-causality from the block of coefficients for the independent variable to the dependent variable based on 0.10 significance levels. The p-values are from F-tests for the null hypothesis of no Granger-causality. The system includes a deterministic constant. The results of the exogenous controls are based on t-test results based on 0.10 significance levels. A "+" represents a positive significant relationship, a "-" represents a negative significant relationship, and "ns" represents not significant. Each of the independent variables in the system includes one monthly lag to control for the inertia of the variables. Lag length is selected by Bayesian Information Criterion (BIC). As a result, VAR estimation with lags performed with data between February 2004 and December 2013. Analysis performed using WinRats Version 8.
Please note: Some tables or figures were omitted from this article.
|Printer friendly Cite/link Email Feedback|
|Publication:||American Journal of Medical Research|
|Date:||Oct 1, 2014|
|Previous Article:||Increased medical malpractice expenditures as a main determinant of growth in health care spending.|
|Next Article:||The economic implications of eHealth and mHealth technologies.|