Knight Trust Regressions

Knight Trust in Media, 2022 Panel Survey – regression and multivariate analysis of key questions

This analysis looks at the 2022 Knight Trust in Media survey, fielded to the Gallup Panel (including 5,073 respondents who took the survey via web and 520 who took it via mail, though this latter group should not confused for the 2022 ABS respondents). The main objective is to whether apparent relationships between how a person answers a particular survery question and the demographic background of that individual (as well other personal characteristics or attitudes which can be treated as ‘independent’ of the main response of interest) are statistically signigicant, after controlling for all other potentially salient factors.

In general, the set of indepedent variables – the items on the right-hand side of the regression equation – will remain constant. This will include:

• Political party affiliation (no leaners): This is based off Q37 in the questionnaire, though Q41 in the datafile. The question reads: “In politics, as of today, with which political party fo you most closely affiliate?”. While there are 4 response options one can choose from – Republican, Democrat, Independent and Other party – this analysis will exclude the latter category (i.e. the 226 individuals who said ‘other party’ are dropped for the regression analysis, as are the 101 individuals who did not answer the question at all).
  
• Generation (or age cohort): This variable has 5 groups, moving from the youngest generation of Gen Z, who are aged between 18 to 26; then Millennials (27-41), Gen X (42-57); Baby Boomer (58-76); and the Silent generation, who is 77+. In practice, the bulk of respondents are in either the Milllennial, Gen X or Baby Boomer cohorts.
  
• Race/ethnicity: This is a recoded version of the question, with the groups consisting of Non-Hispanic White, Non-Hispanic Black, Hispanic and Asian. Note the latter group consists of only 108 individuals, and Gallup’s writing team should avoid focusing on this cohort. This variable drops the other category, which had only 34 individuals.
• Gender: While the survey collected gender in manner that allowed people to identify as non-binary, the total number of respondents who did so is too small to include here. Thus, we will be focusing on males and females alone.
  
• Education: This was collected in a very granular manner on the survey (Q33 on the questionnaire). Past analysis has focused on the question of whether a person has a college degree; here a 4-category variable is used (see below).
  
• Urbanicity (Q42. If you had to choose, would you describe the are you live in as…): A recoded version of this variable is being used, as is shown below. There were six response options featured on the survey, and reasonable people could certainly disagree on how the variable was recoded. However, it rarely produces significant variation.
  
• Household income: While Q41 on the survey allowed for 10 different income brackets, this analysis utilizes the “INCOMENEW” variable which was also a crosstab variable. It has five categories.

While this is not necessarily an exhaustive list of every independent variable which appears in each regression analsis, it is the “usual suspects.”

Some recoding was required for these items. The next section looks at that (and simply shows the distribution) before moving on to the regressions.

Recoding and descriptive statistics of independent variables

In this section, variables are recoded to either collapse (or combine) categories, remove the “no answer,” option (given it’s difficulty in intepretation when looking at model results), consider other transformations which might be necessary (including making most variables factor variables) and examine the distribution of these items.

The first variable of interest is highest level of education completed, or Q33 in the questionnaire but Q37 in the datafile (these different names will be a constant theme through this analysis). The original form of this question features 8 categories. Below a new variable is created, called “educr,” which has 4 categories – high school or less; some college or post-college work which does not include a four-year bachelor’s degree;a four-year bachelor’s degree ONLY and, finally, post-grad degree or work. Note these definitions are in line with a standard variable created for Gallup’s U.S.-based GPSS data (i.e. the traditional polling efforts of the company, which feature questions which stretch back many decades). In the code chunk below, the code is shown as well as a comparison of the distribution of the original variable and the new one.

df$educr<-ifelse(df$Q37 <= 2, 1,
                 ifelse(df$Q37 >= 3 & df$Q37 <=5, 2,
                        ifelse(df$Q37 == 6, 3,
                               ifelse((df$Q37 == 7 | df$Q37 ==8), 4, NA_real_))))

df$educr<-factor(df$educr, levels=c(1,2,3,4), labels=c('HS or less', 'Some college', '4-year degree', 'Postgrad work or degree'))



one.factor.var.topline(df, "Q37", "weight")%>%
  mutate(weighted_n=round(weighted_n,0))%>%
  kbl(caption="Table 1. Q33 original: Highest level of education you have completed")%>%
  kable_styling(font_size=16, bootstrap_options = c("striped"))%>%
  kable_classic_2(full_width =F)

Table 1. Q33 original: Highest level of education you have completed

QTAG	Wording	response.option	weighted_percentage	unweighted_n	weighted_n
Q37	What is the highest level of school you have completed, or the highest degree you have received?	Less than a high school diploma (Grades 1 through 11 or no schooling)	4	152	218
Q37	What is the highest level of school you have completed, or the highest degree you have received?	High school graduate (Grade 12 with diploma or GED certificate)	33	1448	1874
Q37	What is the highest level of school you have completed, or the highest degree you have received?	Technical, trade, vocational or business school or program after high school	5	341	268
Q37	What is the highest level of school you have completed, or the highest degree you have received?	Some college – college, university or community college – but no degree	15	996	829
Q37	What is the highest level of school you have completed, or the highest degree you have received?	Two-year associate degree from a college, university, or community college	7	507	413
Q37	What is the highest level of school you have completed, or the highest degree you have received?	Four-year bachelor’s degree from a college or university (e.g., BS, BA, AB)	18	966	1008
Q37	What is the highest level of school you have completed, or the highest degree you have received?	Some postgraduate or professional schooling after graduating college, but no postgraduate degree (e.g., some graduate school)	4	230	222
Q37	What is the highest level of school you have completed, or the highest degree you have received?	Postgraduate or professional degree, including master’s, doctorate, medical or law degree (e.g., MA, MS, PhD, MD, JD)	13	900	712
Q37	What is the highest level of school you have completed, or the highest degree you have received?	No answer	1	53	51

one.factor.var.topline(df, "educr", "weight")%>%
  mutate(weighted_n=round(weighted_n,0))%>%
  kbl(caption="Table 2. Educr: Highest level of education you have completed (recoded)")%>%
  kable_styling(font_size=16, bootstrap_options = c("striped"))%>%
  kable_classic_2(full_width =F)

Table 2. Educr: Highest level of education you have completed (recoded)

QTAG	Wording	response.option	weighted_percentage	unweighted_n	weighted_n
educr	Q33. What is the highest level of scholl you have completed, or the highest degree you received	HS or less	38	1600	2091
educr	Q33. What is the highest level of scholl you have completed, or the highest degree you received	Some college	27	1844	1509
educr	Q33. What is the highest level of scholl you have completed, or the highest degree you received	4-year degree	18	966	1008
educr	Q33. What is the highest level of scholl you have completed, or the highest degree you received	Postgrad work or degree	17	1130	934

Another recode is the final question on the survey (Q42 in the questionnaire and Q46 in the datafile, though an existing recoded version of this item also exists which is called ‘AREALIVE,’ and was used for the crosstabs). The six categories used here are reduced into three items.

Additionally, formatting variable recodes are required, which are specific to R but are nonetheless shown below. A final substantive recode concerns Q37 (in survey). In politics, as of today, with which political party do you most closely affiliate? This item does NOT allocate leaners, as requested by Knight. However, the current variable in the datafile – POLPARTY – includes “other party,” as an option, which this recode removes. This means the main partisan variable (one which will appear in all regression models) has the name of POLPARTY_NEW. The code for this transformation is shown below, as are the distribution of results

df$area.live.new<-ifelse(df$AREALIVE == 1 | df$AREALIVE == 3, 1,
                        ifelse(df$AREALIVE == 2 | df$AREALIVE == 4, 2, 3))


df$area.live.new<-factor(df$area.live.new, levels=c(1,2,3), labels=c("Lives in big city or suburb", "Lives in small city or suburb", "Lives in town or rural area"))




one.factor.var.topline(df, "AREALIVE", "weight")%>%
  mutate(weighted_n=round(weighted_n,0))%>%
  kbl(caption="Table 3. Q42 (original). If you had to choose, would you describe the are you live in as a...")%>%
  kable_styling(font_size=16, bootstrap_options = c("striped"))%>%
  kable_classic_2(full_width =F)

Table 3. Q42 (original). If you had to choose, would you describe the are you live in as a…

QTAG	Wording	response.option	weighted_percentage	unweighted_n	weighted_n
AREALIVE	Area Live	Big city	17	892	931
AREALIVE	Area Live	Small city	19	1063	1083
AREALIVE	Area Live	Suburb of a big city	24	1377	1315
AREALIVE	Area Live	Suburb of a small city	8	476	461
AREALIVE	Area Live	Town	15	818	811
AREALIVE	Area Live	Rural area	17	939	961

one.factor.var.topline(df, "area.live.new", "weight")%>%
  mutate(weighted_n=round(weighted_n,0))%>%
  kbl(caption="Table 4. Q42 (recoded). If you had to choose, would you describe the are you live in as a...")%>%
  kable_styling(font_size=16, bootstrap_options = c("striped"))%>%
  kable_classic_2(full_width =F)

Table 4. Q42 (recoded). If you had to choose, would you describe the are you live in as a…

QTAG	Wording	response.option	weighted_percentage	unweighted_n	weighted_n
area.live.new	Q42. If you had to choose, would you describe the area you live in…	Lives in big city or suburb	40	2269	2246
area.live.new	Q42. If you had to choose, would you describe the area you live in…	Lives in small city or suburb	28	1539	1543
area.live.new	Q42. If you had to choose, would you describe the area you live in…	Lives in town or rural area	32	1757	1772

#####Turning other demographic items into factors

df$GENERATION<-haven::as_factor(df$GENERATION)
df$GENDER<-haven::as_factor(df$GENDER)
df$RACENEW<-haven::as_factor(df$RACENEW)
df$POLPARTY<-haven::as_factor(df$POLPARTY)
df$INCOMENEW<-haven::as_factor(df$INCOMENEW)
df$Q44AR<-haven::as_factor(df$Q44AR)
df$Q44BR<-haven::as_factor(df$Q44BR)
df$Q44CR<-haven::as_factor(df$Q44CR)
df$Partyr<-haven::as_factor(df$Partyr)
#####Recoding of POLPARTY



df$POLPARTY_NEW<-ifelse(df$POLPARTY != "Other party", df$POLPARTY, NA_real_)

df$POLPARTY_NEW<-factor(df$POLPARTY_NEW, levels=c(1,2,3), labels=c("Republican (no lean)", "Democrat (no lean)", "Independent (includes leaners)"))

label_df<-label_df%>%
  add_row(QTAG="POLPARTY_NEW", Wording="Q37. In politicas, as of today, which political party do you most closely affiliate (drops other party)")


#####Distribution of results



one.factor.var.topline(df, "POLPARTY", "weight")%>%
  mutate(weighted_n=round(weighted_n,0))%>%
  kbl(caption="Table 5. Q41. Party affiliation, no leaners -- original version")%>%
  kable_styling(font_size=16, bootstrap_options = c("striped"))%>%
  kable_classic_2(full_width =F)

Table 5. Q41. Party affiliation, no leaners – original version

QTAG	Wording	response.option	weighted_percentage	unweighted_n	weighted_n
POLPARTY	Political party	Republican	27	1462	1457
POLPARTY	Political party	Democrat	40	2229	2191
POLPARTY	Political party	Independent	29	1575	1576
POLPARTY	Political party	Other party	5	226	257

one.factor.var.topline(df, "POLPARTY_NEW", "weight")%>%
  mutate(weighted_n=round(weighted_n,0))%>%
  kbl(caption="Table 6. Q41. Party affiliation, no leaners -- exclude 'other party'")%>%
  kable_styling(font_size=16, bootstrap_options = c("striped"))%>%
  kable_classic_2(full_width =F)

Table 6. Q41. Party affiliation, no leaners – exclude ‘other party’

QTAG	Wording	response.option	weighted_percentage	unweighted_n	weighted_n
POLPARTY_NEW	Q37. In politicas, as of today, which political party do you most closely affiliate (drops other party)	Republican (no lean)	28	1462	1457
POLPARTY_NEW	Q37. In politicas, as of today, which political party do you most closely affiliate (drops other party)	Democrat (no lean)	42	2229	2191
POLPARTY_NEW	Q37. In politicas, as of today, which political party do you most closely affiliate (drops other party)	Independent (includes leaners)	30	1575	1576

We now look at the weighted distribution for all of the “usual suspect,” independent variables in the below table.

regression.independent.vars<-c("GENERATION", "GENDER", "RACENEW", "POLPARTY_NEW", "educr", "INCOMENEW", "area.live.new")

independent.variable.freqs<-two_tab_long_looper_dep(df, "total", regression.independent.vars, "weight")%>%
  mutate(pct=round(pct,0))%>%
  dplyr::select(-ind_var, -ind_category, -unweighted_n)%>%
  rename(weighted.pct=pct)


DT::datatable(independent.variable.freqs,  class = 'cell-border stripe',
              caption="Table 7. Distribution of all 'usual suspect' independent variables after recoding")

A final note on the independent variables used for the regression modelling. There are a set of three questions which ask an individual how often that person votes in a) federal elections; b) state elections and c) local elections. This is based off a five-point scale, with (in the data file), the number 1 representing “never,” and 5 representing “always.”

Given Knight’s interest in tying the data to the election, these could be useful instruments in that regard – though the crosstab analysis did not show promising results in terms of the relationship of these voting items against the core survey questions. Still, an analysis of the three items (see directly below) revealed two potentially interesting and even useful (from a modelling perspective) findings:

• The three items (using the 5-point scale mentioned above as a ‘continuous’ measurement scale) demonstrate strong inter-item correlation, though the three are not in perfect agreement. In particular, the correlation between how an individual rated that person’s frequency of voting in federal election was different from local elections (Pearson’s R of 0.69 between the two, as seen below).
  
• Building off these reasonably high correlations, the three items could be easily transformed into a scale – a variable which simply takes the average of the three responses. The Cronbach’s alpha score (a measure of reliability) is shown below; it stands at 0.915, which indicates “excellent,” reliability (according to a textbook). If these items were to be incorporated in the analysis, this might be how. However, this does not seem like a critical exercise for the purposes of this regression-based analysis.

######First specify a survey design which will incorporate weights for the analysis.

vote.df.design<-df%>%
  dplyr::select(vote.federal, vote.state, vote.local, weight)%>%
  srvyr::as_survey_design(id=1, weight=weight)


####Inter-item correlations
jtools::svycor(~vote.federal+vote.state+vote.local, na.rm=TRUE, design=vote.df.design)

##              vote.federal vote.state vote.local
## vote.federal         1.00       0.84       0.69
## vote.state           0.84       1.00       0.84
## vote.local           0.69       0.84       1.00

#########Cronbach's alpha score for the three questions about frequency of voting

survey::svycralpha(~vote.federal+vote.state+vote.local, na.rm=TRUE, design=vote.df.design)

##  *alpha* 
## 0.914965

Identifying the salient, significant independent variables on core survey items: Regression analysis

We will now move into the regression analysis. In general, the main form of regression analysis will be a logistic regression which will attempt to ‘predict’ an outcome of interest. This typically requires changing a dependent variable which has multiple response options into one that has only two; all such changes are explained below. Other forms of regression are also used, though with less frequency.

The flow of this document essentially follows the questionnaire, rather than the groupings the Knight analysis and reporting team has used in other settings.

For the logistic regressions, the svyglm function in the survey package will be the main analytical tool. For ease of analysis, a function is first created that can take in a dependent variable and a vector of independent variables and return the results from the analysis. It is not critical to follow the code below, but it is shown for reference.

############Code to run logistic regression

survey_logistic <- function(data, dep, covs) {
  ###NOTE WEIGHT VARIABLE IS ALREADY PASSED 
  
  all.vars<-c(dep, covs, "weight")
  
  data.design<-data%>%
    dplyr::select(all_of(all.vars))
  
  #####Create survey design consisting only of our variables
  
  data.design<-data.design%>%
    srvyr::as_survey_design(ids=1, weight = weight)
  
  form_base <- paste(dep, "~")
  # Create a string that concatenates your covs vector with a "+" between each variable
  form_vars <- paste(covs, collapse = " + ")
  # Paste the two parts together
  formula <- paste(form_base, form_vars)
  
  # Call the lm function on your formula
  logit.results<-svyglm(formula, design = data.design, family=quasibinomial, na.action=na.omit)
  return(logit.results)
}

Q1/K1. For information to be considered true, that information needs to be…

The initial question item on the survey asks, “for information to be considered true, that information needs to be…,” – ‘mostly supported by data and evidence,’ ‘mostly based on personal experience or belief,’ ‘both above equally,’ or ‘none of the above.’

As Table 8 below shows, nearly seven-in-ten Americans (68%) said true information must be “mostly supported by data and evidence.” The response option which represents the precise opposite opinion – that true information needs to be “mostly based on personal experience or belief” – is supported by 3%. Another 25% said both.

Given this distribution, the item was recoded in this manner: if a person said “mostly supported by data and evidence,” ONLY (i.e. not both), that person is coded as a 1. All other responses are coded as 0 (however the ‘no answer’ is not included in this analysis). The logistic regression, then, will predict whether an individual said “mostly supported by data and evidence,” positive coefficients are associated with greater odds of saying this, holding all else equal.

First, the variable is recoded. It was verified to have been recoded correctly.

var_label(df$K1)

## [1] "For information to be considered true, that information needs to be..."

one.factor.var.topline(df, "K1", "weight")%>%
  mutate(weighted_n=round(weighted_n,0))%>%
  kbl(caption="Table 8. Q1. For information to be considered true, that information needs to be...")%>%
  kable_styling(font_size=16)%>%
  kable_material_dark(full_width =T)

Table 8. Q1. For information to be considered true, that information needs to be…

QTAG	Wording	response.option	weighted_percentage	unweighted_n	weighted_n
K1	For information to be considered true, that information needs to be…	Mostly supported by data and evidence	68	3917	3829
K1	For information to be considered true, that information needs to be…	Mostly based on personal experience or belief	3	119	148
K1	For information to be considered true, that information needs to be…	Both of the above equally	25	1369	1413
K1	For information to be considered true, that information needs to be…	None of the above	2	116	119
K1	For information to be considered true, that information needs to be…	No answer	1	72	84

####Recode into binary, such that 1=mostly supported by data and evidence ONLY

df$Q1.binary<-ifelse(df$K1 == 1, 1,
                     ifelse(df$K1 == 98, NA_real_, 0))

###Checking recode counts match original variable

table(df$Q1.binary)

## 
##    0    1 
## 1604 3917

With these transformations complete, the regression analysis can now take place. Again, the aim of this statistical model is to “predict,” those who exclusively said “for information to be considered true, that information needs to be mostly supported by evidence.” Pay attention to the final table in the output below, which shows the exponentiated coefficients related to each variable/response option, or what is better known as the odds ratio. An odds ratio above but without a confidence interval that crosses 1 (and is deemed statistically significant) is a driver of the outcome of interest, here saying “mostly supported by data and evidence.” Odds ratio values below 1, indicate a lower likelihood, which is often expressed in percent terms. If, for instance, the odds ratio coefficient is .80, one can subtract 1- .8 =.2, to find that individuals who give this response are 20% less likely than the other comparison groups to voice the response option of interest.

The regression output appears below. Admittedly, the model fit statistics are low, though this is not unusual for a logistic regression model (particularly since all of our independent variables are themselves factors or categorical variables). However, please keep in mind, the goal of this exercise is not to develop a regression model that captures every potentially salient characteristic or attitude to better ‘predict’ who will say that for information to be true it needs to be supported “mostly by data and evidence,” but simply to establish which key demographics have a statsitcally significant effect, controlling for everything else.

Significant factors include:

• Generation: The reference category is Gen Z. The odds ratio (or “exp(Est.)” in some tables, but this is the same thing) is below 1 for all other groups, though it is not significant with respect to Millennials (notably the confidence interval passes through 1). (Admittedly this difference between Gen Z and Millennials is hard to see in the crosstabs, so it may be best to focus on Gen Z/Millennials compared to the other groups) For all other generations, though, they are statistically significant to provide this response.
  
• Race (variable name RACENEW): The reference category is white. Non-Hispanic Blacks are less likely to provide this response – 58% less likely, to be precise. This is statistically significant. Hispanics are also less likely to provide this response (OR=0.78, or 22% less likely than whites), though the effect size is smaller. Results for Asians are not significant.
  
• Political Party: **The reference category is Republican without leaners*. Democrats are 1.95 times more likely to say this, and this is significant. Even independents WITH leaners are more likely, though the confidence interval flirts with the dreaded “1” line, which signifies no meaningful effect. Note: if one re-runs this model and instead allocates the leaners (modified form of Partyr, such that we have three groups – Republican/lean Republican, Democrat/lean Democrat and “true” independents, then the coefficient associated with independents is no longer statistically significant, however the odds ratio for Democrats rises to 2.11).
  
• Gender: The reference category is men. Women are 29% less likely to say this (OR=0.71). The effect is significant.
  
• Education: The reference category is having a high school education or less. This is arguably the single most important predictor for this question item. As education level grows, so does the likelihood of providing the response "information needs to be most supported by evidence. All levels are significant.
• Income (INCOMENEW): The reference category is household income below $30,000. This is statistically significant. As income rises, so too does the likelihood of giving the response of interest, especially for the top income bracket (OR=2.30).
• Urbanicity (area.live.new): The reference category is living in big city or suburb. This was not significant.

The output is below. **You can also find the coefficients here: "X:_Foundation_data\2022working analysis_q1_to_be_true_most_be_supported_by_evidence.csv"**

Observations	4909
Dependent variable	Q1.binary
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.16
Pseudo-R² (McFadden)	0.10
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	1.75	1.05	2.91	2.14	0.03
GENERATIONMillennial (ages 27-41)	0.70	0.44	1.10	-1.55	0.12
GENERATIONGen X (ages 42-57)	0.60	0.37	0.95	-2.17	0.03
GENERATIONBaby Boomer (ages 58-76)	0.51	0.32	0.81	-2.87	0.00
GENERATIONSilent (ages 77+)	0.51	0.30	0.87	-2.49	0.01
GENDERFemale	0.71	0.61	0.83	-4.23	0.00
RACENEWNon-Hispanic Black	0.42	0.33	0.52	-7.57	0.00
RACENEWHispanic	0.78	0.61	0.99	-2.03	0.04
RACENEWAsian	0.65	0.36	1.14	-1.50	0.13
POLPARTY_NEWDemocrat (no lean)	1.95	1.59	2.39	6.46	0.00
POLPARTY_NEWIndependent (includes leaners)	1.27	1.04	1.56	2.34	0.02
educrSome college	1.49	1.24	1.80	4.20	0.00
educr4-year degree	2.23	1.71	2.91	5.96	0.00
educrPostgrad work or degree	2.43	1.87	3.15	6.68	0.00
INCOMENEW$30,000-$49,999 per year	1.41	1.09	1.83	2.59	0.01
INCOMENEW$50,000 - $99,999 per year	1.54	1.22	1.96	3.58	0.00
INCOMENEW$100,000-$149,999 per year	1.69	1.28	2.23	3.68	0.00
INCOMENEW$150,000 or more per year	2.30	1.69	3.11	5.35	0.00
area.live.newLives in small city or suburb	1.13	0.93	1.36	1.25	0.21
area.live.newLives in town or rural area	1.15	0.95	1.39	1.44	0.15
Standard errors: Robust

We can also plot out the coefficients to see the differences in effect sizes more clearly. Again, this is the odds ratio being plotted, where a value greater than 1 indicates the group is more likely than the reference category to say true information must be mostly supported by evidence and data.

To summarize, our significant main effects include (in semi-order of importance, but no official test of variable importance was performed – this is just assessing effect sizes): education, household income, political affiliation, race, generation, gender. Below we see the crosstabs for these items for the original version of the question. Note this table is interactive when accessed via HTML.

sig.vars<-c("educr", "INCOMENEW", "POLPARTY_NEW", "RACENEW", "GENERATION", "GENDER")


q1.by.sig.vars<-two_tab_long_looper_ind(df, sig.vars, "K1", "weight")%>%
  mutate(pct=round(pct,0))%>%
  dplyr::select(-ind_var, -dep_var)%>%
  mutate(QTAG="Q1")%>%
  relocate(QTAG, .before=Wording)%>%
  pivot_wider(names_from=dep_category, values_from=pct)


DT::datatable(q1.by.sig.vars,  class = 'cell-border stripe', caption="Table 9. Q1. For information to be considered true, it must be..., by significant variables")

Q2A-C (P12_Q1A-C in data file): How much attention are you currently paying to the news? a) Local b) National c) International

All three items are asked on a four-point scale, but Gallup’s analysis on the topline document largely focused on the “great deal,” response where there was interesting variation. With this in mind, these variables are recoded to be binary, where 1= “a great deal,” and 0= all other responses, except “no answer.” Granted, this may be controversial, and we can try diferent approaches if desired.

This section will attempt to move more briskly than the one before it. Please refer to Q1/K1 if you have not to get a sense of the general flow of the analysis, variables being tested and output to expect.

Step 1: Recode the items into binary variables, and then check to see that the recodes were performed correctly. Unlike above, this will not include the code used for the recodes (though it is available in this folder: X: Knight_Foundation MR Final_data 2022 Andrew working analysis).

QTAG	Wording	response.option	weighted_percentage	unweighted_n	weighted_n
P12_Q1A	How much attention are you currently paying to each of the following? Local news	A great deal	18	991	1002
P12_Q1A	How much attention are you currently paying to each of the following? Local news	A moderate amount	47	2654	2620
P12_Q1A	How much attention are you currently paying to each of the following? Local news	Not much	29	1626	1631
P12_Q1A	How much attention are you currently paying to each of the following? Local news	None at all	5	299	307
P12_Q1A	How much attention are you currently paying to each of the following? Local news	No answer	1	23	32

QTAG	Wording	response.option	weighted_percentage	unweighted_n	weighted_n
Q2A.top1	Q2A How much attention currently paying to LOCAL NEWS (great deal)	0	82	4579	4559
Q2A.top1	Q2A How much attention currently paying to LOCAL NEWS (great deal)	1	18	991	1002

QTAG	Wording	response.option	weighted_percentage	unweighted_n	weighted_n
P12_Q1B	How much attention are you currently paying to each of the following? National news	A great deal	32	1887	1786
P12_Q1B	How much attention are you currently paying to each of the following? National news	A moderate amount	44	2450	2487
P12_Q1B	How much attention are you currently paying to each of the following? National news	Not much	19	999	1038
P12_Q1B	How much attention are you currently paying to each of the following? National news	None at all	4	219	230
P12_Q1B	How much attention are you currently paying to each of the following? National news	No answer	1	38	52

QTAG	Wording	response.option	weighted_percentage	unweighted_n	weighted_n
Q2B.top1	Q2B How much attention currently paying to NATIONAL NEWS (great deal)	0	68	3668	3755
Q2B.top1	Q2B How much attention currently paying to NATIONAL NEWS (great deal)	1	32	1887	1786

QTAG	Wording	response.option	weighted_percentage	unweighted_n	weighted_n
P12_Q1C	How much attention are you currently paying to each of the following? International news	A great deal	17	1003	960
P12_Q1C	How much attention are you currently paying to each of the following? International news	A moderate amount	46	2643	2595
P12_Q1C	How much attention are you currently paying to each of the following? International news	Not much	29	1577	1621
P12_Q1C	How much attention are you currently paying to each of the following? International news	None at all	6	318	349
P12_Q1C	How much attention are you currently paying to each of the following? International news	No answer	1	52	68

QTAG	Wording	response.option	weighted_percentage	unweighted_n	weighted_n
Q2C.top1	Q2C How much attention currently paying to INTERNATIONAL NEWS (great deal)	0	83	4538	4565
Q2C.top1	Q2C How much attention currently paying to INTERNATIONAL NEWS (great deal)	1	17	1003	960

We have confirmed that recoding is successful. We now move into the different items, taking them one at a time. Please note all output can also be found in the same folder flagged above.

Q2A: Attention paid to local news (great deal or not)

Overall, 18% of Americans are paying “a great deal of attention,” to local news. This analysis looks at the standard battery of demographic characteristics and other background information of respondents to see which types of people are comparitively more (or less) likely to pay “a great deal,” of attention to local news.

The name of the file which has the regression output is knight_q2a_great_deal_regression_output.csv.

The findings of note are (note: at this point, I will only review those demographics or characteristics which are statistically significant or otherwise merit a discussion. The first regression analysis above walks through all of the variables if you need a recap on the variables included in the analysis, they can also be found in the CSV file mentioned above).

• Generation: Again, the reference category is ‘Gen Z.’ This may be the most influential of all the variables examined here, with older cohorts substantially and significantly more likely to pay a great deal of attention to the news than the youngest cohort – though the odds ratio dips slightly with respect to the smaller, older Silent Generation. Still, it’s fair to characterize having a “great deal of interest,” in local news as rising with age. Note, this is a finding that is fairly apparent in the toplines – as6% of Gen Z says they have a great deal of interest in the news, compared to 12% of Millennials, 19% of Gen Z, 25% of Baby Boomers and 27% of the Silent generation.
  
• Racial identity: This is statistically significant, even with political party included in the analysis. Again, the reference category is white. Non-Hispanic Blacks are 3.66 times more likely to have a great deal of interest in the local news than whites. However, no other effect is significant with respect to race/ethnicity.
• Political Party: Democrats are 1.53 times more likely than Republicans (both groups do not include leaners) to have a great deal of interest in local news; this is significant. However, there is not a significant effect with respect to independents.
• Income: This is a strange relationship – it appears interest declines with higher income. This is supported by the crosstabs as well (27% of those earning less than $30,000 pay a great deal of interest, a rate which declines to 14% for the top two brackets). It is difficult to offer a theory as to why this is.

Observations	4942
Dependent variable	Q2A.top1
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.12
Pseudo-R² (McFadden)	0.08
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	0.09	0.04	0.17	-7.19	0.00
GENERATIONMillennial (ages 27-41)	2.07	1.13	3.81	2.35	0.02
GENERATIONGen X (ages 42-57)	3.89	2.12	7.12	4.40	0.00
GENERATIONBaby Boomer (ages 58-76)	5.21	2.86	9.50	5.38	0.00
GENERATIONSilent (ages 77+)	5.08	2.63	9.81	4.84	0.00
GENDERFemale	1.00	0.84	1.20	0.02	0.98
RACENEWNon-Hispanic Black	1.55	1.23	1.96	3.66	0.00
RACENEWHispanic	0.99	0.75	1.32	-0.04	0.97
RACENEWAsian	0.80	0.37	1.73	-0.58	0.56
POLPARTY_NEWDemocrat (no lean)	1.53	1.22	1.92	3.68	0.00
POLPARTY_NEWIndependent (includes leaners)	0.86	0.67	1.11	-1.16	0.25
educrSome college	0.86	0.69	1.06	-1.40	0.16
educr4-year degree	0.81	0.60	1.10	-1.33	0.18
educrPostgrad work or degree	1.06	0.81	1.40	0.44	0.66
INCOMENEW$30,000-$49,999 per year	0.64	0.48	0.85	-3.09	0.00
INCOMENEW$50,000 - $99,999 per year	0.57	0.44	0.74	-4.18	0.00
INCOMENEW$100,000-$149,999 per year	0.53	0.39	0.73	-3.98	0.00
INCOMENEW$150,000 or more per year	0.56	0.40	0.79	-3.36	0.00
area.live.newLives in small city or suburb	0.93	0.75	1.15	-0.69	0.49
area.live.newLives in town or rural area	0.99	0.80	1.24	-0.07	0.95
Standard errors: Robust

Keeping in mind a number of variables did not have significant effects, here is a plot of the coefficients for this question item.

Q2B: Attention paid to national news

More Americans pay attention to national news than local news, at least when focusing on who is paying " a great deal of attention." Nearly a third of U.S. adults (32%) pay “a great deal of attention,” to national news.

The trends for this item are similar to the item before, but not exact. These include:

• Generation Though the effect isn’t as strong with local news, there again is a tendency for older age cohorts to be more likely to report paying “a great deal,” of attention to the news.
  
• Racial identity Is NOT significant (just flagging to be aware).
• Gender Women are about 23% less likely to have a great deal of interest in national news. In the crosstabs, though, the difference is marginal (but probably statistically significant, though I did not perform a t-test on this), at 34% for men and 30% for women.
  
• Political Party Again, the reference category is Republicans, no leaners. Democrats (no leaners) are once again more likely to pay attention to this type of news than their counterparts in the GOP, 1.85 times more likely to be precise. There is no signicant effect for independents.
• Education This is significant here, whereas that was not the case with local news. Higher levels of education are associated with a greater likelihood to pay “a great deal of attention,” to national news. Note this is not obvious when looking at the crosstabs, and is really only noticeable among post graduates (37% say great deal of interest). So while this is significant, it would be a struggle to explain in the report.

Observations	4932
Dependent variable	Q2B.top1
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.14
Pseudo-R² (McFadden)	0.09
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	0.22	0.14	0.35	-6.28	0.00
GENERATIONMillennial (ages 27-41)	0.88	0.59	1.31	-0.64	0.52
GENERATIONGen X (ages 42-57)	1.68	1.12	2.52	2.50	0.01
GENERATIONBaby Boomer (ages 58-76)	3.01	2.00	4.52	5.30	0.00
GENERATIONSilent (ages 77+)	4.39	2.73	7.07	6.10	0.00
GENDERFemale	0.77	0.67	0.89	-3.50	0.00
RACENEWNon-Hispanic Black	0.86	0.68	1.08	-1.31	0.19
RACENEWHispanic	0.83	0.65	1.05	-1.54	0.12
RACENEWAsian	0.78	0.46	1.32	-0.93	0.35
POLPARTY_NEWDemocrat (no lean)	1.85	1.54	2.24	6.44	0.00
POLPARTY_NEWIndependent (includes leaners)	0.83	0.69	1.02	-1.80	0.07
educrSome college	1.36	1.13	1.63	3.21	0.00
educr4-year degree	1.50	1.17	1.91	3.22	0.00
educrPostgrad work or degree	1.71	1.36	2.14	4.67	0.00
INCOMENEW$30,000-$49,999 per year	0.94	0.72	1.21	-0.50	0.62
INCOMENEW$50,000 - $99,999 per year	0.94	0.75	1.19	-0.49	0.62
INCOMENEW$100,000-$149,999 per year	0.96	0.73	1.25	-0.32	0.75
INCOMENEW$150,000 or more per year	1.09	0.83	1.43	0.60	0.55
area.live.newLives in small city or suburb	0.96	0.81	1.14	-0.49	0.62
area.live.newLives in town or rural area	1.07	0.89	1.28	0.73	0.46
Standard errors: Robust

We will skip the effect plots (if anyone wants it, please reach out).

International News: Great Deal

Overall, 17% of Americans say they pay “a great deal,” of attention to the news, putting it on par with local news. Gallup’s topline analysis has noted that this figure may be higher than previous surveys due to the abundance of gripping foreign affair issues, especially the war in Ukraine. Obviously, the regression analysis is not able to test this very reasonable hypothesis.

Fewer variables “popped” as being significant predictors of having “a great deal of interest” in the news. These include:

• Generation A familiar predictor, but the relationship is much weaker with this item than the other two news items. Only Baby Boomers and the Silent Generation have a significant positive odds ratio (1.69 and 2.21, respectively), and this can be seen in the crosstabs.
  
• Gender A more dramatic relationship here, as women are now 40% less likely than men to have “a great deal of interest in national news,” corresponding to an odds ratio of 0.60 for women. The crosstabs help illustrate this difference – 21% of men say they pay a great deal of attention, compared to 14% of women.
  
• Political Party Democrats again are more likely to pay “a great deal” of attention than Republicans, with the odds ratio registering 1.66.
• Education People with postgraduate degrees or work are 1.75 times more likely than those without a high school degree to say they pay a great deal of attention to international news. This can be illustrated in the crosstabs (which admittedly are not in the standard file): 21% of people with a postgrad degree or work have a great deal of interest in the news, compared to 14% for those with a 4-year degree and 17% for both of the other two groups.
• Income The highest income group has a positive, statistically significant odds ratio value.

Observations	4921
Dependent variable	Q2C.top1
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.09
Pseudo-R² (McFadden)	0.06
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	0.15	0.08	0.27	-6.22	0.00
GENERATIONMillennial (ages 27-41)	0.65	0.39	1.07	-1.71	0.09
GENERATIONGen X (ages 42-57)	1.18	0.70	1.97	0.62	0.53
GENERATIONBaby Boomer (ages 58-76)	1.69	1.01	2.85	1.99	0.05
GENERATIONSilent (ages 77+)	2.21	1.23	3.97	2.65	0.01
GENDERFemale	0.60	0.50	0.71	-5.80	0.00
RACENEWNon-Hispanic Black	0.79	0.59	1.05	-1.64	0.10
RACENEWHispanic	0.94	0.71	1.24	-0.46	0.65
RACENEWAsian	0.90	0.49	1.65	-0.35	0.72
POLPARTY_NEWDemocrat (no lean)	1.66	1.30	2.11	4.12	0.00
POLPARTY_NEWIndependent (includes leaners)	1.27	0.99	1.62	1.89	0.06
educrSome college	1.21	0.97	1.51	1.68	0.09
educr4-year degree	1.17	0.86	1.61	1.01	0.31
educrPostgrad work or degree	1.75	1.34	2.29	4.08	0.00
INCOMENEW$30,000-$49,999 per year	0.87	0.63	1.19	-0.89	0.37
INCOMENEW$50,000 - $99,999 per year	0.80	0.60	1.05	-1.59	0.11
INCOMENEW$100,000-$149,999 per year	0.99	0.73	1.34	-0.08	0.94
INCOMENEW$150,000 or more per year	0.81	0.58	1.11	-1.31	0.19
area.live.newLives in small city or suburb	1.23	1.00	1.51	1.99	0.05
area.live.newLives in town or rural area	1.24	1.00	1.54	2.00	0.05
Standard errors: Robust

	Q2C. Great deal of attention to NATIONAL NEWS
(Intercept)	0.15 ***
	[0.08, 0.27]
GENERATIONMillennial (ages 27-41)	0.65
	[0.39, 1.07]
GENERATIONGen X (ages 42-57)	1.18
	[0.70, 1.97]
GENERATIONBaby Boomer (ages 58-76)	1.69 *
	[1.01, 2.85]
GENERATIONSilent (ages 77+)	2.21 **
	[1.23, 3.97]
GENDERFemale	0.60 ***
	[0.50, 0.71]
RACENEWNon-Hispanic Black	0.79
	[0.59, 1.05]
RACENEWHispanic	0.94
	[0.71, 1.24]
RACENEWAsian	0.90
	[0.49, 1.65]
POLPARTY_NEWDemocrat (no lean)	1.66 ***
	[1.30, 2.11]
POLPARTY_NEWIndependent (includes leaners)	1.27
	[0.99, 1.62]
educrSome college	1.21
	[0.97, 1.51]
educr4-year degree	1.17
	[0.86, 1.61]
educrPostgrad work or degree	1.75 ***
	[1.34, 2.29]
INCOMENEW$30,000-$49,999 per year	0.87
	[0.63, 1.19]
INCOMENEW$50,000 - $99,999 per year	0.80
	[0.60, 1.05]
INCOMENEW$100,000-$149,999 per year	0.99
	[0.73, 1.34]
INCOMENEW$150,000 or more per year	0.81
	[0.58, 1.11]
area.live.newLives in small city or suburb	1.23 *
	[1.00, 1.51]
area.live.newLives in town or rural area	1.24 *
	[1.00, 1.54]
N	4921
R2
*** p < 0.001; ** p < 0.01; * p < 0.05.

A final note on the Q2 series. Though the output is not shown here, two other issues were examined with respect to these items – their inter-item correlation and whether they would make a reliable sclae (one which would simply take an average value of the three questions as a measure of a person’s overall interest in the news). For this analysis, the items were reverse coded, i.e. 1 represented having no interest and 4 represented having a great deal of interest. Findings include:

• Inter-item correlation Note this correlation was calculated withs urvey weights, however the correlation coefficient is Pearson’s, which is an analytical choice that others may object to (if this is of strong interest, there are other correlation tests that can be tested). The inter-item correlation is not great: For local news, it has a 0.49 correlation with national news and 0.39 correlation with international news. For national news, it’s correlation with international news is higher at 0.69 (international news’ correlations are mentioned in the previous two sentences, but is is 0.39 for local and 0.69 for national). So interest in one type of news is not necessarily indicative of a broader interest in the news.

• Scale reliability Despite some subpar inter-item correlations, Cronbach’s alpha still suggests these items would make a reliable scale, with the value standing at 0.767.

Q3 SERIES: How often, if at all, do you use each of the following for staying up to date on the news? (Daily or not)

There are seven items in this question series, including: a newspaper, television station, a website or app, a radio station, a magazine, direct communication with people in your local area and a social media platform.

The Gallup topline document focused largely on “daily,” usage, an approach which will be followed here. The variables are recoded in the below code. Note, in the datafile, this variable is referred to as “Q47”. Simple table outputs are shown to verify the recoding is correct, where we want “daily,” in the original variable to match “1.” Again, "no answer is not considered here.

df<-df%>%
  mutate(across(all_of(q3.vars), haven::zap_missing),
         across(all_of(q3.vars), haven::as_factor))%>%
  mutate(across(all_of(q3.vars), list(rec = ~ recode(., "Daily"=1,
                                                     "Weekly"=0,
                                                     "Monthly"=0,
                                                     "Less than monthly"=0,
                                                     "Never"=0))))


label_df<-label_df%>%
  add_row(QTAG = "Q47A_rec", Wording = "Q3A (recoded): Newspaper: Use daily/not daily to stay yp to date on news")%>%
  add_row(QTAG = "Q47B_rec", Wording = "Q3B (recoded) Television station: Use daily/not daily to stay up to date on news")%>%
  add_row(QTAG = "Q47C_rec", Wording = "Q3C (recoded): Website or app: Use daily/not daily to stay yp to date on news")%>%
  add_row(QTAG = "Q47D_rec", Wording = "Q3D (recoded) Radio station: Use daily/not daily to stay up to date on news")%>%
  add_row(QTAG = "Q47E_rec", Wording = "Q3E (recoded): Magazine: Use daily/not daily to stay yp to date on news")%>%
  add_row(QTAG = "Q47F_rec", Wording = "Q3F (recoded) Direct communication with people: Use daily/not daily to stay up to date on news")%>%
  add_row(QTAG = "Q47G_rec", Wording = "Q3G (recoded): Social media platform: Use daily/not daily to stay yp to date on news")



table(df$Q47A_rec)

## 
##    0    1 
## 4070 1466

table(df$Q47A)

## 
##             Daily            Weekly           Monthly Less than monthly 
##              1466              1029               354              1089 
##             Never         No answer 
##              1598                57

table(df$Q47B)

## 
##             Daily            Weekly           Monthly Less than monthly 
##              2663               965               400               806 
##             Never         No answer 
##               742                17

table(df$Q47B_rec)

## 
##    0    1 
## 2913 2663

table(df$Q47C_rec)

## 
##    0    1 
## 3341 2177

table(df$Q47C)

## 
##             Daily            Weekly           Monthly Less than monthly 
##              2177              1072               452               800 
##             Never         No answer 
##              1017                75

table(df$Q47D)

## 
##             Daily            Weekly           Monthly Less than monthly 
##              1549              1060               380              1071 
##             Never         No answer 
##              1473                60

table(df$Q47D_rec)

## 
##    0    1 
## 3984 1549

table(df$Q47E_rec)

## 
##    0    1 
## 5334  194

table(df$Q47E)

## 
##             Daily            Weekly           Monthly Less than monthly 
##               194               548               716              1514 
##             Never         No answer 
##              2556                65

table(df$Q47F)

## 
##             Daily            Weekly           Monthly Less than monthly 
##              1344              1812               686               928 
##             Never         No answer 
##               759                64

table(df$Q47F_rec)

## 
##    0    1 
## 4185 1344

table(df$Q47G_rec)

## 
##    0    1 
## 3364 2174

table(df$Q47G)

## 
##             Daily            Weekly           Monthly Less than monthly 
##              2174               902               323               654 
##             Never         No answer 
##              1485                55

Q3A: Newspaper – (daily usage to stay up-to-date on the news)

Daily usage of the newspaper stands at 25% overall. There are a number of different factors which are predictive of a higher likelihood to read the paper daily, but these relationships are not necessarily glaringly obvious when one turns to the cross tab (this of course is due to the fact that the regression analysis is taking into account many factors beyond the bi-variate relationship between the dependent variable and any given independent variable).

All output for these regressions will have the name “Q3,” in the file name, all of which are in the same file location.

For newspapers, significant predictors include:

• Generation: Daily usage rises considerably with age – indeed age is the most salient predictor here, and this can be demonstrated via the crosstabs (35% of the Silent Generation, for instance, read a newspaper daily while 14% of Gen Z does the same).
  
• Gender: Women are 24% less likely than men to read the paper daily (odds ratio of 0.76).
  
• Racial identity: Barring Asians (whose result is not statistically significant), the other race/ethnicity groups are less likely than white people to read the paper everyday. In other words, white people are the biggest readers of newspapers, at least on a daily basis.
• Political Party: Democrats are 2.35 times more likely than Republicans to read the newspaper daily, one of the largest effect sizes here.
  
• Education/Income: Grouping both items together here. A similar relationship for both – the more education a person has or the higher up the income ladder a person climbs – the more likely that person reads the newspaper on a daily basis.
  
• Urbanicity (area live in): People in a rural area about 31% less likely to read a newspaper everyday than those living in a big city/suburb of a big city (the reference category).

Observations	4921
Dependent variable	Q47A_rec
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.16
Pseudo-R² (McFadden)	0.10
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	0.06	0.03	0.11	-9.85	0.00
GENERATIONMillennial (ages 27-41)	1.56	1.01	2.42	2.01	0.04
GENERATIONGen X (ages 42-57)	2.58	1.64	4.04	4.13	0.00
GENERATIONBaby Boomer (ages 58-76)	4.28	2.73	6.71	6.32	0.00
GENERATIONSilent (ages 77+)	7.28	4.35	12.18	7.56	0.00
GENDERFemale	0.76	0.65	0.89	-3.35	0.00
RACENEWNon-Hispanic Black	0.67	0.53	0.86	-3.12	0.00
RACENEWHispanic	0.76	0.59	0.98	-2.14	0.03
RACENEWAsian	0.88	0.52	1.50	-0.46	0.65
POLPARTY_NEWDemocrat (no lean)	2.35	1.90	2.91	7.89	0.00
POLPARTY_NEWIndependent (includes leaners)	1.13	0.90	1.42	1.08	0.28
educrSome college	1.31	1.06	1.62	2.46	0.01
educr4-year degree	1.88	1.43	2.47	4.55	0.00
educrPostgrad work or degree	2.33	1.82	2.98	6.74	0.00
INCOMENEW$30,000-$49,999 per year	1.27	0.93	1.73	1.50	0.13
INCOMENEW$50,000 - $99,999 per year	1.47	1.11	1.95	2.70	0.01
INCOMENEW$100,000-$149,999 per year	1.47	1.08	1.99	2.44	0.01
INCOMENEW$150,000 or more per year	1.89	1.38	2.59	3.98	0.00
area.live.newLives in small city or suburb	0.87	0.73	1.04	-1.50	0.13
area.live.newLives in town or rural area	0.69	0.56	0.84	-3.73	0.00
Standard errors: Robust

Q3B: Television: (daily usage to stay up-to-date on the news)

Nearly half of Americans (47%) use television on a daily basis to stay up to date on the news. Due to its relative popularity, there are fewer significant differences between the demographic groups of interest here. Notable ones include:

• Generation The odds ratio estimates for this variable are mind-boggling. Baby Boomers are 28.7 times more likely than Gen Z to use television daily; the Silient Generation is 47.7 times more likely. The crosstabs illustrate this nicely: just 11% of Americans who fall into the Gen Z cohort use telvision daily for news; 81% of the Silent Generation does. There is definitely a generational divide on this form of news.
• Racial identity Black individuals are 1.69 times more likely than whites to use television daily. There are no other significant differences between the other racial/ethnic groups.
• Political Party Democrats are 1.49 times more likely to rely on television daily to stay up on the news. At this point, it just appears that Democrats like to stay up on the news more so than Republicans, and that has a ripple effect across the news platforms (or at least that is one hypothesis).
  
• Education As we might expect, television usage decreases with higher levels of education.
• Income Similar dynamic as education (more income meaning less likely to watch television), but the effect is weaker and for some brackets not always significant. Would focus more on education than income.
• Urbanicity (Area live in) Again, those in a small town or rural area are less likely to watch television daily for the purposes of getting news (23% less likely).

Observations	4949
Dependent variable	Q47B_rec
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.36
Pseudo-R² (McFadden)	0.23
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	0.15	0.08	0.28	-6.02	0.00
GENERATIONMillennial (ages 27-41)	2.74	1.60	4.67	3.69	0.00
GENERATIONGen X (ages 42-57)	10.24	6.01	17.44	8.56	0.00
GENERATIONBaby Boomer (ages 58-76)	28.67	16.75	49.05	12.25	0.00
GENERATIONSilent (ages 77+)	47.69	25.94	87.67	12.44	0.00
GENDERFemale	1.00	0.86	1.17	0.04	0.96
RACENEWNon-Hispanic Black	1.69	1.34	2.12	4.46	0.00
RACENEWHispanic	0.97	0.77	1.23	-0.24	0.81
RACENEWAsian	1.51	0.90	2.52	1.56	0.12
POLPARTY_NEWDemocrat (no lean)	1.49	1.22	1.81	3.89	0.00
POLPARTY_NEWIndependent (includes leaners)	0.91	0.75	1.12	-0.87	0.39
educrSome college	0.77	0.63	0.94	-2.63	0.01
educr4-year degree	0.67	0.52	0.86	-3.19	0.00
educrPostgrad work or degree	0.60	0.47	0.76	-4.11	0.00
INCOMENEW$30,000-$49,999 per year	0.71	0.53	0.94	-2.36	0.02
INCOMENEW$50,000 - $99,999 per year	0.70	0.55	0.91	-2.69	0.01
INCOMENEW$100,000-$149,999 per year	0.74	0.56	0.98	-2.10	0.04
INCOMENEW$150,000 or more per year	0.74	0.55	1.00	-1.96	0.05
area.live.newLives in small city or suburb	0.87	0.72	1.04	-1.54	0.12
area.live.newLives in town or rural area	0.77	0.64	0.94	-2.61	0.01
Standard errors: Robust

Q3C. Website/app (daily usage to stay up-to-date on the news)

This mode of getting the news is used by 38% of Americans on a daily basis to get the news. Here are the findings of the regression analysis:

• Generation Not sure if this was expected, but daily usage of a website or app does increase with age – up until a point, as the result for the Silent Generation is not significant. I think the larger message here is that Gen Z is less keen to use any news platform on a daily basis to consume their news, so the analysis might consider the relative ranking of sources between the generations (i.e. which one is used most often by that cohort, rather than the absolute trends).
  
• Racial identity Not significant.
• Political Party Believe it or not, but this not significant.
  
• Education This is significant and, again, people with higher levels of education are more likely to use this on a daily basis to get the news.
  
• Income Has a very similar trend as education.
  
• Urbanicity (area live in) Both groups who do not live in a big city or a suburb are slightly less likely to use this on a daily basis.

Observations	4904
Dependent variable	Q47C_rec
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.09
Pseudo-R² (McFadden)	0.05
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	0.40	0.25	0.64	-3.89	0.00
GENERATIONMillennial (ages 27-41)	1.53	1.05	2.24	2.22	0.03
GENERATIONGen X (ages 42-57)	2.12	1.44	3.14	3.78	0.00
GENERATIONBaby Boomer (ages 58-76)	1.78	1.20	2.63	2.85	0.00
GENERATIONSilent (ages 77+)	1.15	0.72	1.84	0.58	0.57
GENDERFemale	0.62	0.54	0.71	-6.96	0.00
RACENEWNon-Hispanic Black	0.88	0.70	1.09	-1.18	0.24
RACENEWHispanic	0.85	0.69	1.05	-1.53	0.13
RACENEWAsian	1.46	0.95	2.25	1.73	0.08
POLPARTY_NEWDemocrat (no lean)	0.86	0.72	1.03	-1.64	0.10
POLPARTY_NEWIndependent (includes leaners)	0.85	0.71	1.02	-1.76	0.08
educrSome college	1.22	1.02	1.47	2.17	0.03
educr4-year degree	1.33	1.05	1.67	2.40	0.02
educrPostgrad work or degree	1.35	1.09	1.69	2.70	0.01
INCOMENEW$30,000-$49,999 per year	1.23	0.95	1.60	1.55	0.12
INCOMENEW$50,000 - $99,999 per year	1.29	1.02	1.62	2.13	0.03
INCOMENEW$100,000-$149,999 per year	1.62	1.25	2.10	3.63	0.00
INCOMENEW$150,000 or more per year	1.49	1.14	1.95	2.89	0.00
area.live.newLives in small city or suburb	0.85	0.72	1.00	-1.98	0.05
area.live.newLives in town or rural area	0.81	0.68	0.97	-2.36	0.02
Standard errors: Robust

## # A tibble: 6 x 6
## # Rowwise: 
##   dep.question                   variable odds.ratio   S.E.     p is.significant
##   <chr>                          <chr>         <dbl>  <dbl> <dbl> <chr>         
## 1 Q3C.Use daily/not daily: Webs~ GENERAT~      1.53  0.193  0.026 Yes           
## 2 Q3C.Use daily/not daily: Webs~ GENERAT~      2.12  0.199  0     Yes           
## 3 Q3C.Use daily/not daily: Webs~ GENERAT~      1.78  0.201  0.004 Yes           
## 4 Q3C.Use daily/not daily: Webs~ GENERAT~      1.15  0.240  0.565 No            
## 5 Q3C.Use daily/not daily: Webs~ GENDERF~      0.616 0.0696 0     Yes           
## 6 Q3C.Use daily/not daily: Webs~ RACENEW~      0.876 0.112  0.239 No

Q3D: Radio station (daily usage to stay up-to-date on the news)

This source is utilized by 26% of Americans on a daily basis for the purposes of catching up on the news. There is essentially one significant variable when looking at this item, and the results are somewhat perplexing. It is:

• GENERATION: Radio usage is highest within the Gen X and Baby Boomer generations – both relative to Gen Z (the reference category in the regression), but also the crosstabs suggest as much as well. A third of Gen X use radio on a daily basis for the news, as does 28% of Baby Boomers. About a fifth of the Millennials and the Silent Generation, by contrast, do this. For Gen Z, the figure is 7%.

Note independents here is significant, but this is a result which is difficult to interpret (there is no significant difference between Republicans and Democrats, meanwhile). If the regression is re-run to use the variable which allocates the leaners, then being an independent is no longer a significant predictor.

Observations	4918
Dependent variable	Q47D_rec
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.09
Pseudo-R² (McFadden)	0.05
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	0.09	0.05	0.17	-8.00	0.00
GENERATIONMillennial (ages 27-41)	3.21	1.89	5.46	4.31	0.00
GENERATIONGen X (ages 42-57)	5.82	3.41	9.95	6.44	0.00
GENERATIONBaby Boomer (ages 58-76)	4.71	2.75	8.06	5.65	0.00
GENERATIONSilent (ages 77+)	3.92	2.15	7.12	4.47	0.00
GENDERFemale	0.78	0.67	0.90	-3.35	0.00
RACENEWNon-Hispanic Black	1.17	0.94	1.46	1.43	0.15
RACENEWHispanic	0.99	0.78	1.26	-0.08	0.93
RACENEWAsian	0.56	0.31	1.01	-1.93	0.05
POLPARTY_NEWDemocrat (no lean)	0.89	0.74	1.07	-1.26	0.21
POLPARTY_NEWIndependent (includes leaners)	0.62	0.51	0.75	-4.89	0.00
educrSome college	1.10	0.91	1.33	0.99	0.32
educr4-year degree	1.10	0.86	1.42	0.76	0.44
educrPostgrad work or degree	1.15	0.91	1.47	1.18	0.24
INCOMENEW$30,000-$49,999 per year	0.98	0.74	1.30	-0.13	0.90
INCOMENEW$50,000 - $99,999 per year	1.13	0.88	1.46	0.94	0.35
INCOMENEW$100,000-$149,999 per year	1.44	1.09	1.92	2.54	0.01
INCOMENEW$150,000 or more per year	1.30	0.97	1.75	1.75	0.08
area.live.newLives in small city or suburb	0.99	0.82	1.19	-0.12	0.90
area.live.newLives in town or rural area	1.07	0.89	1.28	0.68	0.49
Standard errors: Robust

## # A tibble: 6 x 6
## # Rowwise: 
##   dep.question                   variable odds.ratio   S.E.     p is.significant
##   <chr>                          <chr>         <dbl>  <dbl> <dbl> <chr>         
## 1 Q3D.Use daily/not daily: Radi~ GENERAT~      3.21  0.271  0     Yes           
## 2 Q3D.Use daily/not daily: Radi~ GENERAT~      5.82  0.273  0     Yes           
## 3 Q3D.Use daily/not daily: Radi~ GENERAT~      4.71  0.274  0     Yes           
## 4 Q3D.Use daily/not daily: Radi~ GENERAT~      3.92  0.305  0     Yes           
## 5 Q3D.Use daily/not daily: Radi~ GENDERF~      0.776 0.0756 0.001 Yes           
## 6 Q3D.Use daily/not daily: Radi~ RACENEW~      1.17  0.112  0.154 No

Q3E. A magazine (daily usage to stay up-to-date on the news)

This is not the greatest model in the world, as 3% use this daily. If this is not necessary to refer to, please ignore, though output appears below (without commentary), and model output is saved in the folder.

Observations	4910
Dependent variable	Q47E_rec
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.11
Pseudo-R² (McFadden)	0.09
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	0.04	0.01	0.11	-6.16	0.00
GENERATIONMillennial (ages 27-41)	0.88	0.38	2.05	-0.29	0.77
GENERATIONGen X (ages 42-57)	1.16	0.49	2.73	0.34	0.73
GENERATIONBaby Boomer (ages 58-76)	1.10	0.45	2.69	0.22	0.83
GENERATIONSilent (ages 77+)	1.33	0.42	4.20	0.49	0.63
GENDERFemale	0.44	0.31	0.62	-4.60	0.00
RACENEWNon-Hispanic Black	1.35	0.86	2.12	1.32	0.19
RACENEWHispanic	1.64	1.02	2.66	2.02	0.04
RACENEWAsian	0.91	0.30	2.79	-0.16	0.87
POLPARTY_NEWDemocrat (no lean)	1.73	1.07	2.81	2.22	0.03
POLPARTY_NEWIndependent (includes leaners)	0.74	0.43	1.28	-1.08	0.28
educrSome college	1.37	0.82	2.31	1.19	0.23
educr4-year degree	1.61	0.85	3.06	1.47	0.14
educrPostgrad work or degree	2.75	1.50	5.04	3.29	0.00
INCOMENEW$30,000-$49,999 per year	0.54	0.29	1.00	-1.95	0.05
INCOMENEW$50,000 - $99,999 per year	0.89	0.52	1.50	-0.45	0.65
INCOMENEW$100,000-$149,999 per year	0.61	0.30	1.27	-1.31	0.19
INCOMENEW$150,000 or more per year	0.82	0.44	1.51	-0.65	0.52
area.live.newLives in small city or suburb	0.57	0.38	0.84	-2.85	0.00
area.live.newLives in town or rural area	0.54	0.34	0.85	-2.68	0.01
Standard errors: Robust

## # A tibble: 6 x 6
## # Rowwise: 
##   dep.question                    variable odds.ratio  S.E.     p is.significant
##   <chr>                           <chr>         <dbl> <dbl> <dbl> <chr>         
## 1 Q3E.Use daily/not daily: Magaz~ GENERAT~      0.882 0.430 0.77  No            
## 2 Q3E.Use daily/not daily: Magaz~ GENERAT~      1.16  0.437 0.733 No            
## 3 Q3E.Use daily/not daily: Magaz~ GENERAT~      1.10  0.453 0.826 No            
## 4 Q3E.Use daily/not daily: Magaz~ GENERAT~      1.33  0.587 0.626 No            
## 5 Q3E.Use daily/not daily: Magaz~ GENDERF~      0.438 0.180 0     Yes           
## 6 Q3E.Use daily/not daily: Magaz~ RACENEW~      1.35  0.230 0.187 No

Q3F. Direct communication (daily usage to stay up-to-date on the news)

Overall, 24% of Americans use “direct communciation with people in your local area (outside of your household), including in person, on the phone or online,” on a daily basis to stay up-to-date on the news.

Very few demographics or characteristics are significant for this item (one exception occurs among Black people, whoa re 1.65 times more likely than white people to use this daily).

Output appears below without commentary, and information is saved to folder.

Observations	4913
Dependent variable	Q47F_rec
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.05
Pseudo-R² (McFadden)	0.03
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	0.40	0.25	0.64	-3.82	0.00
GENERATIONMillennial (ages 27-41)	1.10	0.74	1.63	0.47	0.64
GENERATIONGen X (ages 42-57)	0.95	0.64	1.43	-0.24	0.81
GENERATIONBaby Boomer (ages 58-76)	0.71	0.48	1.07	-1.61	0.11
GENERATIONSilent (ages 77+)	0.75	0.45	1.23	-1.14	0.26
GENDERFemale	0.93	0.80	1.09	-0.88	0.38
RACENEWNon-Hispanic Black	1.65	1.32	2.05	4.45	0.00
RACENEWHispanic	1.01	0.79	1.28	0.06	0.95
RACENEWAsian	1.09	0.66	1.79	0.33	0.74
POLPARTY_NEWDemocrat (no lean)	0.83	0.68	1.01	-1.84	0.07
POLPARTY_NEWIndependent (includes leaners)	0.77	0.63	0.94	-2.51	0.01
educrSome college	1.14	0.94	1.39	1.33	0.18
educr4-year degree	0.92	0.71	1.20	-0.62	0.54
educrPostgrad work or degree	0.96	0.74	1.23	-0.33	0.74
INCOMENEW$30,000-$49,999 per year	0.97	0.73	1.27	-0.24	0.81
INCOMENEW$50,000 - $99,999 per year	0.93	0.72	1.20	-0.56	0.58
INCOMENEW$100,000-$149,999 per year	1.13	0.85	1.50	0.82	0.41
INCOMENEW$150,000 or more per year	0.98	0.73	1.32	-0.14	0.89
area.live.newLives in small city or suburb	0.98	0.81	1.18	-0.23	0.81
area.live.newLives in town or rural area	0.96	0.79	1.16	-0.47	0.64
Standard errors: Robust

## # A tibble: 6 x 6
## # Rowwise: 
##   dep.question                   variable odds.ratio   S.E.     p is.significant
##   <chr>                          <chr>         <dbl>  <dbl> <dbl> <chr>         
## 1 Q3F.Use daily/not daily: Dire~ GENERAT~      1.10  0.200  0.636 No            
## 2 Q3F.Use daily/not daily: Dire~ GENERAT~      0.952 0.206  0.812 No            
## 3 Q3F.Use daily/not daily: Dire~ GENERAT~      0.714 0.208  0.106 No            
## 4 Q3F.Use daily/not daily: Dire~ GENERAT~      0.748 0.255  0.255 No            
## 5 Q3F.Use daily/not daily: Dire~ GENDERF~      0.933 0.0794 0.381 No            
## 6 Q3F.Use daily/not daily: Dire~ RACENEW~      1.65  0.112  0     Yes

Q3G A social media platform (daily usage to stay up-to-date on the news)

Overall, 39% of Americans usea social media platform to stay up to date on the news on a daily basis. In terms of the major dividing lines on this question, it is without a doubt generation. Gen Z is much more likely to use social media on a daily basis, as can be seen by the below regression output – however, the crosstabs tell the story just as well. 65% of Gen Z uses social media on a daily basis to get the news, compared to 50% of Millennials, 40% of Gen X, 26% of Baby Boomers and 18% of Silent Generation.

Beyond this factor, relatively little else stands out as significant in terms of predicting daily usage of social media for the purposes of consuming the news. Regression output is below and saved in the folder.

Observations	4921
Dependent variable	Q47G_rec
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.13
Pseudo-R² (McFadden)	0.07
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	1.90	1.23	2.92	2.91	0.00
GENERATIONMillennial (ages 27-41)	0.49	0.35	0.70	-3.99	0.00
GENERATIONGen X (ages 42-57)	0.35	0.24	0.49	-5.82	0.00
GENERATIONBaby Boomer (ages 58-76)	0.18	0.13	0.26	-9.29	0.00
GENERATIONSilent (ages 77+)	0.13	0.08	0.21	-8.48	0.00
GENDERFemale	1.09	0.95	1.25	1.21	0.23
RACENEWNon-Hispanic Black	1.27	1.03	1.56	2.27	0.02
RACENEWHispanic	1.03	0.83	1.27	0.25	0.80
RACENEWAsian	0.94	0.61	1.45	-0.28	0.78
POLPARTY_NEWDemocrat (no lean)	1.03	0.86	1.24	0.32	0.75
POLPARTY_NEWIndependent (includes leaners)	0.91	0.75	1.09	-1.04	0.30
educrSome college	1.01	0.84	1.21	0.09	0.93
educr4-year degree	1.10	0.87	1.38	0.80	0.42
educrPostgrad work or degree	0.98	0.78	1.23	-0.18	0.85
INCOMENEW$30,000-$49,999 per year	1.02	0.79	1.31	0.13	0.90
INCOMENEW$50,000 - $99,999 per year	1.08	0.85	1.36	0.63	0.53
INCOMENEW$100,000-$149,999 per year	1.11	0.86	1.44	0.82	0.41
INCOMENEW$150,000 or more per year	1.11	0.85	1.46	0.75	0.45
area.live.newLives in small city or suburb	1.04	0.88	1.22	0.42	0.68
area.live.newLives in town or rural area	0.80	0.67	0.95	-2.53	0.01
Standard errors: Robust

Q4: In which format do you get most of your news?

This question offered the following response options: printed newspaper or magazine (3%), television (31%), online using a computer, smartphone or app (58%) and radio (7%). No regression analysis was conducted on this question. If one is desired, please let me know, and specify the outcome of interest.

Q5: Which one of the following do you use the most often?

This question has around 20 response options, including a verbatim. No regression analysis was conducted on this item. If one is desired, it would be necessary to specify the outcome of interest.

Q6. What is your overall opinion of the news media in the United States today? (Q5 in dataset)

This analysis could be handled in multiple different ways, including:

• A linear regression, which treats the 5-point scale as continuous. For ease of interpretation, it is desirable to reverse-recode this variable such that 1=very favorable and 5=very unfavorable (meaning a positive coefficient is associated with a rise in unfavorable feelings). NA would be removed.
• Logistic regression – Here the question is which category to combine. Should it be “very unfavorable,” or “very unfavorable”? Both were tried, and the model for “very unfavorable,” produced clearer results (as combining very unfavorable and unfavorable yields a majority).

All three models confirmed the same general findings. For the sake of brevity, this will show the “very unfavorable,” logistic regression results, though the output is available for all of them.

In the logistic regression predicting who will hold a very unfavorable view of the media, the following groups have significant differences:

• GENERATION Notably, this is not significant.
• Gender Women are 39% less likely to have a very unfavorable opinion of the news media.
• Racial Identity Black Americans are 37% less likely to have a very unfavorable opinion of the news media, a statistically significant effect even when controlling for party identification.
• Political Party By far the largest effect size, as Democrats are 92% less likely to have a very unfavorable overall opinion of the news media than Republicans. Independents (which includes leaners) also are less likely to hold this opinion than outright Republicans. Interestingly, this effect does not really change if we instead allocate the leaners.
• Household Income Somewhat surprising (to this analyst anyway), but holding a “very unfavorable” opinion of the news media decreases as you move up the income bracket.
• Urbanicity (Area live in) People in small towns/rural areas are 1.36 times more likely to hold a very unfavorable opinion of the news media than those living in a big city or suburb of a big city.

The output below shows the odds ratio associated with holding a very unfavorable opinion of the news media in the US.

Observations	4962
Dependent variable	Q5_very_unfavorable
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.31
Pseudo-R² (McFadden)	0.21
AIC	NA

	exp(Est.)	S.E.	t val.	p
(Intercept)	0.99	0.26	-0.06	0.96
GENERATIONMillennial (ages 27-41)	0.88	0.22	-0.60	0.55
GENERATIONGen X (ages 42-57)	0.73	0.22	-1.40	0.16
GENERATIONBaby Boomer (ages 58-76)	0.72	0.22	-1.47	0.14
GENERATIONSilent (ages 77+)	0.71	0.29	-1.21	0.23
GENDERFemale	0.61	0.08	-5.91	0.00
RACENEWNon-Hispanic Black	0.63	0.18	-2.67	0.01
RACENEWHispanic	1.07	0.13	0.51	0.61
RACENEWAsian	0.60	0.35	-1.47	0.14
POLPARTY_NEWDemocrat (no lean)	0.08	0.13	-19.61	0.00
POLPARTY_NEWIndependent (includes leaners)	0.53	0.09	-6.73	0.00
educrSome college	1.21	0.11	1.81	0.07
educr4-year degree	0.90	0.14	-0.71	0.48
educrPostgrad work or degree	0.80	0.14	-1.62	0.10
INCOMENEW$30,000-$49,999 per year	1.29	0.17	1.51	0.13
INCOMENEW$50,000 - $99,999 per year	1.66	0.15	3.45	0.00
INCOMENEW$100,000-$149,999 per year	1.74	0.16	3.46	0.00
INCOMENEW$150,000 or more per year	1.96	0.17	4.01	0.00
area.live.newLives in small city or suburb	1.05	0.10	0.51	0.61
area.live.newLives in town or rural area	1.36	0.10	3.03	0.00
Standard errors: Robust

Q7. To what extent do you see political bias in news coverage? (Q6 in datafile)

For this question, 55% of Americans said “a great deal,” – making this the clear outcome of interest. Again, a logistic regression is the sensible approach.

Political party is the most slaient predictor for this item. Democrats are 86% less likely than Republicans to say they see “a great deal,” of bias in the news media; though a few other variables are significant, including gender (women are 26% less likely to say this) and, again, income (higher income generally translates to a higher likelihood of saying this).

The output below shows the results for predicting who will say they see “a great deal” of political bias in news coverage.

Observations	4962
Dependent variable	q6.great.deal
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.25
Pseudo-R² (McFadden)	0.15
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	3.34	2.08	5.38	4.97	0.00
GENERATIONMillennial (ages 27-41)	0.80	0.55	1.16	-1.19	0.24
GENERATIONGen X (ages 42-57)	0.81	0.55	1.19	-1.06	0.29
GENERATIONBaby Boomer (ages 58-76)	0.80	0.54	1.18	-1.13	0.26
GENERATIONSilent (ages 77+)	0.72	0.46	1.15	-1.37	0.17
GENDERFemale	0.74	0.64	0.86	-3.99	0.00
RACENEWNon-Hispanic Black	0.83	0.66	1.03	-1.68	0.09
RACENEWHispanic	0.93	0.75	1.16	-0.61	0.54
RACENEWAsian	0.44	0.27	0.71	-3.37	0.00
POLPARTY_NEWDemocrat (no lean)	0.14	0.11	0.17	-19.61	0.00
POLPARTY_NEWIndependent (includes leaners)	0.42	0.35	0.52	-8.28	0.00
educrSome college	1.32	1.09	1.60	2.88	0.00
educr4-year degree	1.18	0.93	1.50	1.35	0.18
educrPostgrad work or degree	1.18	0.94	1.49	1.40	0.16
INCOMENEW$30,000-$49,999 per year	1.33	1.01	1.74	2.03	0.04
INCOMENEW$50,000 - $99,999 per year	1.43	1.12	1.83	2.87	0.00
INCOMENEW$100,000-$149,999 per year	1.66	1.26	2.18	3.60	0.00
INCOMENEW$150,000 or more per year	1.76	1.32	2.33	3.91	0.00
area.live.newLives in small city or suburb	1.04	0.88	1.24	0.48	0.63
area.live.newLives in town or rural area	1.19	0.99	1.43	1.83	0.07
Standard errors: Robust

Q8. Please indicate which of these staements come closer to how you feel (though there is some bias, people can sort out facts OR so much bias in news media it is difficult to sort out facts)

This item is on hold per message from Sarah F.

High Priority Item: q28-series “When you are unsure about something in the news and want to find out the truth, which of the following do you typically do?” (K14_1-K14_7 in the data series)

This question is a multiple response item which included a “other,” response – and though this option was selected by 4% of Americans, it will be omitted from the analysis here.

Putting that option aside, it is striking how uneven the distribution between the response options are, inclduing:

• I do my own research using a variety of information sources (q28_2), 68%
• I turn to the news organizations I trust the most (q28_3), 54%
  
• I turn to friends and family (q28_5), 19%
  
• I look for information or opinions from people on social media (q28_1), 15%
• I turn to my faith or a religious institution, or leader, or text (q28_4), 9%
• I don’t usually take additional steps to find out the truth (q28_7) 4%

This analysis will examine each response option in turn, starting from the most popular option (Q28_2, or Q28, response option 2), which is “I do my own research or opinions from people on social media.”

One might argue some element of social desirability is driving the popularity of this response – most people want to be as a person who does their own investigation of matters of import. On any account, a logistic regression was run, where the outcome of interest was whether a person gave this response. All other individuals were categorized as 0 (however no answer was excluded in this analysis).

Note the output for these items consist of six spreadsheets, all of which have the name “Q28” in them and then the specific response option. Please let me know if you cannot locate them.

Q28_2 When unsure about something and want to find out the truth: I do my own research using a variety of information sources

The output of this analysis appears below and is likewise saved in the folder. Key highlights include:

• Generation While not all age cohorts had statistically significant results, the two oldest cohorts did (Baby Boomers and the Silent Generation). In both instances, this group was less likely than the reference group of Gen Z to say they do their own research. This relationship can be seen in the crosstabs as well – while over 70% of the youngest three generations provide this response, the figure drops to 58% among the Baby Boomers and 47% for the Silent Generation.
  
• Gender Women are less likely to provide this response than men, about 31% less likely. Again, the crosstabs hint at this dynamic, as 73% of men said this compared to 64% of women.
• Racial Identity This is not significant and apparent differences should not be reported here.
• Political Party Believe it or not, there are no significant differences between Democrats (no leaners) and Republicans (no leaners). Among independents (with leaners), there is a statistically significant effect – this group is 1.47 times more likely than the GOP to say they do their own research. It is important to note that this relationship disappears entirely if we allocate the leaners – in other words “true,” independents are not more likely to do their own research.
  
• Education This is probably the most influential variable in the model. Compared to those who have a high school degree or less, people who attended some college are 1.69 times more likely to say they do their own research; those with a 4-year degree are 1.72 times more likely and postgrad work are 2.36 times more likely. Again, this is not the education variable in the crosstabs, so exploring this relationship will require a fresh data cut. It does seem to be worth it though.
  
• Income A somewhat similar relationship to education, though it is not always significant and peters out as one moves up the income ladder.

Observations	4945
Dependent variable	q28_2
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.15
Pseudo-R² (McFadden)	0.09
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	1.58	1.00	2.52	1.94	0.05
GENERATIONMillennial (ages 27-41)	1.15	0.77	1.71	0.69	0.49
GENERATIONGen X (ages 42-57)	0.99	0.65	1.49	-0.06	0.95
GENERATIONBaby Boomer (ages 58-76)	0.59	0.39	0.89	-2.52	0.01
GENERATIONSilent (ages 77+)	0.42	0.26	0.67	-3.66	0.00
GENDERFemale	0.69	0.59	0.80	-4.82	0.00
RACENEWNon-Hispanic Black	0.85	0.68	1.05	-1.49	0.14
RACENEWHispanic	0.87	0.68	1.10	-1.17	0.24
RACENEWAsian	1.00	0.59	1.69	-0.02	0.99
POLPARTY_NEWDemocrat (no lean)	1.07	0.89	1.29	0.73	0.47
POLPARTY_NEWIndependent (includes leaners)	1.47	1.20	1.80	3.75	0.00
educrSome college	1.69	1.41	2.03	5.61	0.00
educr4-year degree	1.72	1.34	2.21	4.30	0.00
educrPostgrad work or degree	2.36	1.85	3.02	6.88	0.00
INCOMENEW$30,000-$49,999 per year	1.42	1.10	1.84	2.70	0.01
INCOMENEW$50,000 - $99,999 per year	1.50	1.20	1.89	3.49	0.00
INCOMENEW$100,000-$149,999 per year	1.31	1.00	1.70	2.00	0.05
INCOMENEW$150,000 or more per year	1.30	0.98	1.73	1.83	0.07
area.live.newLives in small city or suburb	0.99	0.83	1.19	-0.09	0.93
area.live.newLives in town or rural area	0.94	0.78	1.13	-0.65	0.51
Standard errors: Robust

Q28_3 When unsure about something in news…I turn to the news organizations I trust the most

The next response option of interest is “I turn to the news organizations I trust the most.” Even though there is a widespread lack of trust in media, 54% of Americans still at least have some source they feel they can rely on.

Who are these Americans? Key highlights are:

• Generation Again, the older age groups show statistically significant effects, but this time they are more likely to say they do turn to their trusted news organizations.
• Racial identity Both Blacks and Hispanics are less likely than whites to provide this response.
• Political Party Democrats are more likely than Republcians to say this (1.83 in odds ratio terms).

Observations	4945
Dependent variable	q28_3
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.12
Pseudo-R² (McFadden)	0.07
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	0.77	0.51	1.17	-1.24	0.22
GENERATIONMillennial (ages 27-41)	0.91	0.65	1.27	-0.58	0.56
GENERATIONGen X (ages 42-57)	0.94	0.66	1.34	-0.33	0.74
GENERATIONBaby Boomer (ages 58-76)	1.50	1.05	2.15	2.25	0.02
GENERATIONSilent (ages 77+)	2.29	1.45	3.60	3.58	0.00
GENDERFemale	1.03	0.90	1.18	0.44	0.66
RACENEWNon-Hispanic Black	0.81	0.66	1.00	-1.95	0.05
RACENEWHispanic	0.61	0.49	0.75	-4.63	0.00
RACENEWAsian	0.95	0.59	1.54	-0.21	0.84
POLPARTY_NEWDemocrat (no lean)	1.83	1.53	2.19	6.63	0.00
POLPARTY_NEWIndependent (includes leaners)	0.85	0.71	1.02	-1.79	0.07
educrSome college	1.08	0.91	1.29	0.90	0.37
educr4-year degree	1.18	0.94	1.48	1.42	0.15
educrPostgrad work or degree	1.43	1.15	1.79	3.18	0.00
INCOMENEW$30,000-$49,999 per year	0.92	0.72	1.18	-0.62	0.53
INCOMENEW$50,000 - $99,999 per year	1.16	0.93	1.44	1.29	0.20
INCOMENEW$100,000-$149,999 per year	1.34	1.04	1.73	2.27	0.02
INCOMENEW$150,000 or more per year	1.61	1.23	2.09	3.51	0.00
area.live.newLives in small city or suburb	1.07	0.90	1.26	0.75	0.45
area.live.newLives in town or rural area	0.80	0.67	0.95	-2.57	0.01
Standard errors: Robust

Q28_5: I turn to my friends and family

Nearly one-in-five Americans (19%) provided this response.Here are the groups where statistically significant effects were observed:

• Generation While Millennials do not have a statistically significant result, the older cohorts do. All of them are less likely than Gen Z to say they would turn to their friends and family. While the effect sizes seem respectable in the below table – for instance Baby Boomers are 47% less likely than Gen Z to provide this response according to their model (corresponding to an odds ratio of 0.53), the differences are not nearly as dramatic in the crosstabs – so this could be a hard relationship to illustrate using simple descriptive statistics.
• Gender Women are 1.39 times more likely than men to say they will turn to their friends and family to find out the truth. * Racial identity There is no uniform pattern here, so it’s debatable if racial identity as an overall concept plays a role here. However, Hispanics are about 26% less likely than whites to say this. Looking at the crosstabs, however, the difference between the two groups is about a percentage point.

In short, despite some apparent differences highlighted above – differences which, I concede, were identified by a multivariate statistical analysis and thus should hold up to a greater amount of scrutiny – none are really of the order that this analyst would recommend including in the report (this only applies to this particular response option).

Observations	4945
Dependent variable	q28_5
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.06
Pseudo-R² (McFadden)	0.04
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	0.51	0.32	0.81	-2.86	0.00
GENERATIONMillennial (ages 27-41)	0.91	0.62	1.34	-0.48	0.63
GENERATIONGen X (ages 42-57)	0.66	0.44	0.98	-2.05	0.04
GENERATIONBaby Boomer (ages 58-76)	0.53	0.36	0.80	-3.04	0.00
GENERATIONSilent (ages 77+)	0.56	0.35	0.91	-2.32	0.02
GENDERFemale	1.39	1.17	1.65	3.83	0.00
RACENEWNon-Hispanic Black	0.89	0.69	1.16	-0.85	0.40
RACENEWHispanic	0.74	0.57	0.97	-2.20	0.03
RACENEWAsian	0.95	0.55	1.64	-0.20	0.85
POLPARTY_NEWDemocrat (no lean)	0.61	0.49	0.75	-4.58	0.00
POLPARTY_NEWIndependent (includes leaners)	0.62	0.50	0.76	-4.46	0.00
educrSome college	0.87	0.69	1.09	-1.23	0.22
educr4-year degree	1.08	0.82	1.42	0.54	0.59
educrPostgrad work or degree	1.12	0.85	1.47	0.82	0.41
INCOMENEW$30,000-$49,999 per year	0.91	0.67	1.24	-0.61	0.54
INCOMENEW$50,000 - $99,999 per year	0.90	0.69	1.18	-0.74	0.46
INCOMENEW$100,000-$149,999 per year	0.86	0.63	1.17	-0.97	0.33
INCOMENEW$150,000 or more per year	0.92	0.67	1.27	-0.49	0.62
area.live.newLives in small city or suburb	0.92	0.75	1.12	-0.85	0.39
area.live.newLives in town or rural area	0.81	0.66	1.00	-1.99	0.05
Standard errors: Robust

Q28_1: When you are unsure about something in the news and want to find out the truth…I look for information or opinions from people on social media

This option was selected by 15% of Americans. The crosstab analysis suggested this option was one favored by younger Americans, an observaton further confirmed by this analysis. As the below regression output table shows:

• Generation With Gen Z as the reference group, we see the odds ratio (or exp(Est.)) drop closer and closer to 0 with each older age cohort. For Baby Boomers, the odds ratio is 0.17, meaning they are 83% less likely to say this than Gen Z. Or, if we were to flip the reference group between these two, Gen Z would be 4.5 times more likely than Baby Boomers to say this (1/0.17=4.5). A similar relationship is at play for the Silent Generation. Only Millennials come anywhere near Gen Z’s level.
• Racial Identity Interestingly, this is a salient characteristic as well. All the non-white groups are more likely to say this (including Asians, and, yes, this is statistically significant). One can see these differences in the crosstabs, but this analysis suggests they are authentic and not the result of a confounding variable.
• Education For the most part, there are no statistically significant effects, with the exception of people who have a postgrad degree or schooling. They are less likely to check out social media for the purposes of finding out the truth.

Observations	4945
Dependent variable	q28_1
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.11
Pseudo-R² (McFadden)	0.08
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	0.60	0.35	1.02	-1.88	0.06
GENERATIONMillennial (ages 27-41)	0.60	0.41	0.87	-2.72	0.01
GENERATIONGen X (ages 42-57)	0.31	0.21	0.47	-5.50	0.00
GENERATIONBaby Boomer (ages 58-76)	0.17	0.11	0.26	-8.11	0.00
GENERATIONSilent (ages 77+)	0.22	0.12	0.41	-4.92	0.00
GENDERFemale	0.95	0.78	1.15	-0.55	0.58
RACENEWNon-Hispanic Black	1.78	1.36	2.34	4.18	0.00
RACENEWHispanic	1.47	1.11	1.95	2.69	0.01
RACENEWAsian	2.12	1.25	3.60	2.81	0.01
POLPARTY_NEWDemocrat (no lean)	1.03	0.79	1.35	0.24	0.81
POLPARTY_NEWIndependent (includes leaners)	0.98	0.75	1.29	-0.13	0.90
educrSome college	0.96	0.74	1.25	-0.28	0.78
educr4-year degree	0.80	0.57	1.11	-1.32	0.19
educrPostgrad work or degree	0.58	0.41	0.82	-3.10	0.00
INCOMENEW$30,000-$49,999 per year	0.71	0.51	1.00	-1.96	0.05
INCOMENEW$50,000 - $99,999 per year	0.70	0.52	0.96	-2.24	0.02
INCOMENEW$100,000-$149,999 per year	0.70	0.49	0.99	-1.99	0.05
INCOMENEW$150,000 or more per year	0.75	0.52	1.08	-1.56	0.12
area.live.newLives in small city or suburb	1.12	0.89	1.41	0.97	0.33
area.live.newLives in town or rural area	0.90	0.69	1.17	-0.78	0.44
Standard errors: Robust

q28_4 When you are unsure about something in the news and want to find out the truth…I turn to faith or a religious institution, leader, or text

This response option was selected by 9% of Americans. It is also a response option that brings together types of people who generally are not on the same wavelength, in terms of political, media and social matters. This includes:

• Racial Identity Black Americans are slightly more than twice as likely to say the turn to their fath than white Americans, a statistically significant difference. While Gallup certainly has research that generally supports a finding of this nature – Black Americans tend to be more religious than other Americans – one issue is that the apparent differences, as shown in the crosstabs, is a simple percentage point. A discussion of this finding would require bringing in the complexities of regression analysis.
  
• Political Party Democrats are especially unlikely to say this compared to Republicans (all without leaners allocated) – about 87% less likely. Or put another way, Republicans are 7.7 times more likely to say they will turn to their faith. Independents are also unlikely to say they would turn to their faith. This holds up even if we re-run the model to allocate the leaners – in other words, the “true independents,” do not flock to this option either.

Everything else is not significant, with the exception of the highest income bracket, but would not recommend calling that group out.

Observations	4945
Dependent variable	q28_4
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.14
Pseudo-R² (McFadden)	0.11
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	0.15	0.07	0.33	-4.77	0.00
GENERATIONMillennial (ages 27-41)	1.21	0.63	2.33	0.57	0.57
GENERATIONGen X (ages 42-57)	1.13	0.58	2.21	0.35	0.72
GENERATIONBaby Boomer (ages 58-76)	1.13	0.58	2.22	0.36	0.72
GENERATIONSilent (ages 77+)	1.43	0.68	3.04	0.94	0.35
GENDERFemale	1.37	1.07	1.75	2.51	0.01
RACENEWNon-Hispanic Black	2.07	1.40	3.07	3.66	0.00
RACENEWHispanic	1.18	0.80	1.75	0.83	0.41
RACENEWAsian	1.79	0.76	4.21	1.33	0.18
POLPARTY_NEWDemocrat (no lean)	0.13	0.09	0.18	-11.83	0.00
POLPARTY_NEWIndependent (includes leaners)	0.36	0.27	0.48	-6.82	0.00
educrSome college	1.04	0.76	1.41	0.22	0.83
educr4-year degree	1.36	0.89	2.07	1.42	0.16
educrPostgrad work or degree	0.91	0.60	1.40	-0.41	0.68
INCOMENEW$30,000-$49,999 per year	0.78	0.51	1.20	-1.13	0.26
INCOMENEW$50,000 - $99,999 per year	0.85	0.57	1.25	-0.83	0.41
INCOMENEW$100,000-$149,999 per year	0.91	0.58	1.45	-0.39	0.70
INCOMENEW$150,000 or more per year	0.49	0.29	0.83	-2.68	0.01
area.live.newLives in small city or suburb	1.02	0.73	1.40	0.09	0.93
area.live.newLives in town or rural area	1.30	0.97	1.76	1.75	0.08
Standard errors: Robust

q28_7 When you are unsure about something in the news and want to find out the truth…I don’t usually take additional steps to find out the truth

A brutally honest response to this question, it was also selected by less than 1 in 20 Americans (or 4%).

In the standard regression model, few demographic traits or personal characteristics register as statistically significant predictors (for or against). Some levels of education are statistically significant, and in the direction one might expect – higher education means lower likelihood of saying they do not take additional steps to find out the truth. Other demographics or key attitudes that have been instrumental in other portions of this analysis – including generation and political party – are found to have no statistical significance here.

However, there is a twist, please scroll past the first output table…

Observations	4945
Dependent variable	q28_7
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.09
Pseudo-R² (McFadden)	0.08
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	0.05	0.02	0.16	-4.99	0.00
GENERATIONMillennial (ages 27-41)	1.57	0.53	4.66	0.82	0.41
GENERATIONGen X (ages 42-57)	1.25	0.38	4.11	0.37	0.71
GENERATIONBaby Boomer (ages 58-76)	1.24	0.39	3.93	0.37	0.71
GENERATIONSilent (ages 77+)	1.43	0.42	4.93	0.57	0.57
GENDERFemale	1.20	0.84	1.72	1.01	0.31
RACENEWNon-Hispanic Black	1.23	0.73	2.07	0.77	0.44
RACENEWHispanic	0.84	0.44	1.58	-0.55	0.58
RACENEWAsian	1.23	0.29	5.20	0.28	0.78
POLPARTY_NEWDemocrat (no lean)	0.68	0.41	1.10	-1.57	0.12
POLPARTY_NEWIndependent (includes leaners)	1.20	0.76	1.89	0.79	0.43
educrSome college	0.47	0.30	0.74	-3.25	0.00
educr4-year degree	0.56	0.30	1.06	-1.77	0.08
educrPostgrad work or degree	0.17	0.07	0.42	-3.94	0.00
INCOMENEW$30,000-$49,999 per year	0.87	0.52	1.47	-0.51	0.61
INCOMENEW$50,000 - $99,999 per year	0.73	0.44	1.23	-1.19	0.23
INCOMENEW$100,000-$149,999 per year	0.79	0.43	1.45	-0.77	0.44
INCOMENEW$150,000 or more per year	0.28	0.11	0.70	-2.72	0.01
area.live.newLives in small city or suburb	1.06	0.67	1.68	0.26	0.79
area.live.newLives in town or rural area	1.48	0.92	2.38	1.62	0.11
Standard errors: Robust

If, however, the model is changed to allocate the leaners – thus leaving us with a political affiliation variable which has three categories consisting of Republican/lean Republican, Democrat/lean Democrat and “true” independents – a finding suddenly emerges. The regression output table below shows the results for this response option “I don’t usually take additional steps to find out the truth” but uses this alternative political affiliation variable.

As can be seen, true independents are 2.44 times more likely than Republicans/lean Republican to say they do not usually take additional steps to find out the truth. This is the biggest effect of all variables in the model. In the crosstabs, 10% of true independents say they don’t usually take additional steps to find out the truth, compared to 4% of Republicans/lean Republican and 3% of Democrats/lean Democrat.

Observations	5151
Dependent variable	q28_7
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.11
Pseudo-R² (McFadden)	0.09
AIC	NA

	exp(Est.)	S.E.	t val.	p
(Intercept)	0.07	0.46	-5.70	0.00
PartyrDemocrat/leans Democrat	0.66	0.22	-1.91	0.06
PartyrIndependent – no lean	2.44	0.22	4.01	0.00
GENERATIONMillennial (ages 27-41)	0.84	0.42	-0.41	0.68
GENERATIONGen X (ages 42-57)	0.79	0.46	-0.51	0.61
GENERATIONBaby Boomer (ages 58-76)	0.76	0.44	-0.63	0.53
GENERATIONSilent (ages 77+)	0.99	0.49	-0.02	0.98
GENDERFemale	1.17	0.17	0.90	0.37
RACENEWNon-Hispanic Black	1.41	0.24	1.43	0.15
RACENEWHispanic	0.92	0.28	-0.32	0.75
RACENEWAsian	1.18	0.74	0.22	0.82
educrSome college	0.52	0.22	-2.98	0.00
educr4-year degree	0.62	0.29	-1.68	0.09
educrPostgrad work or degree	0.20	0.41	-3.93	0.00
INCOMENEW$30,000-$49,999 per year	0.81	0.25	-0.87	0.39
INCOMENEW$50,000 - $99,999 per year	0.72	0.25	-1.30	0.19
INCOMENEW$100,000-$149,999 per year	0.76	0.30	-0.91	0.36
INCOMENEW$150,000 or more per year	0.41	0.40	-2.22	0.03
area.live.newLives in small city or suburb	1.05	0.22	0.23	0.82
area.live.newLives in town or rural area	1.57	0.22	2.04	0.04
Standard errors: Robust

This concludes the analysis of the Q28 series items.

Sub-Module about Paying for News Content: Sarah F’s section

This section now considers items about the role and priorities of the media from a financial perspective. This analysis will beginw ith Q9 on the questionnaire (K3 in the datafile).

Q9: Have you ever donated money to a news organization or paid to acces news such as paying for a subscriptions or article, or buying anews organizaition?

**Please note that the regression output can be found here "X:_Foundation_data\2022working analysis_q9_paid_for_news_regression.csv"**

This question is named K3 in the datafile. In total, slightly over a quarter of Americans (26%) said they have donated money to a news organization or paid to access the news.

The regression analysis is a logistic regression, predicting the response of “yes.” The :no answer," category (which is only 2% of all responses) has been excluded for this purpose. This analysis is using the standard battery of independent variables, which appear in the regression output below.

A number of different variables are significant in this model, perhaps to a greater degree than virtually any other analysis run thus far. Main findings include (as a reminder the column “exp(Est. )” represent the odds ratio, where a value greater than 1 indicates a higher likelihood to say “yes,” to this question relative to the reference group):

• Generation Recall, the reference category is Gen Z, the youngest group. The odds ratio increases in almost linear like fashion with each older generation – Millennials are 1.6 times more likely than Gen Z to pay for the news, for Gen X it is 1.95 and, with the Baby Boomers, the odds ratio (OR) eclipses the 2.0 mark. All age cohorts have statistically significant effects. Somewhat interestingly, the crosstabs, though, do not show such a simple relationship (Millennials have the highest paid news rate, at 30%), so that is something to consider when writing about this topic.

• Racial identity All effects are statistically significant and suggest that white Americans (who are the reference category and thus do not appear in the below table) are more likely than other racial/ethnic groups to pay for news, a finding that is supported by the crosstab, at least among Blacks and Hispanics.

• Gender Women are about 23% less likely to say they pay for news.

• Political Party Democrats (without leaners being allocated) are more likely to pay for news than “core” Republicans (i.e. without leaners). There is no statistically significant effect with Independents.Note: these results look very similar if instead the analysis is performed using the political party variable which allocates the leaners.

This discussion is continued below the output

## [1] "Have you ever donated money to a news organization or paid to access news, such as paying for a subscription or article, or buying a news magazine?"

Have you ever donated money to a news organization or paid to access news, such as paying for a subscription or article, or buying a news magazine?

QTAG	Wording	response.option	weighted_percentage	unweighted_n	weighted_n
K3	Have you ever donated money to a news organization or paid to access news, such as paying for a subscription or article, or buying a news magazine?	Yes	26	1636	1475
K3	Have you ever donated money to a news organization or paid to access news, such as paying for a subscription or article, or buying a news magazine?	No	72	3882	4031
K3	Have you ever donated money to a news organization or paid to access news, such as paying for a subscription or article, or buying a news magazine?	No answer	2	75	87

Observations	4904
Dependent variable	q9.yes
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.19
Pseudo-R² (McFadden)	0.12
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	0.07	0.04	0.11	-10.15	0.00
GENERATIONMillennial (ages 27-41)	1.61	1.07	2.41	2.29	0.02
GENERATIONGen X (ages 42-57)	1.95	1.28	2.96	3.13	0.00
GENERATIONBaby Boomer (ages 58-76)	2.05	1.35	3.12	3.35	0.00
GENERATIONSilent (ages 77+)	2.28	1.39	3.74	3.26	0.00
GENDERFemale	0.77	0.66	0.90	-3.36	0.00
RACENEWNon-Hispanic Black	0.54	0.43	0.68	-5.18	0.00
RACENEWHispanic	0.60	0.47	0.77	-4.05	0.00
RACENEWAsian	0.46	0.29	0.75	-3.17	0.00
POLPARTY_NEWDemocrat (no lean)	2.07	1.69	2.53	7.04	0.00
POLPARTY_NEWIndependent (includes leaners)	1.08	0.88	1.34	0.75	0.45
educrSome college	2.17	1.74	2.69	6.94	0.00
educr4-year degree	3.25	2.50	4.24	8.73	0.00
educrPostgrad work or degree	4.01	3.13	5.16	10.88	0.00
INCOMENEW$30,000-$49,999 per year	1.27	0.93	1.73	1.53	0.13
INCOMENEW$50,000 - $99,999 per year	1.52	1.16	2.00	3.00	0.00
INCOMENEW$100,000-$149,999 per year	1.29	0.95	1.75	1.64	0.10
INCOMENEW$150,000 or more per year	2.51	1.84	3.44	5.75	0.00
area.live.newLives in small city or suburb	0.95	0.80	1.14	-0.54	0.59
area.live.newLives in town or rural area	0.80	0.67	0.97	-2.29	0.02
Standard errors: Robust

• Education Using the custom four-category version of this variable, we see an increasing likelihood in previously paying for the news as we move up the educational attainment ladder. Indeed, this is arguably the most influential variable for this item – as the large odds ratio values suggest. Postgrads, for instance, are 4.01 times more likely than the reference category – those with no more than a high school diploma – to have paid for news.

Table Q9/K3 below shows the crosstab between this custom education variable and the original survey item, though only the “yes” response option is displayed (however the percentages were calculated using the full base of respondents). It further shows this gulf in paying for news in educational attainment.

Table Q9/K3. Yes, have paid for news by educational attainment

ind.wording	ind_category	Wording	dep_category	pct
Educational attainment recoded	HS or less	Have you ever donated money to a news organization or paid to access news, such as paying for a subscription or article, or buying a news magazine?	Yes	14
Educational attainment recoded	Some college	Have you ever donated money to a news organization or paid to access news, such as paying for a subscription or article, or buying a news magazine?	Yes	25
Educational attainment recoded	4-year degree	Have you ever donated money to a news organization or paid to access news, such as paying for a subscription or article, or buying a news magazine?	Yes	36
Educational attainment recoded	Postgrad work or degree	Have you ever donated money to a news organization or paid to access news, such as paying for a subscription or article, or buying a news magazine?	Yes	46

Income is also a significant predictor of paying for news, though the effect is not as strong as education. Furthermore, the effect is not consistently significant across the brackets.

As a final note, this analysis did look at the relationship between this question and overall opinion of the news media in the US (Q6 in the questionnaire). Remarkably, attitudes about the news media in the U.S. today have little bearing on whether you have paid for media in the past – the results are fairly steady across the five response options of Q6 (very unfavorable, somewhat favorable, neutral, somewhat favorable and very favorable). These can be seen in Table Q9/K3 by Q6 below.

Q9/K3 by Q6. Ever paid for news by overall opinion of the news media

ind.wording	ind_category	Wording	dep_category	pct
Overall opinion of news media in U.S. today	Very unfavorable	Have you ever donated money to a news organization or paid to access news, such as paying for a subscription or article, or buying a news magazine?	Yes	23
Overall opinion of news media in U.S. today	Somewhat unfavorable	Have you ever donated money to a news organization or paid to access news, such as paying for a subscription or article, or buying a news magazine?	Yes	26
Overall opinion of news media in U.S. today	Neutral	Have you ever donated money to a news organization or paid to access news, such as paying for a subscription or article, or buying a news magazine?	Yes	26
Overall opinion of news media in U.S. today	Somewhat favorable	Have you ever donated money to a news organization or paid to access news, such as paying for a subscription or article, or buying a news magazine?	Yes	34
Overall opinion of news media in U.S. today	Very favorable	Have you ever donated money to a news organization or paid to access news, such as paying for a subscription or article, or buying a news magazine?	Yes	25

Q10 SERIES: Which of the following payment methods have you ever used to access the news? (P3_11A-E in data file)

The Q10 series is only asked of individuals who confirmed they have paid for the news – a group, as we have seen, which constitutes 26% of Americans. This series asks about five items, including:

• Micropayments – 10% of people who have paid for the news said they have used this form of payment to access the news
• Day pass – 5% of news-paying Americans have used this method
  
• Membership – A comparatively more popular option, at 36%.
  
• Subscription – This is by far the most common method, as 86% of Americans have used this.
• Donation – Somewhat more common than one might expect, at 39%.

The regression output appears below, though few interesting, statistically significant relationships were discovered through them (at least to this analyst, please take a closer look to see if something pops). You can also find the output in the normal location. At best, there are a few instances where there is a statistically significant effect for theupper income brackets, but it is inconsistent. Generation can sometimes play a role.

On any account, the regression outputgoes through each of the 5 forms of payment tested.

Q10A. Micropayment Regression Output (outcome of interest is “yes”)

Observations	1446
Dependent variable	Q10A
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.23
Pseudo-R² (McFadden)	0.18
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	0.12	0.03	0.51	-2.91	0.00
GENERATIONMillennial (ages 27-41)	0.62	0.20	1.87	-0.85	0.39
GENERATIONGen X (ages 42-57)	0.70	0.22	2.26	-0.60	0.55
GENERATIONBaby Boomer (ages 58-76)	0.42	0.12	1.41	-1.40	0.16
GENERATIONSilent (ages 77+)	0.08	0.01	0.52	-2.67	0.01
GENDERFemale	1.18	0.78	1.78	0.78	0.43
RACENEWNon-Hispanic Black	3.64	2.08	6.37	4.51	0.00
RACENEWHispanic	1.42	0.77	2.60	1.13	0.26
RACENEWAsian	0.99	0.25	3.96	-0.01	0.99
POLPARTY_NEWDemocrat (no lean)	0.77	0.45	1.32	-0.94	0.35
POLPARTY_NEWIndependent (includes leaners)	1.04	0.58	1.85	0.12	0.90
educrSome college	2.21	1.08	4.56	2.16	0.03
educr4-year degree	1.45	0.56	3.74	0.77	0.44
educrPostgrad work or degree	1.75	0.78	3.94	1.36	0.17
INCOMENEW$30,000-$49,999 per year	0.44	0.16	1.20	-1.60	0.11
INCOMENEW$50,000 - $99,999 per year	0.86	0.39	1.94	-0.35	0.72
INCOMENEW$100,000-$149,999 per year	0.65	0.26	1.61	-0.94	0.35
INCOMENEW$150,000 or more per year	0.62	0.26	1.49	-1.06	0.29
area.live.newLives in small city or suburb	1.06	0.65	1.72	0.24	0.81
area.live.newLives in town or rural area	1.58	0.92	2.70	1.66	0.10
Standard errors: Robust

Q10B. Day pass regression output

Observations	1444
Dependent variable	Q10B
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.23
Pseudo-R² (McFadden)	0.19
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	0.11	0.03	0.48	-2.97	0.00
GENERATIONMillennial (ages 27-41)	0.44	0.12	1.57	-1.27	0.21
GENERATIONGen X (ages 42-57)	0.78	0.19	3.17	-0.35	0.73
GENERATIONBaby Boomer (ages 58-76)	0.37	0.09	1.47	-1.41	0.16
GENERATIONSilent (ages 77+)	0.28	0.05	1.43	-1.53	0.13
GENDERFemale	1.11	0.65	1.91	0.39	0.70
RACENEWNon-Hispanic Black	3.09	1.68	5.69	3.63	0.00
RACENEWHispanic	0.91	0.39	2.11	-0.23	0.82
RACENEWAsian	0.59	0.07	4.79	-0.49	0.63
POLPARTY_NEWDemocrat (no lean)	1.17	0.51	2.67	0.36	0.72
POLPARTY_NEWIndependent (includes leaners)	2.20	1.01	4.79	1.99	0.05
educrSome college	0.71	0.30	1.70	-0.76	0.45
educr4-year degree	1.03	0.35	3.03	0.06	0.95
educrPostgrad work or degree	1.11	0.43	2.86	0.22	0.83
INCOMENEW$30,000-$49,999 per year	0.62	0.20	1.97	-0.81	0.42
INCOMENEW$50,000 - $99,999 per year	0.68	0.27	1.71	-0.81	0.42
INCOMENEW$100,000-$149,999 per year	0.42	0.15	1.19	-1.63	0.10
INCOMENEW$150,000 or more per year	0.30	0.11	0.83	-2.31	0.02
area.live.newLives in small city or suburb	1.67	0.93	3.01	1.70	0.09
area.live.newLives in town or rural area	1.34	0.67	2.69	0.82	0.41
Standard errors: Robust

Q10C. Membership regression output

Observations	1449
Dependent variable	Q10C
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.25
Pseudo-R² (McFadden)	0.15
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	0.37	0.15	0.86	-2.29	0.02
GENERATIONMillennial (ages 27-41)	1.54	0.79	2.98	1.27	0.20
GENERATIONGen X (ages 42-57)	1.29	0.65	2.57	0.73	0.47
GENERATIONBaby Boomer (ages 58-76)	1.15	0.57	2.30	0.39	0.69
GENERATIONSilent (ages 77+)	0.77	0.33	1.77	-0.62	0.54
GENDERFemale	0.80	0.62	1.02	-1.78	0.07
RACENEWNon-Hispanic Black	1.66	1.11	2.47	2.46	0.01
RACENEWHispanic	1.00	0.66	1.51	-0.01	0.99
RACENEWAsian	0.92	0.40	2.10	-0.20	0.84
POLPARTY_NEWDemocrat (no lean)	0.66	0.46	0.93	-2.37	0.02
POLPARTY_NEWIndependent (includes leaners)	0.77	0.53	1.12	-1.38	0.17
educrSome college	1.43	0.94	2.19	1.65	0.10
educr4-year degree	1.18	0.72	1.92	0.65	0.51
educrPostgrad work or degree	1.37	0.86	2.18	1.34	0.18
INCOMENEW$30,000-$49,999 per year	0.63	0.34	1.18	-1.45	0.15
INCOMENEW$50,000 - $99,999 per year	1.63	0.96	2.77	1.80	0.07
INCOMENEW$100,000-$149,999 per year	1.63	0.92	2.89	1.67	0.09
INCOMENEW$150,000 or more per year	1.55	0.88	2.74	1.53	0.13
area.live.newLives in small city or suburb	0.90	0.67	1.21	-0.68	0.50
area.live.newLives in town or rural area	0.98	0.70	1.37	-0.14	0.89
Standard errors: Robust

Q10D. Subscription regression output

Income plays a sizeable role here, as does education.

Observations	1474
Dependent variable	Q10D
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.27
Pseudo-R² (McFadden)	0.20
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	2.57	0.73	9.06	1.47	0.14
GENERATIONMillennial (ages 27-41)	0.77	0.26	2.32	-0.46	0.65
GENERATIONGen X (ages 42-57)	0.89	0.29	2.71	-0.21	0.84
GENERATIONBaby Boomer (ages 58-76)	2.22	0.70	7.01	1.36	0.18
GENERATIONSilent (ages 77+)	6.24	1.23	31.72	2.21	0.03
GENDERFemale	1.27	0.87	1.85	1.22	0.22
RACENEWNon-Hispanic Black	0.44	0.26	0.76	-2.93	0.00
RACENEWHispanic	0.90	0.51	1.58	-0.37	0.71
RACENEWAsian	1.63	0.32	8.40	0.59	0.56
POLPARTY_NEWDemocrat (no lean)	0.69	0.39	1.23	-1.26	0.21
POLPARTY_NEWIndependent (includes leaners)	0.61	0.35	1.09	-1.68	0.09
educrSome college	1.23	0.72	2.12	0.76	0.45
educr4-year degree	1.83	0.93	3.58	1.75	0.08
educrPostgrad work or degree	2.82	1.49	5.33	3.19	0.00
INCOMENEW$30,000-$49,999 per year	1.35	0.67	2.71	0.84	0.40
INCOMENEW$50,000 - $99,999 per year	2.55	1.31	4.95	2.76	0.01
INCOMENEW$100,000-$149,999 per year	3.36	1.65	6.86	3.33	0.00
INCOMENEW$150,000 or more per year	3.24	1.60	6.59	3.26	0.00
area.live.newLives in small city or suburb	0.91	0.58	1.42	-0.41	0.68
area.live.newLives in town or rural area	0.66	0.40	1.08	-1.64	0.10
Standard errors: Robust

Q10E. Donation regression output

Democrats are 3.4 times more likely than Republicans to say they have donated to a news organization – seeming to confirm the popular stereotype of NPR’s popularity with Democrats. Independents (all groups) are also slightly more likely than Republicans to donate, but this disappears if we do allocate the leaners. Notice, income is no longer significant, whereas it has often been the best predictor of using a particular form of payment.

Observations	1452
Dependent variable	Q10E
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.28
Pseudo-R² (McFadden)	0.17
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	0.12	0.05	0.31	-4.36	0.00
GENERATIONMillennial (ages 27-41)	1.50	0.70	3.19	1.05	0.29
GENERATIONGen X (ages 42-57)	1.64	0.75	3.58	1.24	0.22
GENERATIONBaby Boomer (ages 58-76)	1.47	0.67	3.22	0.96	0.34
GENERATIONSilent (ages 77+)	0.63	0.26	1.53	-1.02	0.31
GENDERFemale	0.90	0.70	1.17	-0.77	0.44
RACENEWNon-Hispanic Black	0.83	0.55	1.25	-0.90	0.37
RACENEWHispanic	1.13	0.73	1.77	0.55	0.58
RACENEWAsian	0.88	0.40	1.94	-0.31	0.75
POLPARTY_NEWDemocrat (no lean)	3.41	2.31	5.03	6.18	0.00
POLPARTY_NEWIndependent (includes leaners)	1.57	1.04	2.39	2.13	0.03
educrSome college	1.66	1.06	2.61	2.20	0.03
educr4-year degree	1.66	1.00	2.76	1.95	0.05
educrPostgrad work or degree	1.51	0.94	2.42	1.71	0.09
INCOMENEW$30,000-$49,999 per year	0.95	0.52	1.74	-0.17	0.87
INCOMENEW$50,000 - $99,999 per year	1.09	0.63	1.88	0.30	0.76
INCOMENEW$100,000-$149,999 per year	1.33	0.74	2.38	0.97	0.33
INCOMENEW$150,000 or more per year	1.28	0.71	2.28	0.82	0.41
area.live.newLives in small city or suburb	0.99	0.74	1.33	-0.05	0.96
area.live.newLives in town or rural area	1.15	0.81	1.62	0.77	0.44
Standard errors: Robust

Q11 Would you pay to access the news (K4 in datafile)

This question is now asked of the entire sample. The results seem bleak for the news media – overall, 17% say they would pay to access the news in the future.

Again, this analysis will use a logistic regression, aiming to understand what type of person says “yes,” they are willing to pay for news in the future. Initially we will look at our standard battery of demographic features and personal characteristics. The output for this model has the name “Q11,” in the title in the same folder.

Highlights from the main model are as follows:

• Gender Women are about 30% less likely to say they will pay for news in the future, a finding that is also visdible in the cross-tabulations (there 20% of men say yes and 14% of women say likewise).
  
• Racial identity On the face of it, it appears that all other racial/ethnic groups are less willing to pay for news in the future (with the exception of Asians, as the coefficient is not statistically significant). The crosstabs present a more muddled picture (and, indeed, finds that Asians are the most willing to pay for news in the future, at 33%, well above the 18% of white Americans). Proably best to skip this one.
  
• Political Party Democrats, who already are more likely to have a history for paying for news in the past, are 2.22 times more likely than Republicans to say they will do so in the future.
• Education Again, education has a very strong effect here. Against the reference category of having a high school diploma orless, those with some college are 1.58 times more likely to say they will pay for news in the future; 4-year degree holders are 2.95 times more likely and those who have been inthe postgrad space are 4.5 times more likely.
• Income Less clear relationship here, though the top bracket is 2.04 times more likely than the lowest bracket to say this. This is visible in the crosstabulations – 34% of those making $150,000 or more say they are willing – however, there isn’t much variation between the lower income groups.

Observations	4926
Dependent variable	q11.yes
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.21
Pseudo-R² (McFadden)	0.15
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	0.10	0.05	0.18	-7.81	0.00
GENERATIONMillennial (ages 27-41)	0.84	0.56	1.25	-0.87	0.38
GENERATIONGen X (ages 42-57)	0.71	0.46	1.09	-1.56	0.12
GENERATIONBaby Boomer (ages 58-76)	0.76	0.49	1.17	-1.24	0.22
GENERATIONSilent (ages 77+)	1.03	0.59	1.80	0.10	0.92
GENDERFemale	0.70	0.58	0.84	-3.86	0.00
RACENEWNon-Hispanic Black	0.53	0.40	0.71	-4.25	0.00
RACENEWHispanic	0.72	0.54	0.96	-2.23	0.03
RACENEWAsian	0.90	0.56	1.44	-0.44	0.66
POLPARTY_NEWDemocrat (no lean)	2.22	1.72	2.87	6.15	0.00
POLPARTY_NEWIndependent (includes leaners)	0.98	0.75	1.29	-0.13	0.90
educrSome college	1.58	1.19	2.10	3.18	0.00
educr4-year degree	2.95	2.16	4.02	6.85	0.00
educrPostgrad work or degree	4.54	3.40	6.05	10.30	0.00
INCOMENEW$30,000-$49,999 per year	0.72	0.48	1.08	-1.59	0.11
INCOMENEW$50,000 - $99,999 per year	1.18	0.83	1.68	0.92	0.36
INCOMENEW$100,000-$149,999 per year	1.13	0.78	1.63	0.66	0.51
INCOMENEW$150,000 or more per year	2.04	1.39	2.99	3.66	0.00
area.live.newLives in small city or suburb	0.99	0.81	1.21	-0.07	0.95
area.live.newLives in town or rural area	0.94	0.75	1.18	-0.51	0.61
Standard errors: Robust

A final note on Q11, willingness to access news in the future. This item has a clear relationship with two other core survey items (i.e. not demographics), even if the relationship is not a surprising one.

First, people who have paid for news in the past are much, much more likely to pay for news in the future, at 49% to 6%. This is shown in the below table.

Q11. Willingness to pay for media in the future by having paid in the past

ind.wording	ind_category	Wording	dep_category	pct
Q9 Have you ever donated money to a news organization or paid to access news, such as paying for a subscription or article, or buying a news magazine?	Yes	Would you pay to access news in the future?	Yes	49
Q9 Have you ever donated money to a news organization or paid to access news, such as paying for a subscription or article, or buying a news magazine?	No	Would you pay to access news in the future?	Yes	6

Secondly, this may be the question where negative attitudes about the news media finally “catch up,” to the media. Those who hold a very unfavorable opinion of the news media are around 10 percentage points less likely than those holding some degree of a favorable opinion to say they will pay for news in the future. More to the point, 12% of those with a very unfavorable opinion would consider paying for news in the future.These results are in the below table.

Q11. Would you pay to access news in the future, by overall opinion of news media

ind.wording	ind_category	Wording	dep_category	pct
Overall opinion of news media in U.S. today	Very unfavorable	Would you pay to access news in the future?	Yes	12
Overall opinion of news media in U.S. today	Somewhat unfavorable	Would you pay to access news in the future?	Yes	16
Overall opinion of news media in U.S. today	Neutral	Would you pay to access news in the future?	Yes	16
Overall opinion of news media in U.S. today	Somewhat favorable	Would you pay to access news in the future?	Yes	25
Overall opinion of news media in U.S. today	Very favorable	Would you pay to access news in the future?	Yes	23

Q16 Which describes your opinion: Most news organizations are first and foremost businesses motivated by financial interests and goals

Q16, or K7_1 in the data file, presents a five point scale, but is actually asking respondents to select between two different statements, including “most news organizations are first and foremost businesses, motivated by financial interests and goals.” A person agreeing with this strongly would be coded as 1 and if the persone agreed “somewhat,” the code would be 2.

The alternative statement reads: “Most news organizations are first and foremost civic institutions, motivated by serving the public interest.”

Overall, 45% of Americans said they strongly agreed with the first statement – the one indicating that news organizations are profit-seeking ventures. Another 31% say this strongly describes their opinion – meaning 76% agree with this notion to one extent or another.

Given this broad consensus, this analysis will focus on “predicting,” those who strongly agree that most news organizations are business motivated by financial interests. This, of course, means a logistic regression will again be utilized. Odds ratio values above 1 mean that group is more likely to strongly agree with this sentiment.

The regression output appears below. Important highlights include:

• Generation Gen Z is truly the most likely of all the cohorts to say this. The odds ratio for each generation is below 1 (and for the Silent generation at 0.39), indicating all of these generations are appreciably less likely to say this than Gen Z, controlling for all other factors. But the crosstabs are not misleading here: they show that 53% of Gen Z strongly agree that news organizations are businesses motivated by profit; at least 5-points higher than any other generation.
  
• Gender Women are about 19% less likely than men to say this (corresponding to an OR value of 0.81). At the bivariate level, 49% of men strongly agree with this sentiment compared to 41% of women.
  
• Racial identity Non-Hispanic Blacks are slightly outside the chosen level of significance, meaning the result is not significant – somewhat surprising given how different their results are from the other racial/ethnic groups, especially whites. 48% of whites strongly agree news organizations are businesses seeking profit; just 34% of Blacks say the same. However, it is likely that partisan identification is playing the larger role here.
  
• Political Party Democrats (without leaners) are 67% less likely than Republicans to strongly agree with this statement, an effect which is definitely statistically significant. Independents (with leaners included) register an odds ratio of 0.81, suggesting they are also less likely to say this than Republicans, but the confidence interval is between 0.68 and 0.97. In other words, the confidence interval comes close to the value of 1, and thus we should not read too much into this.
• Education Relationship here is less clear, though it does appear that people with some college or a college degree are more likely to strongly agree with this statement (the effect for postgrad is not significant).
• Income Has significant efffects, but the Odds Ratio bounces up and down. In general, people with a higher household income are more likely to agree with statement, but the pattern is not monotonic.

## 
##    0    1 
## 2897 2545

Q18-Q21 SERIES: Americans’ attitudes about how news organizations can earn money or receive funding

The Q18-Q21 series are (pseudo) multiple response items which ask Americans to consider whether “it is reasonable,’ for news organizations to engage in different activities to make money or, alternatively, what other entity should help foot the bill for these media groups. An analytical difficulty in these questions, at least from a regression standpoint, is that there are a number of”it depends," options for any given question, and respondents were allowed to select more than one of these options. However, those who selected “yes, always,” or “no, never,” were unable to select any other option.

This is one reason these extremes will be the focus of the analysis – though which one of these extremes will differ according to the question, depending on the popularity of the response.

Q18: Do you think it is reasonable for individuals to have to pay for the news they watch or read? (K9A_1 to K9A_4 in datafile)

Overall, 4% of Americans said “yes,” it is always reasonable for individuals to have to pay for the news they watch or read, meaning there are few absolutists within the ranks of this country, at least with respect to paying for news content. Another 67% offered some version of “it depends,” – either on the content or on the cost – while 48% said it is never reasonable for people to pay for these types of services.

This latter category will be the focus of the regression (the “yes, always” group was too small for meaningful results). While the odds ratio will be indicating who is more likely to say “no, never,” to this question, they are also saying something about the types of people who would be relatively more likely to pay for content (either all the time or on a “it depends” basis). So while this analysis is focused on one response category, it can say something about both.

The output predicting the “never,” response appears below (and is saved in a marked CSV file). Here are notable highlights:

• Generation This is a striking, surprising relationship, at least to this analyst. Older generations – especially the Baby Boomers – are much more likely than Gen Z (the reference category, though this would likely be true of Milennials as well) to say “no, never.” Given Gen Z and to a lesser extent Millennials have come of age at a time when free content is far more abundant, whereas paying for newspaper subscriptions or cable (for cable news) would be activities far more ensconced in the memories of the older generations, it is counterintuitive that the older generations are absloutists, in terms of nevery paying for content. This dynamic is also clear in the crosstab for the question (60% of Baby Boomers, for instance, say this). * Gender Women are more likely to say this than men, supporting the 9-point difference on this question in the tabs.
  
• Racial identity There are statistically significant differences here as well. Blacks are 1.41 times more likely than whites to say “no, never,” a relationship which can be seen in the cross-tabs. Hispanics, while having a statistically significant, positive, odds ratio (making them more likely than whites to say this), show no real difference between whites if only looking at the crosstabs.
• Political Party Democrats, meanwhile, are largely not behind the idea of never paying for content – they are 44% less likely than Republicans to provide this response. The coefficient for independents is not significant, though in the tabs this group is also less supportive of never paying for content than Republicans, 56% of whom support the idea.
  
• Education A strong relationship here, with college graduates and post-grads arguably taking the most uniform stand against this idea. Compared to those with a high school degree or less, people with a 4-year degree are 50% less likel to support never paying for content and post grads are 59% less likely. The crosstabs (though using a different measure of education) show this very well: 31% of college graduates say ‘no, never’ while 57% of non-grads say this.

Observations	4930
Dependent variable	q18.never
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.17
Pseudo-R² (McFadden)	0.10
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	1.03	0.67	1.60	0.15	0.88
GENERATIONMillennial (ages 27-41)	1.12	0.78	1.61	0.59	0.56
GENERATIONGen X (ages 42-57)	1.66	1.14	2.41	2.64	0.01
GENERATIONBaby Boomer (ages 58-76)	2.15	1.48	3.13	4.01	0.00
GENERATIONSilent (ages 77+)	1.58	1.02	2.46	2.06	0.04
GENDERFemale	1.41	1.23	1.62	4.82	0.00
RACENEWNon-Hispanic Black	1.39	1.12	1.73	2.98	0.00
RACENEWHispanic	1.33	1.07	1.66	2.56	0.01
RACENEWAsian	0.74	0.45	1.23	-1.15	0.25
POLPARTY_NEWDemocrat (no lean)	0.64	0.54	0.77	-4.90	0.00
POLPARTY_NEWIndependent (includes leaners)	0.88	0.73	1.06	-1.37	0.17
educrSome college	0.79	0.66	0.95	-2.58	0.01
educr4-year degree	0.50	0.40	0.63	-5.81	0.00
educrPostgrad work or degree	0.41	0.32	0.51	-7.92	0.00
INCOMENEW$30,000-$49,999 per year	0.98	0.76	1.26	-0.15	0.88
INCOMENEW$50,000 - $99,999 per year	0.70	0.56	0.88	-3.05	0.00
INCOMENEW$100,000-$149,999 per year	0.80	0.62	1.03	-1.75	0.08
INCOMENEW$150,000 or more per year	0.60	0.45	0.79	-3.62	0.00
area.live.newLives in small city or suburb	1.02	0.86	1.20	0.18	0.85
area.live.newLives in town or rural area	1.02	0.86	1.21	0.26	0.79
Standard errors: Robust

Q19 Do you think it is reasonable for news organiations to make money through advertising on their channels, print editions and websites? (K9B_1 to K9B_5 in file)

Slightly more than three-in-ten Americans (31%) said “yes, always,” to the question of whether it is reasonable for news organizations to make money through advertising, while the bulk of Americans fell somewhere into the “it depends,” camp (it could depend on what was being advertised or on the quantity and length of the advertisements or on the person or the organization behind the advertising). Just 7% said “never.”

While the choice of which response category to focus on is less clear here, the “yes, always,” shows interesting variation in the crosstabs and will be the focal point of the regression output below.

Q19 Reasonable for news organizations to make money through advertising: Yes, always

• Generation Generation X comes out as most supportive of this idea, both compared to Gen Z, but in an overall sense as well. 38% of Gen X’ers say “yes,always,” higher than other cohort.
  
• Gender Women are less keen on this idea than men, being 23% less supportive, after controlling for all other demographics and personal characteristics. In the simple tabs, 275 of wmen say “yes, always,” to 35% of men.
  
• Racial identity None of these differences are significant.
  
• Political Party Democrats are 34% less supportive than Republicans according to the model (OR=0.66). Independents (with leaners grouped in) are also less supportive, though the confidence interval comes very close to 1.
  
• Education Not significant.
  
• Income There is a statistically significant relationship here, but the effect size ebbs and flows as we move up the income laddder. In general, it appears support for allowing news organizations to raise revenue via advertising increases with household income, which can be seen in the crosstabs as well.

## [1] "Q19 Reasonable for news organizations to make money through advertising: Yes, always"

Observations	4929
Dependent variable	q19.always
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.07
Pseudo-R² (McFadden)	0.04
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	0.30	0.18	0.51	-4.53	0.00
GENERATIONMillennial (ages 27-41)	1.35	0.89	2.07	1.41	0.16
GENERATIONGen X (ages 42-57)	2.17	1.40	3.36	3.49	0.00
GENERATIONBaby Boomer (ages 58-76)	1.65	1.06	2.57	2.22	0.03
GENERATIONSilent (ages 77+)	1.48	0.88	2.47	1.48	0.14
GENDERFemale	0.77	0.67	0.89	-3.55	0.00
RACENEWNon-Hispanic Black	1.06	0.85	1.32	0.50	0.62
RACENEWHispanic	0.93	0.74	1.17	-0.62	0.54
RACENEWAsian	0.86	0.53	1.42	-0.57	0.57
POLPARTY_NEWDemocrat (no lean)	0.66	0.55	0.79	-4.49	0.00
POLPARTY_NEWIndependent (includes leaners)	0.82	0.68	0.99	-2.11	0.03
educrSome college	1.03	0.85	1.24	0.28	0.78
educr4-year degree	0.99	0.78	1.27	-0.05	0.96
educrPostgrad work or degree	0.92	0.73	1.15	-0.75	0.45
INCOMENEW$30,000-$49,999 per year	1.29	0.98	1.70	1.81	0.07
INCOMENEW$50,000 - $99,999 per year	1.26	0.98	1.61	1.84	0.07
INCOMENEW$100,000-$149,999 per year	1.45	1.10	1.90	2.67	0.01
INCOMENEW$150,000 or more per year	1.44	1.08	1.92	2.52	0.01
area.live.newLives in small city or suburb	1.09	0.91	1.29	0.93	0.35
area.live.newLives in town or rural area	1.11	0.93	1.33	1.19	0.24
Standard errors: Robust

This analysis was also conducted on the “no,never” response. In general, few variables proved statistically significant, with the interesting exception of gender (women are 1.43 times more likely to say news organizations should never make money via advertising); household income was significant, but simply showed the opposite relationship which was on display with respct to the “yes, always,” option. Political party, though, was not significant when focusing on the “no, never,” category.

Q20 Do you think donations should be used to ensure news is available for everyone free of charge?

This is is K9C_1 to K9C_6 in the data file. There are six different options: “yes, always,” “no, never,” and a number of different versions of “it depends.” About a quarter of Americans (24%) said “yes, always,” while a slightly smaller 21% said “no, never.”

This analysis first focuses on those individuals who said “yes, always” that donations should be used to ensure news is available for everyone free of charge.

Highlights from the modelling include:

• Generation Gen Z is the most supportive of this notion, with Millennials essentially at the same place (and the difference between the two is not significant). Even Gen X does not register a statistically significant difference, even if the percentage saying “yes, always,” falls nominally to 25%. However both of the older generations are much less supportive than younger ones.
  
• Gender Women are 1.32 times more likely than men to support the use of donations “always,” for the purposes of providing free news content.
  
• Racial identity Blacks far more supportive of this idea compared to whites – a finding that is true even when controlling for political party. Hispanics are slightly more supportive, and in a statistically significant manner.
• Political Party Democrats are 1.43 times more likely than Republicans to back the idea of using donations; however, Independents are a wash (Statistically speaking).
• Education Most categories do not have a statistically significant effect.
  
• Income Support sinks like a stone as we ascend the income brackets.

Observations	4917
Dependent variable	q20.always
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.09
Pseudo-R² (McFadden)	0.06
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	0.41	0.26	0.66	-3.74	0.00
GENERATIONMillennial (ages 27-41)	0.94	0.65	1.35	-0.34	0.73
GENERATIONGen X (ages 42-57)	0.77	0.53	1.13	-1.35	0.18
GENERATIONBaby Boomer (ages 58-76)	0.49	0.33	0.72	-3.62	0.00
GENERATIONSilent (ages 77+)	0.45	0.28	0.73	-3.21	0.00
GENDERFemale	1.32	1.12	1.55	3.31	0.00
RACENEWNon-Hispanic Black	1.48	1.18	1.85	3.40	0.00
RACENEWHispanic	1.29	1.02	1.63	2.09	0.04
RACENEWAsian	1.16	0.70	1.92	0.57	0.57
POLPARTY_NEWDemocrat (no lean)	1.43	1.16	1.77	3.33	0.00
POLPARTY_NEWIndependent (includes leaners)	0.96	0.76	1.20	-0.39	0.70
educrSome college	0.96	0.78	1.18	-0.41	0.68
educr4-year degree	1.06	0.82	1.37	0.43	0.66
educrPostgrad work or degree	0.72	0.56	0.94	-2.38	0.02
INCOMENEW$30,000-$49,999 per year	0.89	0.67	1.18	-0.81	0.42
INCOMENEW$50,000 - $99,999 per year	0.72	0.56	0.93	-2.53	0.01
INCOMENEW$100,000-$149,999 per year	0.74	0.55	1.00	-1.97	0.05
INCOMENEW$150,000 or more per year	0.57	0.41	0.77	-3.56	0.00
area.live.newLives in small city or suburb	1.02	0.84	1.23	0.18	0.86
area.live.newLives in town or rural area	0.91	0.74	1.12	-0.93	0.35
Standard errors: Robust

If the focus turns to the “no, never,” response option on the question of using donations, we essentially see the same relationships as above, but in reverse.

Q21 Do you think government fuuncting should be used to ensure news is available for everyone free of charge?

The Q21 series corresponds to K9C_1 to K9D_6 in the datafile. Again, we have six options, but we will look at the extremes: including the 23% who said “yes, always,” but especially the 44% who said “no, never.”

Following the other items in this batch of questions, a logistic regression is set up. The focus is first on “no, never,” which show very interesting variation in the tabs.

Highlights from this analysis include:

• Generation Older generations are much keener on the idea that government funding should “never” be used to ensure news is available for everyone free of charge. The Baby Boomers and Silent Generation are particularly hostile to th idea of the government funding news organizations.
  
• Gender Women are less likely to say this, and this is a statistically significant difference.
• Racial identity Blacks are less supportive of saying the government should “never” support news organizations, a finding that holds after controlling for political partisanship. However, the difference between Hispanics and whites is not statistically significant (though, for Asians, it is, which is rare, but keep in mind this is a small sample size).
• Political Party Results are as dramatic as you would expect, with Democrats 82% less likely than Republicans to say “no, never” with respect to the government’s role in this sphere. Independents are also less liekly than Republicans.
• Income Support increases as household income rises.

Observations	4918
Dependent variable	q21.never
Type	Survey-weighted generalized linear model
Family	quasibinomial
Link	logit

Pseudo-R² (Cragg-Uhler)	0.29
Pseudo-R² (McFadden)	0.18
AIC	NA

	exp(Est.)	2.5%	97.5%	t val.	p
(Intercept)	0.71	0.43	1.16	-1.37	0.17
GENERATIONMillennial (ages 27-41)	0.90	0.60	1.36	-0.50	0.62
GENERATIONGen X (ages 42-57)	1.96	1.28	3.00	3.10	0.00
GENERATIONBaby Boomer (ages 58-76)	2.50	1.63	3.84	4.20	0.00
GENERATIONSilent (ages 77+)	2.75	1.67	4.54	3.97	0.00
GENDERFemale	0.77	0.67	0.89	-3.50	0.00
RACENEWNon-Hispanic Black	0.73	0.58	0.92	-2.70	0.01
RACENEWHispanic	0.88	0.70	1.12	-1.04	0.30
RACENEWAsian	0.43	0.24	0.77	-2.85	0.00
POLPARTY_NEWDemocrat (no lean)	0.18	0.15	0.22	-17.35	0.00
POLPARTY_NEWIndependent (includes leaners)	0.42	0.35	0.51	-9.13	0.00
educrSome college	1.23	1.02	1.50	2.14	0.03
educr4-year degree	0.85	0.67	1.09	-1.25	0.21
educrPostgrad work or degree	0.87	0.68	1.10	-1.16	0.25
INCOMENEW$30,000-$49,999 per year	1.90	1.43	2.53	4.43	0.00
INCOMENEW$50,000 - $99,999 per year	2.13	1.64	2.78	5.63	0.00
INCOMENEW$100,000-$149,999 per year	2.81	2.09	3.77	6.88	0.00
INCOMENEW$150,000 or more per year	2.34	1.73	3.17	5.47	0.00
area.live.newLives in small city or suburb	0.99	0.83	1.19	-0.07	0.94
area.live.newLives in town or rural area	1.33	1.11	1.60	3.08	0.00
Standard errors: Robust

An analysis on the “yes always,” shows similar relationships but in reverse.

Current stopping point