ArticlePDF Available

Biases in Fiscal Multiplier Estimates


Abstract and Figures

The size of fiscal multipliers is intensively debated as large (small) multipliers provide arguments to expand (cut) public spending. We use data on multiplier estimates from over a hundred scholarly studies, and ask whether the national imprint and various incentives that the authors face can help explain the large observed variance in these estimates. We complement this meta-analytical data with information on economists’ personal characteristics collected from their biographies and through a self-conducted survey. Our evidence is consistent with the hypothesis that national background and policy orientation of researchers matter for the size of multiplier estimates. We only find weak support for the hypothesis that the interests of donors financing the research are relevant. Significant biases largely disappear for teams of international co-authors.
Content may be subject to copyright.
Journal Pre-proof
Biases in Fiscal Multiplier Estimates
Zareh Asatryan, Annika Havlik, Friedrich Heinemann, Justus Nover
PII: S0176-2680(20)30009-4
Reference: POLECO 101861
To appear in: European Journal of Political Economy
Received Date: 15 July 2019
Revised Date: 18 February 2020
Accepted Date: 20 February 2020
Please cite this article as: Asatryan, Z., Havlik, A., Heinemann, F., Nover, J., Biases in Fiscal Multiplier
Estimates, European Journal of Political Economy,
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.
© 2020 Published by Elsevier B.V.
Biases in Fiscal Multiplier Estimates
This version: February 24, 2020
The size of fiscal multipliers is intensively debated as large (small) multi-
pliers provide arguments to expand (cut) public spending. We use data on
multiplier estimates from over a hundred scholarly studies, and ask whether
the national imprint and various incentives that the authors face can help
explain the large observed variance in these estimates. We complement this
meta-analytical data with information on economists’ personal characteris-
tics collected from their biographies and through a self-conducted survey.
Our evidence is consistent with the hypothesis that national background and
policy orientation of researchers matter for the size of multiplier estimates.
We only find weak support for the hypothesis that the interests of donors
financing the research are relevant. Significant biases largely disappear for
teams of international co-authors.
JEL codes: B4, E62, H11
Keywords: fiscal multipliers, meta-analysis, economic policy ideology, funding
bias, publication bias
1 Introduction
In this paper we ask whether the personal beliefs and career incentives of economists
impact their policy-relevant research findings. In particular, we study a large sub-
field of public economics that is interested in the estimation of fiscal multipliers.1
The literature on fiscal multipliers is a relevant testing ground to explore our re-
search question since multipliers can guide the appropriate level and timing of
government interventions, and thus serve as crucial inputs for policy design. More
specifically, we ask whether an author’s national background and economic policy
orientation, the financial sources funding the research, and career-related publica-
tion incentives help explain the size of fiscal multiplier estimates.
We follow Paldam (2018)’s model of an academic economist who maximizes
her utility across several objectives, and hypothesize that the following three types
of potential biases may play a role in explaining research results. First, economic
policy preferences, partly determined by a researcher’s national background, may
influence her empirical findings (see Section 2.2). Saint-Paul (2018, p. 216) ob-
serves that “people seem to adopt views about underlying parameters that are
conducive to the policies they would otherwise favor for ideological reasons”, and
provides anecdotal evidence for such behavior by leading US macro-economists.
Moreover, there are frequent observations that contributions to macro-economic
policy debates are regularly influenced by a national imprint. A country’s specific
economic policy experience, its dominant national schools of thought and its pop-
1Multipliers are typically defined as the ratio of a change in output at a particular horizon
as a response to a change in fiscal policy (see, e.g., Batini et al. 2014, and Figure 1 for the
considered types of changes). For recent work on fiscal multipliers see, e.g., Christiano et al.
(2011); Ilzetzki et al. (2013); Ramey (2011, 2019).
ulation’s shared values may also leave its mark on researchers that originate from
that specific environment (Alesina et al. 2017; Brunnermeier et al. 2016).
A second potential bias may emerge from the interests of the donors that fund
the research. In pharmaceutical research, there is evidence that industry-financed
studies tend to differ from independent analyses by finding more industry-favorable
results (see Section 2.3). Analogous forces could be at work in macro-economic
research if private donors have certain differing views on the appropriate size of
government as opposed to civil servants and politicians that influence funding
decisions of public donors.
A third type of bias is related to the researcher’s career incentives. A well
studied example of such behavior is the phenomenon of publication bias, which
has been documented in various contexts (see Section 2.4). This behavior does not
necessarily have an ideological dimension but may arise if reviewers and editors
discriminate against insignificant or non-surprising findings.
Macro-economic research in general and fiscal multiplier literature in partic-
ular offer an especially promising field to analyze the impact of author and donor
interests in research findings. The flexibility of macro-models opens rich opportu-
nities for authors to vary assumptions on multipliers and Phillips curve trade-offs
in a way that respects the internal consistency of the underlying model and its
coherence with observable data (Saint-Paul 2018). The “credibility revolution”
with its emphasis on (natural) experiments is still in its infancy in macro-economic
research (Leamer 2010). Consequently, as put forward by Kirchg¨assner (2014, p.
1), “there is quite a lot of consensus with respect to microeconomic questions, but
much less with respect to macroeconomic or macro policy questions.” Ioannidis
(2005) predicts that biases will be particularly large in research fields that offer
a great flexibility in designs and analytical models. This condition is clearly ful-
filled for multiplier research where authors have plenty of opportunity (so-called
“researcher degrees of freedom”, Simmons et al. 2011) to cherry-pick the method,
model structure, identification strategy, data, and/or the context, among other
variables of choice. This argument is consistent with Figure 1 where we show that
fiscal multiplier estimates, as collected in the meta-analysis of Gechert (2015)2
from over a hundred scholarly studies, have a very wide distribution.3
Our data is based on this meta-data of fiscal multipliers which we augment by
various author- and funding-related variables. As proxies for national imprint, we
use the government-spending-to-GDP ratio and the level of economic freedom of
the author’s country of origin, and also rely on an author-specific preference indi-
cator derived from a self-conducted survey among the authors of primary studies.
To measure a potential funding bias, we collect data on project financing as well as
study the type of a researcher’s workplace. We study publication bias by testing
for asymmetries in the relation between the precision and the size of the multiplier
estimate, as well as by searching for systematic differences between journal articles
and working papers and those between the findings of tenured and non-tenured
2Glass (1976, p.3) explains a meta-analysis to be the analysis of analyses, and defines the
term as “the statistical analysis of a large collection of [...] results from individual studies for
the purpose of integrating the findings”. Anderson and Kichkha (2017); Nelson and Kennedy
(2009); Stanley et al. (2013); Stanley (2001); Stanley and Doucouliagos (2012) present reviews
of meta-analytical methods.
3Multipliers in Figure 1 range from -0.19 to 2.27 at the bottom and top 5 percentiles of
the distribution, respectively. They vary substantially across policy instruments having smaller
average values for transfers (0.39) and tax cuts (0.52), and larger averages for general spending
(0.97) and investment expenditures (1.27).
Figure 1: Distribution of fiscal multipliers on spending and taxes (left) and in-
vestment and transfers (right)
(a) Spending and Tax Reliefs (b) Investment and Transfers
Notes: Fiscal multiplier estimates are taken from Gechert (2015). The histograms ex-
clude outliers outside the interval of [-1.7, 3.4], which is three times the standard de-
viation around the mean value of 0.85 of the total sample. The sample includes 2,283
observations of which 33 are dropped as outliers.
We find evidence that is consistent with the national and ideology bias. We do
not detect any evidence for publication selection in the considered fiscal multiplier
studies, and see only weak support for the hypothesis that funding sources matter.
We then document evidence of factors that amplify or mitigate this distortion.
We find a mitigating effect from co-author monitoring that is most pronounced
for teams of international authors. We can exclude that this result is driven by a
higher ranking of journals for international research teams. Our interpretation is
that authors exert mutual control for professional standards so that author-teams
are better able to contain biases compared to a single author. We also find some
weak evidence that an author’s active role in the media amplifies the ideology
bias. Our interpretation is that media presence – as measured by publishing on
the policy blog VoxEU – tends to identify researchers who have a “mission” and,
hence, a stronger tendency to present research insights that raise public attention.
Three clarifying notes of caution are necessary. First, our approach only pro-
vides a lower bound for the possible presence of biases as our meta-analytical
regression includes controls for the multiplier, model type etc., i.e., methodolog-
ical decisions that may themselves be driven by author incentives. Second, if
evidence for biases can be detected, this does not necessarily point to conscious
manipulations (Kirchg¨assner 2014). Instead, a researcher’s ideological position
or self-interests could impact her choice of modeling or testing approaches. Ad-
ditionally, certain priors could unconsciously affect the author’s trust in differing
findings, thus creating a bias of judgment in the selection of results. Moreover, a
financing and a publication bias can be largely outside authors’ responsibility as
donors and editors/reviewers may select researchers on the basis of their (past)
results. The third caveat refers to causality. Our data structure does not offer an
opportunity to exploit a natural experiment. Endogeneity concerns differ across
hypotheses: they are more pronounced for the funding source but are smaller for
the national background. The national background though is still associated with
various other dimensions than just economic policy orientation, and we cannot ex-
clude that omitted variables drive the results. Thus, we are able to show to which
extent observable correlations are consistent with the existence of biases (and their
antidotes) without claiming the identification of the causal channel. We continue
to discuss the relevance and impact of a possible endogeneity for several of the key
findings in the presentation of detailed results.
2 Relevant Literature
2.1 Selective publication of research results
The first related strand of literature comprises of contributions that question the
neutrality of quantitative research in economics. In an anonymous survey among
members of the European Economic Association, almost half of respondents ad-
mit to present empirical findings selectively to confirm one’s argument (Necker
2014). Over recent years, meta-analytical approaches have indirectly confirmed
that empirical researchers benefit from considerable discretion in selecting results
and that they use this freedom. The direction of selection bias will correspond to
author interests along various dimensions, such as successful publications, financial
interests, and intrinsic or ideological motives (Ioannidis 2005).
Employing a collection of 159 meta-analyses from all fields of economics, Ioan-
nidis et al. (2017) show that most research designs suffer from low power and a
consequential bias towards exaggerated effect sizes that provide significant find-
ings despite low power. The same argument is made by Vasishth et al. (2018)
who claim that publishing these low powered findings arouse a wrong impression
of replicability and are a main contributor to the current “replicability crisis”.
Concerns are not necessarily reduced by the tendency towards more credible iden-
tification: Young (2017) claims that one third to one half of instrumental variable
point estimates are wrongly declared statistically significant and replications of lab
experiments are frequently unable to reproduce original findings (Camerer et al.
2016). In general, replications that cross-check published empirical results are still
very rare and largely limited to high-impact articles (Mueller-Langer et al. 2019).
Recent articles propose alternatives to the concept of statistical significance that
should make estimates more credible (see, e.g., Gelman and Carlin 2014; McShane
et al. 2019).
The problem of exaggerated research results is especially problematic if policy-
makers want to rely on these results for the implementation of new policies.
Doucouliagos et al. (2018) tackle a related problem, which is challenging for policy-
makers, that is “thin evidence”. This term describes the issue that for new policy
issues, only very few studies exist when the issue is pressing and the policy-maker
has to make a decision which policy to choose and how.
With a closer link to our fiscal policy research question, Gr¨undler and Po-
trafke (2019) look into the biased use of fiscal terminology. The authors demon-
strate that terms like “austerity” and “fiscal consolidation” are used unequally
across researchers: the former is frequently contained in articles of heterodox jour-
nals, whereas mainstream journals predominantly publish articles using the term
“fiscal consolidation”. The authors conclude from their analysis that the various
degrees of freedom on the choice of methods and austerity measures allow “schol-
ars to arrive at any desired results about the economic effects of austerity periods”
(Gr¨undler and Potrafke 2019, p. 3).
2.2 Impact of prior beliefs
What has been established in the literature a long time ago, is the evidence on
partisan politics, i.e., how different political parties have different preferences for
different sets of public policies and government size (Tufte 1978). Leftist gov-
ernments often have larger public expenditure (Cusack 1997), prefer capital over
labor taxes (Angelopoulos et al. 2012), and spend more on social expenditures
(Herwartz and Theilen 2014). Relatedly, Baskaran (2011) finds that decentral-
ization leads to higher aggregate public expenditure when the federal government
is leftist instead of right-wing. With respect to individual politicians, Jochim-
sen and Thomasius (2014) show how a finance minister’s beliefs and educational
background impact on her country’s deficit performance.
In an analogous way, political preferences may also impact on economic re-
searchers who may use their discretion to produce research consistent with prior
beliefs and a self-serving bias. A first relevant finding is that the significant differ-
ences in perceptions, beliefs, and economic policy preferences between economists
and non-economists can to some extent be explained by the typical socio-economic
status of economists (Blendon et al. 1997; Caplan 2002): economists may have
a more optimistic perspective on trade, liberalized labor markets, small govern-
ments, and low taxes because they are typically part of an affluent population
segment with above-average incomes and low labor market risks.
Apart from the fact that economists constitute a specific sample of the popu-
lation, the role of researchers’ ideological biases has gained greater attention. For
example, in the empirical literature on the deterrence effect of the death penalty,
Kirchg¨assner (2014) finds evidence of an impact from prior political convictions
on results, even if advanced statistical methods are applied.
The link between national tradition and economic policy preferences is an-
other field with mounting evidence on the importance of economists’ prior beliefs.
These aspects have received considerable attention in a macro-context since the
outbreak of the euro area debt crisis. The discussion on how to appropriately
manage the crisis has revealed systematically different views among both politi-
cians and economists from Northern and Southern Europe. Alesina et al. (2017)
show that the economic integration in Europe has not reduced the heterogeneity
of general norms (including the value of hard work or obedience). Guiso et al.
(2016) analyze the Greek crisis and identify a cultural clash as a fundamental
cause. Dyson (1999) and Brunnermeier et al. (2016) recognize a substantive di-
vide in economic policy approaches between both economists and politicians in
France and Germany. While German “ordoliberalism” stresses the importance
of rules over discretion and recommends structural reforms and budgetary con-
solidation, French economic policy prescriptions are often more Keynesian and
demand-oriented. Hien (2019) puts an emphasis on religious influence and iden-
tifies the dividing line between Northern European Protestantism and its impact
on ordoliberalism and the Catholic and Orthodox Christian denominations in the
South with their emphasis of unconditional solidarity. De Ville and Berckvens
(2015) exploit a survey among European economists on euro area reform prefer-
ences and confirm German economists have distinct positions compared to their
European colleagues.
Overall, this literature indicates that a researcher’s national background is
a promising proxy for her prior beliefs on the appropriate size and role of gov-
ernment because of a country’s specific values, intellectual history and economic
policy experience. However, this literature so far rests on surveys or anecdotal
evidence, and largely concentrates on economic policy preferences. We add a more
substantive dimension as we focus on the link between national background and
actual empirical research results.
2.3 Funding bias
Funding-induced biases have received considerable attention in the pharmacoeco-
nomic discipline through comparisons between publicly- and industry-financed re-
search results on new drugs. Some examples that find a positive correlation be-
tween private sponsorship and a favorable test outcome for the pharmaceutical
company are given by Friedberg et al. (1999), Baker et al. (2003), and Bell et al.
(2006). Bekelman et al. (2003) review that a conflict of interest in biomedical
research through financial relationships between researchers and industry alter re-
search results. This is especially surprising as most of these studies make use of
randomized control trials, which are otherwise thought of as the “gold standard”
of empirical research. For a systematic overview of research on pharmaceutical
industry funding and its impact on study outcomes see Sismondo (2008).
Evidence for funding biases are also seen in media studies and health eco-
nomics. DellaVigna and Hermle (2014) analyze movie reviews by media outlets
which are owned by a production company. Their results suggest the absence of
any bias. With a similar focus but arriving at the opposite conclusion, Dobrescu
et al. (2013) test the independence of book reviews when the author is connected
to a media outlet. Li (2017) scrutinizes the interdependency of experts’ conflict
of interest and the quality of their judgment in the context of peer review at the
National Institute of Health. Her findings suggest the existence of a bias in favor
of projects close to the evaluators’ own research.
We add to this literature on funding bias by applying it to macro-economic
research. In this context, a possible funding bias is unrelated to any specific
private business interest. Instead, it could be the result of public research donors’
incentives and their interest in demonstrating the usefulness of public spending.
2.4 Publication bias
The publication bias is perhaps the most extensively researched bias in empirical
economic research (De Long and Lang 1992). It results from the competitive strive
of authors for scarce space in reputed journals and can be present when referees,
researchers, or editors have an ex-ante preference for statistically significant or
other specific results (see, e.g., Frey (2003) for a discussion of editors’ and referees’
influence on research articles). Doucouliagos and Stanley (2013) find evidence for
widespread and substantial publication bias in the majority of the 87 economic
areas. According to their results, literatures where theory allows for a larger range
of results are less selective. Their explanation is that a high community consen-
sus on a specific theoretical prediction should make editors and reviewers more
critical against empirical results contradicting this dominant theory. At the same
time, they find macroeconomic research to be more prone to a selection bias than
micro-research even when they control for the degree of theory consensus. Ioanni-
dis et al. (2017) cover a wide range of fields of economics including international
economics, labor economics, growth and development, microeconomics, macroeco-
nomics, finance, and public economics. This “meta-meta-analysis” suggests that
publication bias is omnipresent and is closely related to a low power of research
designs that “forces” researchers to search for results until effect sizes are so large
that they reach significance (Stanley and Doucouliagos 2012).
The range of theoretical predictions for the size of fiscal multipliers is large
given the long-standing macro-controversies around the existence of Ricardian
Equivalence and the competition of various macro-models. This theoretical open-
ness for different results in itself should reduce selectivity in published results
(Doucouliagos and Stanley 2013). In his meta-analysis, Gechert (2015) provides
an initial analysis for a publication bias in the multiplier literature. Following
standard meta-analytical methods, he hypothesizes that there should not be sub-
stantial asymmetries in the direction and magnitude of estimates which are very
imprecisely estimated (small value for the inverse of the standard error as a mea-
sure of precision) as compared to the precisely estimated ones. Any systematic
asymmetry around imprecise estimates would indicate that the published estimates
are not a representative sample of the underlying population (Figure 3 replicates
this scatterplot and discusses it in more details). Gechert finds only weak evidence
for a publication bias in the multiplier literature. According to these results, if
such a bias exists at all, it benefits smaller multiplier estimates. This could re-
late to the attraction of “surprising” results. We take this as our starting point
and augment Gechert’s approach through tests for the impact of author-specific
features that approximate different degrees of publication pressure (e.g., tenured
versus non-tenured authors). If surprisingly small multipliers can catch the atten-
tion of reviewers, authors under particular publication pressure have an incentive
into that direction.
3 Hypotheses
Our hypotheses cover three biases that relate to (i) the impact of an author’s
own national and ideological imprint, (ii) donor interests, and (iii) the publication
process. In addition, we take account of the potential bias-enforcing effect of a
researcher’s involvement in media debates and the potential bias-mitigating effect
of co-authorship.
With the first hypothesis, we follow the observation that a researcher’s po-
litical convictions can have an impact on results (Kirchg¨assner 2014; Saint-Paul
2018). This is more likely the higher the political relevance of the parameter un-
der scrutiny. Fiscal multipliers clearly satisfy this condition as they are important
guides for economic policy decisions. In the debate on appropriate macro-economic
policy, large multipliers support additional government spending and an activist
fiscal policy, whereas low or even negative multipliers rather point to the merits of
austerity. Should a researcher have a prior position in this debate, this might affect
her impartiality. The estimates of market-liberal supply-side economists could be
biased downwards while those of pro-government demand-side researchers might
be biased upwards. Our main focus in the search for this ideological impact is
the researcher’s national background. In line with the literature on systematic
differences in economic policy preferences across countries, (see Section 2.2) our
first hypothesis is as follows:
H1: Researchers from countries with a large government and high level of
regulation present larger fiscal multiplier estimates than researchers
from countries with a small government and low regulation.
Our second hypothesis deals with the possible funding bias that results from
donor interests (see Section 2.3). Here, we focus on the interests of governmental
institutions that fund multiplier studies directly through project grants or institu-
tional support. From a Niskanen perspective (Niskanen 1971, 1975), bureaucrats
who steer the use of research funds might have a preference for a large government
and therefore an interest in results supportive of more public spending. Even if
the influence of bureaucrats on the allocation of research money is constrained in
most national systems through a strong role of academic peer review, there re-
mains room for bureaucratic influence. Elected politicians also have an interest in
proving to voters a responsible and effective use of public resources. The emerg-
ing funding bias evolves through two different channels. First, the selection of
researchers may favor those who are more aligned with donor preferences. Second,
grant-receiving researchers might be influenced by donors in their research design
as well as their choice and interpretation of results. Resulting biases are likely to
differ across different types of affiliations and career ambitions (see, e.g., Paldam
2018). Hence, our second hypothesis is:
H2: Government-funded research provides larger fiscal multiplier estimates
than non-government-funded research.
The third hypothesis is about the well-known and frequently documented bias
coming from publication selection (see Section 2.4). Insofar editors and reviewers
prefer significant and/or surprising results, the wealth of published studies may
not be representative of the underlying overall population of estimates. This may
result from both editor/reviewer selection and author behavior if authors do not
write up and submit “uninteresting” results in the first place.
In the context of the multiplier literature, the direction of bias is not obvi-
ous. In this literature, the crucial controversy is on the size rather than on the
significance of multipliers. As mentioned in Section 2.4, Gechert (2015) finds weak
evidence that the publication bias in the multiplier literature is, if anything, nega-
tive. Given these features of the multiplier literature, the search for a publication
bias should target two distinctive symptoms that are first, suspicious asymmetries
in the precision of estimates around the most precise estimate (see Section 2.4)
and, second, a preference for surprisingly small multipliers. The latter can be
detected by comparing results from different author types (e.g., tenured vs. non-
tenured researchers, as both groups differ with respect to publication incentives) or
publication types (e.g., working paper vs. journal article, as a working paper rep-
resents an earlier stage of scientific production before the editor/reviewer selection
sets in). Published articles or researchers with high publication pressure should
provide smaller multipliers. Therefore, we construct the following hypothesis:
H3: Multiplier estimates are subject to a publication bias that leads to
asymmetries in the precision of estimates and, possibly, smaller es-
timates in published studies (compared to working papers) and from
authors with high publication pressure.
As a bias amplifier, we take account of an author’s involvement in media
debates. Any such activity can be taken as signal of a “mission” and, hence, a
stronger policy interest. We expect that the amplifier can potentially be important
for both the ideology bias (H1 ) and the funding bias (H2 ). Researchers with strong
positions in public debates might also be more willing to oversell results with the
help of like-minded external donors. Note that this bias does not define a genuine
direction of bias. Instead, it reinforces an existing primary bias (that originates
from national imprint or funding). Therefore, this amplifier is tested through
an interaction of the media involvement indicator with the proxy for primary
bias. We do not see any theoretical argument to expect an amplifying effect of an
author’s media exposure on publication bias since a high media presence is hardly
an informative indicator for academic career ambitions and, hence, publication
pressure (H3 ). This leads us to the following hypothesis:
H4: Active participation in the media debate on economic policy increases
the effects of country imprint (H1) and financing source (H2) on
multiplier estimates.
Our final hypothesis relates to a potential monitoring effect that originates
from co-authorship. Several papers show that monitoring agents can decrease tax
evasion or corruption.4Moreover, it is a robust finding of the tax morale literature
that singles are more likely to evade taxes than people living in marriage (Alm and
Torgler 2006). The explanation is that close social interactions have a monitoring
function that tends to enforce both written and social norms. In this sense, a
single agent is less constrained compared to an entity of individuals that has to
agree on joint decisions. Likewise, interaction in researcher teams can be expected
to activate professional norms and improve authors’ respect for high scientific
standards. We expect that mutual monitoring in researcher-teams should mitigate
all three biases (ideology, funding, publication). There is a specific modification
for the national imprint hypothesis H1. Here, we would expect a bias-reducing
4See, e.g., Almunia and Lopez-Rodriguez (forthcoming), Avis et al. (2016), Battiston et al.
(2020), Bobonis et al. (2016), Ferraz and Finan (2011), Kleven et al. (2011), and Olken (2007).
effect primarily from international teams. Co-authors from the same country are
less likely to challenge each other on specific views that originate from the common
national background. Again, our considerations for co-authorship does not suggest
a primary bias but only an effect relative to existing primary biases. In contrast
to the amplifier of media involvement, co-authorship should moderate the primary
bias. As with H4, the detection strategy will make use of interaction terms. Hence,
we evaluate the validity of the following hypothesis:
H5: Mutual monitoring from (international) co-authors reduces the biases
related to the hypotheses H1 (national imprint), H2 (donor interests),
and H3 (publication bias).
4 Data and Method
4.1 Data
We construct our database by combining the meta-analytical data from Gechert
(2015) with our own collection of author characteristics. Gechert’s database pro-
vides 1,069 observations on fiscal multiplier values taken from 104 studies from
1992 to 2012, with the majority of studies published from 2007 onwards. These
contributions furthermore come from 171 different authors. The switch from an
estimate perspective (where several authors together provide one estimate) to an
author-estimate perspective (treating each author-estimate combination as one
separate observation) results in a total of 2,283 observations.
We obtain information on authors from hand-collected CVs and personal web-
sites. This allows us to identify the authors’ country in two different definitions:
the country where the author received the highest educational degree, and the
country of work (at the time of publication). From the CVs, we also collect the
institution of employment. From the published (working) papers, we collect infor-
mation about project grants. Summary statistics for the employed variables are
provided in Table A1 in the Appendix. Figure 2 shows the distribution of authors
across the countries represented in our sample.
Figure 2: Country variation
(a) Work country
0 20 40 60 80
Number of authors
(b) Degree country
0 20 40 60 80
Number of authors
Notes: For our sample of multiplier studies, the two graphs show the number of authors
working in country x and having received their highest degree in country x, respectively.
Our proxies to test the impact of national imprint on an author’s ideological
stance according to H1 are the size of government and the degree of economic
freedom. To limit issues from omitted national variables, we also add an author-
specific measure. To obtain this individual score, we conducted a survey among
all authors to learn more about their policy preferences in macro-policy debates.
From mid-February to mid-March 2019, we contacted 159 of the authors5and
received 54 replies (34%).6Figure A1 in the Appendix shows the questionnaire
comprising seven statements on fiscal and monetary policy issues. Researchers
could agree (= 9) or disagree (= 1) with the statements in incremental steps of
1. Based on their responses, we calculate a dummy that classifies an author as
We also employ four further data sources to obtain a proxy for the market
orientation of authors that did not respond to our survey. First, we use data on
petitions signed by economists and classified by Hedengren et al. (2010) as liberty-
augmenting vs. liberty-reducing. Second, we hand-collect data on additional pe-
titions and open letters.8Third, the “(European) IGM Economic Experts Panel”
surveys a large amount of economists on their opinion on different policy topics,
which we use to determine their preference for market-oriented policies.9Finally,
we collect data on campaign contributions for US authors from the Federal Elec-
tion Commission and code a contribution to the Republican Party as an indication
of market-orientation and support for the Democratic Party as a pro-government
5For the remaining 12 authors in our sample, we were either not able to find a valid email
address or the authors passed away.
6We conduct a response analysis to evaluate the possibility of systematic differences between
the samples of (non-)responding authors. While the size of multiplier estimates has no effect on
individual author response, correlations with some study and author characteristics exist. We
control for these variables in all specifications. Results are available upon request.
7We construct the dummy as follows: The coding of responses for questions one and six was
reversed such that a higher response number indicates a more market-oriented attitude for all
questions. The dummy variable was then simply coded as 1 (i.e. relatively market-oriented) if
the resulting average score was above the median value.
8Petition urging Congress not to increase public spending in the light of a possible recession
(USA 2009) and petition for more government spending and tax credits (USA 2010).
signal.10 This approach allows us to code 22 additional authors as market- versus
4.2 Estimation
We conduct a meta-analytical regression analysis to test the hypotheses developed
above. Our dependent variable is the fiscal multiplier as it is derived in the under-
lying primary study. No further normalization is needed as this measure is already
dimensionless and comparable across all studies (Gechert 2015)11 . Our unit of
observation is the author-estimate. Hence, one estimate from an n-author team
provides nobservations. In order to prevent studies with multiple authors to have
a larger weight in the analysis, we weight each observation by the inverse of the
number of authors. The resulting coefficients are weighted least squares estimates.
We specify the estimation model as follows:
Multai =β0+β1Sourceai +γ1Model Xai +γ2T ype Xai
+γ3Country Xai +γ4Xai +εai,
where Multai is our dependent variable and captures the size of the fiscal multi-
plier. The index adenotes the author and iis the particular estimate from this
author. Individual researchers may be the author of several studies and many
articles contain numerous estimates due to different specifications and robustness
checks. The coefficient β1represents our coefficient of interest and measures the
11Fiscal multipliers are defined as the ratio of a change in output at a particular horizon as a
response to a change in fiscal policy (see Figure 1 for the considered types of changes) and are
therefore perfectly comparable.
impact of the bias-inducing source. Sourceai can be author-dependent (e.g., na-
tional background) or study-dependent (e.g., project grant). For hypotheses H4
and H5, we focus on interaction effects in order to assess the substance of the
claimed statements. Model Xai covers controls about the model employed in the
study.12 13 T ype Xai accounts for the type of multiplier (e.g., spending or tax
multiplier) with its obvious relevance for size differentials.14 Country Xai includes
a battery of dummy variables for the country coverage of the underlying study.
Finally, Xai summarizes other controls to account for the time horizon of the study
and peak vs. cumulative fiscal multipliers.15 The error term is clustered at the
study level.
5 Results
5.1 National Imprint (H1 )
To study the impact of an author’s national background, we need an indicator
that rates countries according to how much free-market oriented they are. For
that purpose, we make use of two proxies: the government expenditure-to-GDP
ratio and the Economic Freedom of the World indicator (EFW) provided by the
12These include: RBC, VAR, DSGE, Macro models, and single equation estimation models.
13The type of model may matter substantially for the multiplier size. Due to their assumption
on Ricardian Equivalence and market clearing, RBC models, for example, should provide sys-
tematically smaller estimates than other approaches (Gechert 2015). See Gechert (2015) also
for multiplier estimates for certain reference specifications.
14We include dummies to differentiate between multipliers for: public consumption, public
investment, public military spending, public unspecified spending, tax reliefs to the private sec-
tor, transfers to households, direct public employment, and unspecified tax reliefs or spending
15Summary statistics and descriptions for all variables can be found in Table A1 in the Ap-
Fraser Institute. The EFW is an index that ranges from 0 to 10 where a higher
value reflects more economic freedom.16 We measure the country indicator in the
year of the (working) paper publication.
For the expenditure ratio and the EFW, our national imprint hypothesis pre-
dicts a positive and negative coefficient, respectively. This represents the view
that living in a more pro-market country with a smaller government reflects a
government-skeptical position and causes the fiscal multipliers to be biased to-
wards zero. The underlying assumption is one of revealed preferences: through
the observable size of government and the extent of governmental interference
with market processes, a country’s population reveals its fundamental economic
policy preferences. Thus, we are able to test whether authors, in their research,
are influenced by the overall policy orientation of their country of origin. For
internationally mobile researchers, “country of origin” is, of course, ambiguous.
Therefore, we work with two different definitions: country of work and country of
the highest educational degree.
Besides the two country indicators, we also use our author-specific survey in-
dicator of policy orientation. This author-specific indicator provides a particularly
important robustness check as a correlation between our two country indicators
and the size of multipliers can be driven by omitted national variables. If results
for the author indicator are similar to those for the country indicators, this sig-
nals that the empirical support for H1 is not merely an artifact driven by omitted
national variables.
16For 1950-2000 the index is only available every 5 years. For the years in between we use a
linear interpolation.
Table 1 summarizes regression results for the tests of hypothesis H1. The
dependent variable is the fiscal multiplier estimate. We cluster the standard er-
rors at the study level. Columns (1)-(4), (5)-(8), and (9)-(12) present various
specifications for our three indicators of ideological orientation: expenditure ra-
tio, economic freedom, and our author-specific indicator of ideological orientation,
For the expenditure ratio and economic freedom regressions, we provide two
variants that relate those indicators to either the author’s country of workplace
or education. For both country definitions, an obvious endogeneity exists. First,
economists (or students in economics) might migrate to those countries that offer
a public sector in line with their preferences so that neither the country of educa-
tion, nor of work is truly exogenous. However, this kind of endogeneity does not
compromise our testing strategy. If such a Tiebout-migration does actually charac-
terize economists’ location choices (Tiebout 1956), this would even strengthen the
case that a country’s governmental features are a useful proxy for author ideology.
For the author-specific dimension, we use the indicator that takes our survey
results only, and the augmented one that adds data on campaign donations and
other sources. All specifications are presented with and without country fixed
effects. We always include the full set of control variables accounting for the
type of multiplier, model, country-coverage17, and other features of the underlying
17Country fixed effects refer to author country; country-coverage controls refer to the country
groups that are included in the primary study. The inclusion of the latter is a safeguard against
the risk that a correlation between country origin and country coverage drives the results.
study. Detailed results for all these controls are presented in Table A2 in the
The results are consistent with our hypothesis in all but one specification.
Authors from countries with larger governments or lower economic freedom come
up with larger multipliers. The same holds for authors that are classified as having
a pro-government orientation through our author-specific indicator. With one
exception, all specifications that use country fixed effects are estimated with high
statistical precision. Effect sizes are fairly large: A 10 percentage point increase
in spending-to-GDP ratio increases the fiscal multiplier by 0.07 to 0.47 points
on average (or 8-55% of the mean). A one point increase in the EFW indicator
is associated with a decrease in the multiplier of up to 0.62 points. Lastly, the
average difference between market- and government-oriented researchers amounts
to a magnitude of between 0.1 and 0.21 points in the size of multipliers.
These estimated effects are based on a weighted least squares regression as is
common in meta-analytical studies. Such weighting procedures are used to account
for the heterogeneity in methodological approaches and sample sizes in primary
studies and lead to a more efficient estimation compared to ordinary least squares
(Greene 2003). For our main results, we use weights based on the inverse of the
number of authors per study as described in Section 4.2. As the appropriate choice
of weights is non-trivial in our case, we re-estimate the models in Table 1 based
on an alternative weight specification. Similar to Gechert (2015) and Heinemann
18The point estimates of the controls for the model and multiplier type presented in Table
A2 in the Appendix are in line with the findings of Gechert (2015). For example, relative to
government consumption, investment multipliers are large and tax or transfer multipliers small.
Among the models, RBC approaches tend to arrive at the smallest multipliers as expected.
et al. (2018), we also weight individual multiplier estimates using the inverse of
the share of observations per study in relation to the full sample. By doing so,
each study is given an equal weight. Results are documented in Table A3 in the
Appendix and largely support our above-drawn conclusions.
Another concern regarding the results in Table 1 is their potential sensitivity
to the US since a large share of researchers work or were educated in the US (see
Figure 2). Table A4 in the Appendix reports the exercise on excluding the US
from the analysis. The sample shrinks by 40% and this affects the results to some
degree. Importantly, the direction of the effect is unaffected when excluding the
US for all of the employed ideology measures. Despite the drop in the precision of
estimates, the effect remains to be statistically significant for 2 out of 5 measures.
Table 1: National imprint & individual market orientation
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Variable Fiscal multiplier estimate
Expenditure/GDP 0.6939 2.8127***
(workplace) (0.6663) (1.0205)
Expenditure/GDP 1.1285* 4.6641***
(education) (0.6580) (1.3454)
Economic freedom 0.0659 -0.6207***
(workplace) (0.1113) (0.2110)
Economic freedom -0.0795 -0.5472**
(education) (0.1123) (0.2173)
Dummy: market orientation -0.1583* -0.1011
(survey responses) (0.0892) (0.0763)
Dummy: market orientation -0.1820** -0.2122***
(survey & other sources) (0.0839) (0.0799)
Multiplier type controls × × × × × × × × × × × ×
Other controls × × × × × × × × × × × ×
Model controls × × × × × × × × × × × ×
Country coverage × × × × × × × × × × × ×
Country fixed effects × × × × × ×
Observations 2,250 2,250 2,044 2,044 2,250 2,250 2,044 2,044 794 794 905 905
R-squared 0.2692 0.3348 0.2731 0.3046 0.2677 0.3409 0.2677 0.2935 0.3386 0.4336 0.3512 0.4211
Notes: *** p<0.01, ** p<0.05, * p<0.1. Standard errors are clustered by study. The expenditure-to-GDP ratio (columns 1-4) and the Fraser economic
freedom index (columns 5-8) correspond to the year of publication. Results for the control variables can be found in Table A2 in the Appendix.
5.2 Funding (H2 )
As formulated in H2, the funding bias should lead to higher multiplier estimates
for government-financed studies. In the following, we test this hypothesis for di-
rect funding through research grants and indirectly through institutional funding.
Table 2 summarizes the various specifications. As in the preceding section, all re-
gressions include the full set of control variables as previously described. Likewise,
we again provide specifications with and without country fixed effects.
Columns (1) and (2) present results for the simple distinction between studies
that received any type of project grant (independent of the funding source) or not
(or at least without reporting the funding in the final publication). Studies that
include an explicit reference to an external funding are less frequent and account
for 30 out of 104 studies included. In line with the reasoning above, project grants
seem to have a positive partial correlation with the size of the multiplier. The
point estimates are statistically significant across all specifications.
In order to obtain a more detailed picture, we differentiate between the vari-
ous sources for project grants in columns (3) and (4). These finer-grained results
indicate that grants coming directly from national institutions (either the na-
tional government, the national science funding agencies, or the central banks) are
associated with higher multiplier estimates than the reference case of unfunded
studies. Interestingly, this effect is only precisely estimated for projects funded
by national science agencies and partially for central banks, though not for those
which received a research grant directly from the government. Grants from pri-
vately financed foundations or from research institutes are significantly associated
with smaller multiplier estimates than studies without a grant. This supports
the hypothesis as these donors should not share the same interest in proving the
government to be efficient.19
Table 2: Funding – project grants and workplace
(1) (2) (3) (4) (5) (6)
Variable Fiscal multiplier estimate
Project grant 0.2801*** 0.1754*
(30 out of the 104 studies) (0.1067) (0.1030)
Project grant [ref.: no grant]
National science funding agency 0.5397*** 0.4290**
(0.1232) (0.1813)
Government / ministry 0.0991 0.1944
(0.1462) (0.1703)
European Commission 0.0987 -0.0139
(0.2222) (0.1365)
National central bank 0.2804* 0.2431
(0.1463) (0.1524)
(Research) foundation / institute -0.3582*** -0.3752**
(0.1321) (0.1502)
Workplace [ref.: university]
Government institution 0.0724 0.1283
(0.1047) (0.1039)
Private institution -0.0172 -0.0360
(0.1106) (0.1034)
International organization -0.0210 -0.0420
(0.0799) (0.0903)
Central bank -0.0240 0.0708
(0.0795) (0.0820)
Multiplier type controls × × × × × ×
Other controls × × × × × ×
Model controls × × × × × ×
Country coverage × × × × × ×
Country fixed effects × × ×
Observations 2,250 2,250 2,250 2,250 2,250 2,250
R-squared 0.2852 0.3333 0.3080 0.3446 0.2684 0.3299
Notes: *** p<0.01, ** p<0.05, * p<0.1. Standard errors are clustered by study.
19The coding of identified grants into the 5 categories is depicted in Table A5 in the Appendix.
Columns (5) and (6) look at the impact of government financing when it
is given in an indirect way to finance the researchers’ workplace. The reference
workplace is university employment.20 The positive correlations for government
and the negative for private institutions correspond to our expectations. However,
all estimates lack statistical precision. The weaker link compared to project grants
is not surprising since project grants offer a more direct channel for bureaucratic
and political influence on research outcomes compared to institutional financing.
Overall, we find some evidence for a funding bias. As usual in the funding bias
literature, we abstain from speculating on the direction of causality that drives the
correlation. It is only one possible case that the source of funding has an impact
on the conduct and results of a research project. It may also well be the case that
a researcher’s (prior) work changes her success rate in obtaining external finance.
5.3 Publication Bias (H3 )
We test for a publication bias by means of three approaches. First, we search for
asymmetries in the relationship between an estimate and its precision. Second, we
look for systematic differences between journal articles and working papers. Fi-
nally, we ask whether non-tenured researchers (due to higher publication pressure)
come up with different estimates on average than tenured researchers. We base
our second and third approach on Gechert’s (2015) finding that the publication
bias in the multiplier literature, if it exists, favors smaller estimates. This could
20The coding of workplaces into the 4 categories is depicted in Table A6 in the Appendix.
be explainable by an incentive to provide surprising results in order to convince
reviewers about the added value of another multiplier study.
To assess asymmetries in statistical precision, one would usually rely on the
standard errors of the respective estimate (Doucouliagos and Stanley 2009). How-
ever, we lack information on the standard errors of the employed studies such that,
similar to Gechert (2015) and suggested by Stanley and Doucouliagos (2012), we
rely on the number of observations used to obtain the multiplier estimates as a
second best proxy for precision.21 We start out with a graphical investigation of
the relationship between the multiplier estimates and the underlying number of
observations. Figure 3 depicts a funnel graph with the two variables. No asymme-
tries are visible that would point to the relevance of the publication bias in that
Since any such graphical investigation is prone to subjective
(mis-)interpretation, we additionally rely on an econometric analysis and estimate
the following simple model similar to Doucouliagos and Stanley (2009):
multi=β0+β1f(Ni) + εi,(2)
where f(Ni) are various functions of the number of observations Niwhich where
used to estimate the fiscal multiplier multiin primary study i. The error term
continues to be clustered at the study level. This testing approach relates to the
interpretation of a publication bias as introducing a dependency between the stan-
dard error and the coefficient estimate, which would not exist in the absence of a
21The standard errors of the multiplier estimates are not part of the original meta-analysis
by Gechert (2015). This is because in some studies (i) standard errors were not reported in the
primary study or (ii) multiplier estimates had to be retrieved from graphical representations of
results which did not reveal (sufficient) information on the standard errors.
bias (Stanley and Doucouliagos 2012). Results for these asymmetry test regres-
sions are presented in Table 3. For columns (1)-(4) and (7)-(10), a publication bias
would imply statistically significant coefficients for f(N). For columns (5) and (6),
where also the dependent variable is weighted (by log(N) and N, respectively),
one would observe a statistically significant intercept. Since none of the specifica-
tions suggests the presence of such a bias, we do not find any significant evidence
for its relevance in the underlying fiscal multiplier studies.
Figure 3: Publication bias – funnel graph
0 500 1000 1500 2000 2500
Number of observations
-2 0 2 4
Fiscal multiplier estimates (mean=0.85)
Notes: the vertical red line shows the mean of the multiplier estimates in our sample
which is given by 0.85. The level of observation is study-estimate.
Columns (1) and (2) of Table 4 answer the direct question of whether the size
of multipliers reported in journal articles with their entry barriers differs systemat-
ically from those in working papers or other non-journal publications. Depending
on the inclusion of country fixed effects, the sign switches and results are far from
being statistically significant.
Columns (3) to (6) present regression results that look into the findings of
researchers that, due to safe academic positions, are under lower publication pres-
sure compared to non-tenured colleagues. Non-tenured researchers tend to pro-
duce larger multipliers, which runs against the expectation that the publication
bias would favor smaller multipliers. None of the coefficients is significant. As an
alternative proxy for publication pressure, we look at researchers with a full profes-
sorship versus those without. These results again do not support the hypothesis.
Overall, while we cannot prove the absence of any selection in the publication
process we do not find significant support for a publication bias in the multiplier
Table 3: Publication bias – asymmetries
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Variable Fiscal multiplier estimate
Const. 0.767* 0.862*** 0.979** 0.906*** 0.266 2.363 0.768 0.627 0.579* 0.648***
(0.446) (0.151) (0.469) (0.212) (2.398) (2.190) (0.662) (0.438) (0.337) (0.222)
f(N) 0.012 -0.002 -0.750 -0.919 0.781 0.662*** -0.015 0.001 0.596 0.577
(0.086) (0.009) (2.276) (2.274) (0.488) (0.183) (0.093) (0.013) (1.785) (1.618)
f(N)-specification log(N)N1/log (N) 1/N log(N)N log(N)N1/log(N) 1/N
Multiplier type controls × × × ×
Other controls × × × ×
Model controls × × × ×
Country coverage × × × ×
Country fixed effects × × × ×
Observations 1,616 1,616 1,616 1,616 1,616 1,616 1,616 1,616 1,616 1,616
R-squared 0.0002 0.0005 0.0011 0.0017 0.0300 0.1409 0.3293 0.3292 0.3294 0.3294
Notes: *** p<0.01, ** p<0.05, * p<0.1. Standard errors are clustered by study. The models include all data-based
observations (i.e., excluding non-estimated DSGE, structural Macro and RBC models) which explains the reduced
sample size. For columns (1) - (4) and (7) - (10), the dependent variable is the (unweighted) fiscal multiplier estimate
from the primary studies. For columns (5) and (6), we follow Stanley and Doucouliagos (2012) and also weight the
dependent variable (i.e., the fiscal multiplier estimate) by log(N) and N, respectively.
Table 4: Publication bias – type of publication and tenure
(1) (2) (3) (4) (5) (6)
Variable Fiscal multiplier estimate
Journal publication -0.0085 -0.0069
(34% of studies) (0.0697) (0.0632)
Tenure position -0.0613 -0.0340
(35% of authors) (0.0824) (0.0619)
Full professorship -0.0124 0.0020
(27% of authors) (0.0812) (0.0536)
Multiplier type controls ××××××
Other controls ××××××
Model controls ××××××
Country coverage ××××××
Country fixed effects ×××
Observations 2,250 2,250 1,246 1,246 1,246 1,246
R-squared 0.2671 0.3279 0.2663 0.3470 0.2651 0.3467
Notes: *** p<0.01, ** p<0.05, * p<0.1. Standard errors are clustered by study.
5.4 Media Involvement (H4 )
Hypothesis H4 claims that researchers who actively participate in media debates
push their opinions in their research as well. We would then expect that both
the ideological bias and the funding bias are reinforced. We therefore focus on
the interaction of the H1 - and H2 -related variables with our indicator for media
involvement. We proxy media involvement through an author’s presence on the
VoxEU blog. The blog was set up by the Centre for Economic Policy Research in
June 2007 to promote “research-based policy analysis and commentary by leading
economists”.22 It has become one of the leading platforms for economic policy de-
bates based on academic research. 70 authors in our sample have already published
on VoxEU.23
23From these 70 authors, 34 are non-European in the workplace definition. This number even
rises to 42 in the education definition. We experimented with Twitter activity as an alternative
Table 5 reports the results for the H1 -related ideology variables. The coef-
ficient estimates for the interaction terms show the expected sign for only five of
the eight specifications. For models (1) to (4), a positive coefficient is in line with
our hypothesis and suggests that authors from countries with higher government
spending show even larger multipliers if they are present in the public debate.
With statistical significance, this is the case for the educational definition of au-
thor origin. For the interaction with the economic freedom indicator in models (5)
to (8), we again find the expected (negative) sign for the country of education but
without statistical significance.24 Results are therefore mixed and provide only
indicative evidence for an amplifying effect of media involvement.
Table 6 illustrates the interaction between media involvement and the funding
bias. The findings do not provide for a strong confirmation of the hypothesis.
The key interaction is insignificant for grants in general (columns (1) and (2)).
Looking at specific funding sources (columns (3) and (4)), only the interaction with
private donors yields a significant estimate with the expected sign. Concerning
the workplace perspective (columns (5) and (6)), the interactions rather point
to a bias-moderating role: authors from government institutions who publish on
VoxEU show a diminished tendency to come up with large multipliers.
measure of media involvement. Due to a low number of authors active on this channel, resulting
estimates are inconclusive (obtainable from the authors).
24Figures A2 and A3 plot the overall marginal effect of media involvement on the multiplier
estimates for different levels of government size and economic freedom. We do not have an ex
ante expectation about this overall effect and the graphs also do not show any robust systematic
Table 5: Interaction media involvement with national imprint
(1) (2) (3) (4) (5) (6) (7) (8)
Variable Fiscal multiplier estimate
Expenditure/GDP 0.4495 3.0015***
(workplace) (0.7092) (1.0544)
Expenditure/GDP -0.0209 3.5321***
(education) (0.5833) (1.2530)
Exp/GDP ×VoxEU 0.6113 -0.3664 2.3785** 1.6788*
(1.4052) (0.7636) (1.0220) (0.9144)
Economic freedom 0.0130 -0.6627***
(workplace) (0.1110) (0.2143)
Economic freedom 0.0561 -0.4219*
(education) (0.1069) (0.2142)
Economic freedom ×VoxEU 0.1115 0.1343 -0.2967 -0.2575
(0.2514) (0.1063) (0.1877) (0.1744)
Publication on VoxEU -0.2280 0.2071 -0.9026** -0.6467 -0.8325 -0.9895 2.4024 2.0866
(0.5820) (0.3097) (0.4401) (0.3991) (1.9570) (0.8343) (1.4626) (1.3618)
Multiplier type controls × × × × × × × ×
Other controls × × × × × × × ×
Model controls × × × × × × × ×
Country coverage × × × × × × × ×
Country fixed effects × × × ×
Observations 2,250 2,250 2,044 2,044 2,250 2,250 2,044 2,044
R-squared 0.2702 0.3358 0.2833 0.3085 0.2687 0.3424 0.2743 0.2978
Notes: *** p<0.01, ** p<0.05, * p<0.1. Standard errors are clustered by study.
Table 6: Interaction media involvement with funding
(1) (2) (3) (4) (5) (6)
Variable Fiscal multiplier estimate
Project grant 0.3508** 0.2497*
(0.1592) (0.1458)
Project grant ×VoxEU -0.1022 -0.1318
(0.1621) (0.1489)
Grant category [ref.: no grant]
National science funding agency 0.7021** 0.5360*
(0.3007) (0.3032)
interaction with VoxEU -0.2060 -0.1654
(0.3124) (0.3148)
Government / ministry -0.0254 -0.0433
(0.1378) (0.1465)
interaction with VoxEU 0.1900 0.3187
(0.1986) (0.2147)
European Commission 0.0944 -0.0279
(0.2203) (0.1320)
National central bank 0.1745 0.1252
(0.1650) (0.1801)
(Research) foundation / institute -0.0145 0.0656
(0.1736) (0.1902)
interaction with VoxEU -0.3993* -0.5077*
(0.2271) (0.2664)
Workplace [ref.: university]
Government institution 0.1353 0.1907*
(0.1197) (0.1051)
Private institution 0.0130 0.0067
(0.1156) (0.1132)
International organization -0.0108 -0.0244
(0.0787) (0.0961)
Central bank -0.0032 0.1023
(0.0732) (0.0842)
Government institution ×VoxEU -0.2081 -0.2585*
(0.1432) (0.1354)
Publication on VoxEU 0.0001 0.0378 -0.0048 0.0472 0.0516 0.0792
(0.0859) (0.0630) (0.0829) (0.0600) (0.0727) (0.0581)
Multiplier type controls × × × × × ×
Other controls × × × × × ×
Model controls × × × × × ×
Country coverage × × × × × ×
Country fixed effects × × ×
Observations 2,250 2,250 2,250 2,250 2,250 2,250
R-squared 0.2860 0.3342 0.3109 0.3471 0.2699 0.3324
Notes: *** p<0.01, ** p<0.05, * p<0.1. Standard errors are clustered by study.
5.5 Co-authorship (H5 )
In this final section, we summarize the evidence on the bias-mitigating effects of
co-authorship. We limit the analysis to the ideological bias H1 and the funding
bias H2. Since we could not detect any hint of the existence of a publication
bias, the search for a counterbalancing effect from coauthors becomes redundant.
Like for media involvement above, we are merely interested in the bias-modifying
effect of co-authorship and have no theoretical expectation for its direct impact
on empirical results. Therefore, we focus our attention on the coefficient for the
interaction of co-authorship with the bias-indicators.
We start with the ideological bias. We expect that monitoring from interna-
tional author teams should be particularly effective in mitigating an ideological
bias compared to purely national collaborations. Table 7 makes use of a dummy
for authors coming from different countries. The effect of co-author monitoring is
indeed highly visible and significant through all specifications that assign author
preferences to their country of education. In all cases, the interaction counterbal-
ances the original direction of the bias to a large extent and with a high statistical
significance.25 26
As collaboration of (international) authors appears to be helpful in contain-
ing biases from one’s national background, mutual monitoring and exchange may
equally serve to reduce funding biases. We test for this mitigating effect by in-
teracting our monitoring dummy for the presence of co-authorships with funding
indicators (Table 8). For aggregated (columns (1) and (2)) and disaggregated
project grants (columns (3) and (4)), the monitoring dummy identifies multiple
25Table A7 in the Appendix presents results for a multiple author dummy, irrespective of
country composition. Interactions have the right sign but often fail to be significant.
26To rule out the possibility that international researcher teams publish in higher ranked and
potentially less biased journals, as a robustness check we also control for publication status and
journal impact factor. Our conclusions are robust to this extension. Results are available upon
Table 7: Interaction co-authorship with national imprint, authors from different countries
(1) (2) (3) (4) (5) (6) (7) (8)
Variable Fiscal multiplier estimate
Expenditure/GDP (workplace) 1.1192 2.7700**
(0.8266) (1.0767)
Expenditure/GDP (education) 2.8092*** 5.7708***
(0.9453) (1.4015)
Exp/GDP ×mult. authors -1.7665* 0.2588 -3.3821*** -3.6312***
from different countries (0.9805) (0.9292) (1.0260) (1.1825)
Economic freedom (workplace) -0.0043 -0.6084***
(0.1667) (0.2205)
Economic freedom (education) -0.3752** -0.7213***
(0.1782) (0.2347)
Economic freedom ×mult. authors 0.1458 -0.0378 0.5812*** 0.5034***
from different countries (0.1849) (0.1448) (0.1965) (0.1881)
Multiple authors from 0.6726 -0.0960 1.4144*** 1.5388*** -1.2044 0.2783 -4.5322*** -3.8932**
different countries (0.4272) (0.3961) (0.4337) (0.4764) (1.4379) (1.1177) (1.5442) (1.4864)
Multiplier type controls × × × × × × × ×
Other controls × × × × × × × ×
Model controls × × × × × × × ×
Country coverage × × × × × × × ×
Country fixed effects × × × ×
Observations 2,250 2,250 2,044 2,044 2,250 2,250 2,044 2,044
R-squared 0.2746 0.3349 0.2882 0.3157 0.2707 0.3409 0.2845 0.3051
Notes: *** p<0.01, ** p<0.05, * p<0.1. Standard errors are clustered by study.
authors from different countries. For the workplace definition, the monitoring
variable is a dummy for multiple authors (columns (5) and (6)) or for teams with
at least one co-author working at a non-government institution (columns (7) and
For aggregate grants, no significant impact of co-authorship emerges, though
there is a clear pattern in line with our expectations for the disaggregation. In-
ternational co-authorship out-balances the impact of both government grants and
private grants on the size of multiplier estimates. More specifically, the signs of
the interaction coefficients are reversed compared to the plain effects and the mag-
nitude of the counter-balancing effect is sufficient to neutralize the bias. Similar
clear effects cannot be detected for the workplace analysis. However, the absence
of a moderating effect is of less relevance given that there was no strong evidence
for the existence of a workplace effect at all.
Table 8: Interaction co-authorship with funding
(1) (2) (3) (4) (5) (6) (7) (8)
Variable Fiscal multiplier estimate
Project grant 0.3154** 0.1966
(0.1322) (0.1243)
Project grant ×monitoring -0.1944 -0.0799
(0.1615) (0.1632)
Grant category [ref.: no grant]
National science funding agency 0.6089*** 0.5669***
(0.1320) (0.2059)
interaction with monitoring. -0.8717*** -0.8843**
(0.3279) (0.4146)
Government / ministry 0.1328 0.2537
(0.1732) (0.2047)
interaction with monitoring 0.1957 0.2649
(0.2403) (0.2515)
European Commission 0.0735 -0.0250
(0.2355) (0.1323)
National central bank -0.0400 -0.1774
(0.1207) (0.1409)
(Research) foundation / institute -0.5014*** -0.5389***
(0.1552) (0.1984)
interaction with monitoring 0.3765** 0.3511*
(0.1728) (0.2015)
Workplace [ref.: university]
Government institution 0.1958 0.1997 0.0676 0.0672
(0.1820) (0.1777) (0.1119) (0.1290)
Private institution -0.0048 -0.0332 -0.0173 -0.0399
(0.1202) (0.1053) (0.1111) (0.1034)
International organization -0.0003 -0.0452 -0.0204 -0.0464
(0.0906) (0.0913) (0.0802) (0.0919)
Central bank -0.0402 0.0734 -0.0276 0.0694
(0.0715) (0.0842) (0.0797) (0.0824)
Gov. institution ×monitoring -0.2503 -0.1174 -0.0557 0.1348
(0.1987) (0.1864) (0.1034) (0.1384)
Monitoring variable -0.0497 0.0221 -0.0636 -0.0012 0.0694 -0.0034 0.1177 0.1301*
(0.0864) (0.0825) (0.0880) (0.0830) (0.1255) (0.1072) (0.0787) (0.0687)
Definition monitoring mult. authors from diff. countries mult. authors non-gov. coauthor
Multiplier type controls × × × × × × × ×
Other controls × × × × × × × ×
Model controls × × × × × × × ×
Country coverage × × × × × × × ×
Country fixed effects × × × ×
Observations 2,250 2,250 2,250 2,250 2,250 2,250 2,250 2,250
R-squared 0.2886 0.3335 0.3190 0.3532 0.2710 0.3304 0.2687 0.3308
Notes: *** p<0.01, ** p<0.05, * p<0.1. Standard errors are clustered by study.
Our findings on multiple authors and biases might not necessarily reflect a
causal impact of co-authorship. An alternative explanation is that more biased au-
thors may self-select into single-authorship. No matter which of these mechanisms
drive the results, the essential finding is that multiplier estimates of (international)
author teams tend to show less symptoms of an ideological and funding bias.
6 Conclusion
It is well known that fiscal multiplier estimates vary largely because of different
country contexts, multiplier types, or econometric models. Our contribution is that
we shed light on more subjective reasons behind the observed variance of estimates
related to authors’ ideology, incentive effects of external research funding, and
rules of the academic publication process. We find that the variance of multiplier
estimates can indeed be better explained if we take account of author-specific
We show that a researcher’s economic policy orientation correlates with her
estimated multipliers. This result also survives if we replace the country indica-
tors by individual survey-based author indicators. The latter is important since
our country indicator regressions could obviously suffer from an omitted variable
bias. In addition, our evidence is consistent with the hypothesis that government-
financed projects are associated with larger multipliers. Our analysis does not
detect evidence for a publication bias in this strand of literature. The analysis
delivers some evidence that researchers with an active involvement in media de-
bates are particularly prone to the production of multipliers that support their
prior (national) beliefs on the role of government.
This work underlines the need for great caution and scientific neutrality when
designing research projects. Moreover, it emphasizes that it is important for policy
makers to carefully compare various sources when seeking guidance from empirical
research and to take into account the conditions and schools of thought under which
research projects were conducted.
Our results also suggest that co-authorship in general, and international teams
in particular, are an antidote to the distorting effects of national or funding biases.
If this particular insight from our results can also be applied to other strands of
literature, this would carry great significance for economics in general. Of most
significance in this case is the conclusion that biases, which originate from the nar-
rowness of national debates, might be counterbalanced through more international
collaboration and mutual surveillance of research teams.
Alesina, A., G. Tabellini, and F. Trebbi (2017). Is Europe an optimal political area?
NBER Working Papers 23325.
Alm, J. and B. Torgler (2006). Culture differences and tax morale in the United States
and in Europe. Journal of Economic Psychology 27 (2), 224–246.
Almunia, M. and D. Lopez-Rodriguez (forthcoming). Under the radar: The effects of
monitoring firms on tax compliance. American Economic Journal: Economic Policy.
Anderson, R. G. and A. Kichkha (2017). Replication, meta-analysis, and research syn-
thesis in economics. American Economic Review 107 (5), 56–59.
Angelopoulos, K., G. Economides, and P. Kammas (2012). Does cabinet ideology matter
for the structure of tax policies? European Journal of Political Economy 28 (4), 620
– 635.
Avis, E., C. Ferraz, and F. Finan (2016). Do government audits reduce corruption?
Estimating the impacts of exposing corrupt politicians. NBER Working Paper No.
Baker, C. B., M. T. Johnsrud, M. L. Crismon, R. A. Rosenheck, and S. W. Woods (2003).
Quantitative analysis of sponsorship bias in economic studies of antidepressants. The
British Journal of Psychiatry 183 (6), 498–506.
Baskaran, T. (2011). Fiscal decentralization, ideology, and the size of the public sector.
European Journal of Political Economy 27 (3), 485 – 506.
Batini, N., L. Eyraud, L. Forni, and A. Weber (2014). Fiscal multipliers: Size, deter-
minants, and use in macroeconomic projections. IMF Technical Notes and Manuals
Battiston, P., D. Duncan, S. Gamba, and A. Santoro (2020). Audit publicity and tax
compliance: A natural experiment. The Scandinavian Journal of Economics 122 (1),
Bekelman, J., Y. Li, and C. Gross (2003). Scope and impact of financial conflicts of
interest in biomedical research: A systematic review. JAMA 289 (4), 454–465.
Bell, C. M., D. R. Urbach, J. G. Ray, A. Bayoumi, A. B. Rosen, D. Greenberg, and P. J.
Neumann (2006). Bias in published cost effectiveness studies: Systematic review.
BMJ 332 (7543), 699–703.
Blendon, R. J., J. M. Benson, M. Brodie, R. Morin, D. E. Altman, D. Gitterman,
M. Brossard, and M. James (1997). Bridging the gap between the public’s and
economists’ views of the economy. Journal of Economic Perspectives 11 (3), 105–118.
Bobonis, G. J., L. R. C´amara Fuertes, and R. Schwabe (2016). Monitoring corruptible
politicians. American Economic Review 106 (8), 2371–2405.
Brunnermeier, M. K., H. James, and J.-P. Landau (2016). The Euro and the battle of
ideas. Princeton University Press.
Camerer, C. F., A. Dreber, E. Forsell, T.-H. Ho, J. Huber, M. Johannesson, M. Kirch-
ler, J. Almenberg, A. Altmejd, T. Chan, et al. (2016). Evaluating replicability of
laboratory experiments in economics. Science 351 (6280), 1433–1436.
Caplan, B. (2002). Systematically biased beliefs about economics: Robust evidence of
judgemental anomalies from the survey of Americans and economists on the economy.
The Economic Journal 112 (479), 433–458.
Christiano, L., M. Eichenbaum, and S. Rebelo (2011). When is the government spending
multiplier large? Journal of Political Economy 119 (1), 78–121.
Cusack, T. R. (1997). Partisan politics and public finance: Changes in public spending
in the industrialized democracies, 1955-1989. Public Choice 91, 375–395.
De Long, J. B. and K. Lang (1992). Are all economic hypotheses false? Journal of
Political Economy 100 (6), 1257–1272.
De Ville, F. and D. Berckvens (2015). What do Eurozone academics think about EMU
reform? On broad support and German exceptionalism. Bruges Political Research
Papers 41.
DellaVigna, S. and J. Hermle (2014). Does conflict of interest lead to biased coverage?
Evidence from movie reviews. NBER Working Paper No. 20661.
Dobrescu, L. I., M. Luca, and A. Motta (2013). What makes a critic tick? Connected
authors and the determinants of book reviews. Journal of Economic Behavior &
Organization 96, 85 – 103.
Doucouliagos, C. and T. D. Stanley (2013). Are all economic facts greatly exaggerated?
Theory competition and selectivity. Journal of Economic Surveys 27 (2), 316–339.
Doucouliagos, H., M. Paldam, and T. Stanley (2018). Skating on thin evidence: Impli-
cations for public policy. European Journal of Political Economy 54, 16 – 25.
Doucouliagos, H. and T. D. Stanley (2009). Publication selection bias in minimum-wage
research? A meta-regression analysis. British Journal of Industrial Relations 47 (2),
Dyson, K. (1999). The franco-german relationship and economic and monetary union:
Using europe to ’bind Leviathan’. West European Politics 22(1), 25–44.
Ferraz, C. and F. Finan (2011). Electoral accountability and corruption: Evidence from
the audits of local governments. American Economic Review 101 (4), 1274–1311.
Frey, B. S. (2003). Publishing as prostitution? – Choosing between one’s own ideas and
academic success. Public Choice 116 (1-2), 205–223.
Friedberg, M., B. Saffran, T. Stinson, W. Nelson, and C. Bennett (1999). Evaluation of
conflict of interest in economic analyses of new drugs used in oncology. JAMA 282 (15),
Gechert, S. (2015). What fiscal policy is most effective? A meta-regression analysis.
Oxford Economic Papers 67 (3), 553–580.
Gelman, A. and J. Carlin (2014). Beyond power calculations: Assessing type s (sign)
and type m (magnitude) errors. Perspectives on Psychological Science 9 (6), 641–651.
Glass, G. V. (1976). Primary, secondary, and meta-analysis of research. Educational
researcher 5 (10), 3–8.
Greene, W. H. (2003). Econometric analysis. Pearson Education India.
Gr¨undler, K. and N. Potrafke (2019). Ideologically-charged terminology: Austerity, fiscal
consolidation, and sustainable governance. CESifo Working Paper 7613.
Guiso, L., H. Herrera, and M. Morelli (2016). Cultural differences and institutional
integration. Journal of International Economics 99, 97 – 113.
Hedengren, D., D. B. Klein, and C. Milton (2010). Economist petitions: Ideology re-
vealed. Econ Journal Watch 7 (3), 288–319.
Heinemann, F., M.-D. Moessinger, and M. Yeter (2018). Do fiscal rules constrain fiscal
policy? A meta-regression-analysis. European Journal of Political Economy 51, 69–92.
Herwartz, H. and B. Theilen (2014). Partisan influence on social spending under market
integration, fiscal pressure and institutional change. European Journal of Political
Economy 34, 409 – 424.
Hien, J. (2019). The religious foundations of the European crisis. JCMS: Journal of
Common Market Studies 57 (2), 185–204.
Ilzetzki, E., E. G. Mendoza, and C. A. V´egh (2013). How big (small?) are fiscal multi-
pliers? Journal of Monetary Economics 60 (2), 239–254.
Ioannidis, J. P. A. (2005). Why most published research findings are false. PLOS
Medicine 2 (8).
Ioannidis, J. P. A., T. D. Stanley, and H. Doucouliagos (2017). The power of bias in
economics research. The Economic Journal 127 (605), F236–F265.
Jochimsen, B. and S. Thomasius (2014, 06). The perfect finance minister: Whom to
appoint as finance minister to balance the budget. European Journal of Political
Economy 34, 390–408.
Kirchg¨assner, G. (2014). On the process of scientific policy advice - with special reference
to economic policy. CESifo Working Paper Series 5144.
Kleven, H. J., M. B. Knudsen, C. T. Kreiner, S. Pedersen, and E. Saez (2011). Unwilling
or unable to cheat? Evidence from a tax audit experiment in Denmark. Economet-
rica 79 (3), 651–692.
Leamer, E. E. (2010). Tantalus on the road to asymptopia. Journal of Economic
Perspectives 24 (2), 31–46.
Li, D. (2017). Expertise versus bias in evaluation: Evidence from the NIH. American
Economic Journal: Applied Economics 9 (2), 60–92.
McShane, B. B., D. Gal, A. Gelman, C. Robert, and J. L. Tackett (2019). Abandon
statistical significance. The American Statistician 73 (sup1), 235–245.
Mueller-Langer, F., B. Fecher, D. Harhoff, and G. G. Wagner (2019). Replication studies
in economics - how many and which papers are chosen for replication, and why?
Research Policy 48 (1), 62 – 83.
Necker, S. (2014). Scientific misbehavior in economics. Research Policy 43 (10), 1747 –
Nelson, J. P. and P. E. Kennedy (2009). The use (and abuse) of meta-analysis in
environmental and natural resource economics: An assessment. Environmental and
resource economics 42 (3), 345–377.
Niskanen, W. (1971). Representative government and bureaucracy. Chicago: Al-
Niskanen, W. A. (1975). Bureaucrats and politicians. The Journal of Law and Eco-
nomics 18 (3), 617–643.
Olken, B. A. (2007). Monitoring corruption: Evidence from a field experiment in In-
donesia. Journal of Political Economy 115 (2), 200–249.
Paldam, M. (2018). A model of the representative economist, as researcher and policy
advisor. European Journal of Political Economy 54 (C), 5–15.
Ramey, V. A. (2011). Can government purchases stimulate the economy? Journal of
Economic Literature 49 (3), 673–85.
Ramey, V. A. (2019). Ten years after the financial crisis: What have we learned from
the renaissance in fiscal research? Journal of Economic Perspectives 33 (2), 89–114.
Saint-Paul, G. (2018, January). The possibility of ideological bias in structural macroe-
conomic models. American Economic Journal: Macroeconomics 10 (1), 216–41.
Simmons, J. P., L. D. Nelson, and U. Simonsohn (2011). False-positive psychology:
Undisclosed flexibility in data collection and analysis allows presenting anything as
significant. Psychological Science 22 (11), 1359–1366.
Sismondo, S. (2008). Pharmaceutical company funding and its consequences: A quali-
tative systematic review. Contemporary Clinical Trials 29 (2), 109 – 113.
Stanley, T., H. Doucouliagos, M. Giles, J. Heckemeyer, R. Johnston, P. Laroche, J. Nel-
son, M. Paldam, J. Poot, G. Pugh, and R. Rosenberger (2013). Meta-analysis of
economics research reporting guidelines. Journal of Economic Surveys 27 (2), 390–
Stanley, T. D. (2001). Wheat from chaff: Meta-analysis as quantitative literature review.
Journal of Economic Perspectives 15 (3), 131–150.
Stanley, T. D. and H. Doucouliagos (2012). Meta-regression analysis in economics and
business. Routledge.
Tiebout, C. M. (1956). A pure theory of local expenditures. Journal of Political Econ-
omy 64 (5), 416–424.
Tufte, E. R. (1978). Political Control of the Economy. Princeton University Press.
Vasishth, S., D. Mertzen, L. A. J¨ager, and A. Gelman (2018). The statistical significance
filter leads to overoptimistic expectations of replicability. Journal of Memory and
Language 103, 151–175.
Young, A. (2017). Consistency without inference: Instrumental variables in practical
application. Mimeo: LSE.
Figure A1: Survey questions – overview
Figure A2: Marginal effects plots, interaction media involvement with expendi-
ture level
(a) Country of work
-.6 -.4 -.2 0 .2 .4 .6 .8 1 1.2
Marginal effect of VoxEU publication
0 2 4 6 8 10 12
.3 .4 .5 .6 .7
Gov. expenditure/GDP (country of work)
% of observations
(b) Country of highest degree
-.6 -.4 -.2 0 .2 .4 .6 .8 1 1.2
Marginal effect of VoxEU publication
0 2 4 6 8 10 12
.3 .4 .5 .6 .7
Gov. expenditure/GDP (country of highest degree)
% of observations
Notes: The graphs show the marginal effects of the VoxEU variable on the multiplier
estimates. The regressions include country fixed effects and correspond to specifications
(2) and (4) of Table 5, respectively.
Figure A3: Marginal effects plots, interaction media involvement with economic
(a) Country of work
-.6 -.4 -.2 0 .2 .4 .6 .8 1 1.2
Marginal effect of VoxEU publication
0 2 4 6 8 10 12 14 16 18 20 22 24
6 7 8 9 10
Economic freedom (country of work)
% of observations
(b) Country of highest degree
-.6 -.4 -.2 0 .2 .4 .6 .8 1 1.2
Marginal effect of VoxEU publication
0 2 4 6 8 10 12 14 16 18 20 22 24
6 7 8 9 10
Economic freedom (country of highest degree)
% of observations
Notes: The graphs show the marginal effects of the VoxEU variable on the multiplier
estimates. The regressions include country fixed effects and correspond to specifications
(6) and (8) of Table 5, respectively.
Table A1: Summary statistics
Variable Variable definition Obs Mean Std.Dev. Min Max
Fiscal multiplier estimate Fiscal multiplier estimate 2,250 0.829 0.696 -1.700 3.400
Expenditure/GDP (workplace) Government expenditure to GDP Ratio (country of workplace) 2,250 0.429 0.056 0.216 0.653
Expenditure/GDP (education) Gov. expenditure to GDP ratio (country of highest degree) 2,044 0.417 0.055 0.327 0.653
Economic freedom (workplace) Fraser index of economic freedom (country of workplace) 2,250 7.722 0.344 6.982 8.690
Economic freedom (education) Fraser index of economic freedom (country of highest degree) 2,044 7.835 0.343 6.982 8.443
Economic freedom (year of birth) Fraser index of economic freedom (country and year of birth) 974 5.925 0.824 3.089 7.191
Economic freedom (year of publication) Fraser index of economic freedom (country of birth, year of publication) 1,410 7.369 0.504 5.597 8.414
Dummy: market orientation (survey) Dummy whether intensity score above its median value 794 0.496 0.500 0 1
Dummy: market orientation (survey & Dummy for market orientation measured with survey answers, open letters
and campaign contributions
905 0.470 0.499 0 1
other sources)
CONS Public consumption 2,250 0.177 0.382 0 1
SPEND Unspecified public spending 2,250 0.381 0.486 0 1
INVEST Public investment 2,250 0.0987 0.298 0 1
MILIT Public military spending 2,250 0.0227 0.149 0 1
TAX Tax reliefs to private sector 2,250 0.225 0.418 0 1
TRANS Transfers to households 2,250 0.0502 0.218 0 1
EMPLOY Direct public employment 2,250 0.0222 0.147 0 1
DEF Unspecified tax relief or spending increase 2,250 0.0227 0.149 0 1
group: EU/EMU/OECD Multiplier estimated for a group of EU, EMU, and OECD countries 2,250 0.071 0.257 0 1
group: EU/EMU Multiplier estimated for a group of EU and EMU countries 2,250 0.138 0.345 0 1
group: Ind. & Dev. Multiplier estimated for a group of industrial and developing countries 2,250 0.011 0.105 0 1
group: Dev. Multiplier estimated for a group of developing countries 2,250 0.015 0.120 0 1
single: Ind. (low exp/GDP) Multiplier estimated for a single industrial country (low expenditure/GDP) 2,250 0.507 0.500 0 1
single: Ind. (high exp/GDP) Multiplier estimated for a single industrial country (high expenditure/GDP) 2,250 0.214 0.410 0 1
Subnational governm. Multiplier estimated for a group of subnational gov. entities 2,250 0.014 0.118 0 1
Theoretical/NA Multiplier estimated from a purely theoretical model 2,250 0.029 0.169 0 1
VAR Vector Autoregression Model 2,250 0.406 0.491 0 1
RBC Real Business Cycle Model 2,250 0.052 0.222 0 1
NK DSGE New Keynesian DSGE Model 2,250 0.358 0.480 0 1
MACRO Macro Model 2,250 0.088 0.283 0 1
SEE Model Single Equation Estimation Model 2,250 0.096 0.294 0 1
PEAK Peak Multiplier 2,250 0.302 0.459 0 1
HORIZON Horizon of measurement 2,250 1.687 0.991 0 3.871
HORIZON2Horizon of measurement squared 2,250 3.827 3.407 0 14.99
PEAK×HOR Peak multiplier ×Horizon 2,250 0.330 0.715 0 3.178
PEAK×HOR2Peak multiplier ×Horizon squared 2,250 0.620 1.584 0 10.10
M/GDP (in %) Average Import-to-GDP ratio 2,250 20.84 11.34 6 63
Project grant Study received at least one project grant 2,250 0.185 0.388 0 1
National science funding agency Study received a grant from a nat. science funding agency 2,250 0.093 0.291 0 1
Government / ministry Study received a grant from a government / ministry 2,250 0.057 0.232 0 1
European Commission Study received a grant from the European Commission 2,250 0.036 0.185 0 1
National central bank Study received a grant from a national central bank 2,250 0.040 0.196 0 1
(Research) foundation / institute Study received a grant from a research foundation / institute 2,250 0.037 0.190 0 1
University Working at a university 2,250 0.554 0.497 0 1
Government institution Working at a public institution 2,250 0.074 0.261 0 1
Private institution Working at a private institution 2,250 0.021 0.145 0 1
International organization Working at an international organization 2,250 0.156 0.363 0 1
Central bank Working at a central bank 2,250 0.214 0.410 0 1
Journal publication Refereed journal article 2,250 0.420 0.494 0 1
Tenure position Researcher has a tenure position 1,246 0.660 0.474 0 1
Full professorship Researcher is a full professor 1,246 0.521 0.500 0 1
Publication on VoxEU Author has published on 2,250 0.458 0.498 0 1
Multiple authors Study written by multiple authors from different countries 2,250 0.859 0.348 0 1
Multiple authors from diff. countries (workplace) Study by multiple authors from different countries (workplace) 2,250 0.335 0.472 0 1
Multiple authors from diff. countries (education) Multiple authors who received their highest degree in different countries 2,250 0.469 0.499 0 1
Coauthors not from governm. institution At least one author working at a government and one elsewhere 2,250 0.024 0.154 0 1
Table A2: National imprint & individual market orientation – all covariate estimates
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Variable Fiscal multiplier estimate
Expenditure/GDP 0.6939 2.8127***
(workplace) (0.6663) (1.0205)
Expenditure/GDP 1.1285* 4.6641***
(education) (0.6580) (1.3454)
Economic freedom 0.0659 -0.6207***
(workplace) (0.1113) (0.2110)
Economic freedom -0.0795 -0.5472**
(education) (0.1123) (0.2173)
Dummy: market orientation -0.1583* -0.1011
(survey responses) (0.0892) (0.0763)
Dummy: market orientation -0.1820** -0.2122***
(survey & other sources) (0.0839) (0.0799)
Model Type
(baseline: VAR)
RBC -0.5249*** -0.5205*** -0.5183** -0.5565*** -0.6115*** -0.4956*** -0.5379** -0.5363** -0.8096** -0.9158*** -0.7171** -0.6254***
(0.1928) (0.1500) (0.2134) (0.2005) (0.2127) (0.1592) (0.2217) (0.2155) (0.3370) (0.1868) (0.3166) (0.2117)
NK DSGE -0.1151 -0.2038** -0.0666 -0.1617* -0.1661 -0.2037** -0.0573 -0.1152 -0.0805 0.0448 -0.0328 0.0967
(0.1023) (0.0881) (0.0988) (0.0969) (0.1076) (0.0890) (0.1100) (0.1080) (0.1068) (0.1001) (0.1099) (0.1058)
MACRO 0.1780 0.1787** 0.1395 0.2012** 0.1718* 0.1628* 0.1766* 0.2114** -0.0057 0.2228** 0.0314 0.1518
(0.1075) (0.0874) (0.0971) (0.0859) (0.0994) (0.0854) (0.0904) (0.0986) (0.0892) (0.0987) (0.0920) (0.1117)
SEE Model -0.0496 -0.1191 -0.0186 -0.0480 -0.1273 -0.1112 -0.0455 -0.0717 0.2072 0.4104 -0.1519 -0.1841
(0.1626) (0.1467) (0.1721) (0.1567) (0.1659) (0.1414) (0.1775) (0.1577) (0.2550) (0.2918) (0.2332) (0.2462)
Multiplier Type
(baseline: Governm. Consumption)
SPEND -0.0341 -0.0014 -0.0406 -0.0035 -0.0318 0.0336 -0.0449 -0.0060 0.0229 -0.0378 0.0337 0.0044
(0.0808) (0.0843) (0.0790) (0.0802) (0.0837) (0.0831) (0.0809) (0.0837) (0.0740) (0.0822) (0.0811) (0.1019)
INVEST 0.3788** 0.3492** 0.4206** 0.4093** 0.3803** 0.3534** 0.4161** 0.4178** 0.5530*** 0.5105** 0.6328*** 0.5803***
(0.1538) (0.1561) (0.1826) (0.1822) (0.1538) (0.1574) (0.1842) (0.1851) (0.1883) (0.1926) (0.1778) (0.1904)
MILIT -0.0635 -0.1795 -0.0569 -0.1271 -0.0861 -0.1618 -0.0729 -0.1175 0.5108** 0.3108 -0.0168 -0.0981
(0.1858) (0.1763) (0.1760) (0.1812) (0.1877) (0.1849) (0.1742) (0.1864) (0.1935) (0.2125) (0.2291) (0.2215)
TAX -0.3978*** -0.3551*** -0.3468*** -0.3222*** -0.3967*** -0.3437*** -0.3434*** -0.3317*** -0.4287*** -0.3923*** -0.2038 -0.1866
(0.1072) (0.1132) (0.1121) (0.1127) (0.1070) (0.1117) (0.1141) (0.1144) (0.1231) (0.1384) (0.1306) (0.1431)
TRANS -0.5363*** -0.5467*** -0.4178*** -0.4168*** -0.5374*** -0.5360*** -0.4271*** -0.4131*** -0.6497*** -0.6591*** -0.5115*** -0.5180***
(0.1075) (0.1112) (0.1218) (0.1195) (0.1075) (0.1094) (0.1223) (0.1173) (0.1138) (0.1163) (0.1222) (0.1332)
EMPLOY -0.0086 0.1676 0.0887 0.1807 0.0309 0.1472 0.0644 0.1772 0.1245 0.0837 0.0184 -0.0234
(0.1466) (0.1323) (0.1526) (0.1457) (0.1378) (0.1281) (0.1523) (0.1453) (0.1626) (0.1713) (0.2953) (0.2725)
DEF -0.1207 -0.1312 -0.1436 -0.1354 -0.1087 -0.1298 -0.1349 -0.1083 - - -0.0615 -0.0944
(0.1010) (0.1074) (0.1192) (0.1208) (0.1015) (0.1133) (0.1206) (0.1249) (0.1679) (0.1827)
Country coverage
(baseline: group: EU/EMU/OECD)
group: EU/EMU 0.5346*** 0.4839*** 0.5145*** 0.4585** 0.5608*** 0.5387*** 0.5498*** 0.5180** 0.4363** 0.0222 0.5657*** 0.3687
(0.1710) (0.1744) (0.1898) (0.1768) (0.1716) (0.1786) (0.1937) (0.2050) (0.1721) (0.1423) (0.1994) (0.2235)
group: Ind. & Dev. 0.6463* 0.5743* 0.9964*** 0.9495*** 0.6787* 0.7387** 1.0878*** 1.1461*** 0.4673 0.3866* 1.1081*** 1.2847***
(0.3403) (0.3011) (0.3544) (0.2609) (0.3560) (0.3053) (0.3369) (0.2474) (0.3199) (0.1950) (0.2859) (0.2630)
group: Dev. 0.0862 0.1020 0.1210 0.1276 0.0703 0.1094 0.1137 0.1316 -0.1775 -0.1514* 0.0155 0.0614
(0.1381) (0.1424) (0.1465) (0.1552) (0.1382) (0.1499) (0.1462) (0.1659) (0.1248) (0.0897) (0.1671) (0.1601)
single: Ind. 0.4468** 0.5428*** 0.4225** 0.4743*** 0.4509*** 0.5902*** 0.4346** 0.4961*** 0.2399 0.0786 0.3794** 0.3854**
(low exp/GDP) (0.1707) (0.1675) (0.1746) (0.1800) (0.1709) (0.1716) (0.1745) (0.1877) (0.1609) (0.1184) (0.1841) (0.1929)
single: Ind. 0.3942*** 0.5772*** 0.3891** 0.4771*** 0.4263*** 0.6285*** 0.4018** 0.5047*** 0.2825 0.1798 0.3242* 0.3822*
(high exp/GDP) (0.1500) (0.1594) (0.1558) (0.1681) (0.1529) (0.1653) (0.1531) (0.1810) (0.1702) (0.1160) (0.1933) (0.1999)
Subnational gov. 0.6972 0.8287 0.6801 0.6500 0.7514 0.8365 0.7006 0.6957 -0.4203 -0.8185** 0.0465 0.1477
(0.6220) (0.6244) (0.6182) (0.6139) (0.6174) (0.5965) (0.6100) (0.5727) (0.3816) (0.3199) (0.3345) (0.3843)
Theoretical/NA 0.0021 -0.0350 0.0095 0.0075 -0.0326 -0.0544 -0.0005 -0.0613 -0.2584 -0.1788* -0.0634 0.0087
(0.1461) (0.1382) (0.1559) (0.1478) (0.1493) (0.1449) (0.1583) (0.1643) (0.1580) (0.0914) (0.1698) (0.1486)
Control Variables
(baseline: cumulative multiplier)
PEAK 0.1933 0.2988** 0.1554 0.1757 0.1907 0.3164** 0.1318 0.1516 0.0218 0.0058 0.1751 0.1904*
(0.1403) (0.1357) (0.1389) (0.1342) (0.1368) (0.1321) (0.1379) (0.1333) (0.1306) (0.1007) (0.1293) (0.1071)
HORIZON -0.0668 0.0196 -0.1951 -0.1795 -0.0482 0.0058 -0.2029 -0.1928 -0.1078 -0.0630 -0.0869 -0.0064
(0.1538) (0.1337) (0.1321) (0.1281) (0.1536) (0.1313) (0.1396) (0.1325) (0.1036) (0.1014) (0.1016) (0.1109)
HORIZON20.0488 0.0291 0.0864** 0.0817** 0.0417 0.0347 0.0867** 0.0831** 0.0555* 0.0489* 0.0544* 0.0366
(0.0396) (0.0335) (0.0345) (0.0339) (0.0397) (0.0332) (0.0366) (0.0356) (0.0308) (0.0291) (0.0283) (0.0317)
PEAK×HOR -0.0725 -0.1840 0.0501 0.0576 -0.1239 -0.1591 0.0443 0.0432 0.0874 0.0732 -0.0010 -0.0406
(0.2149) (0.1821) (0.2068) (0.1932) (0.2200) (0.1861) (0.2165) (0.2062) (0.2011) (0.1643) (0.1939) (0.1972)
PEAK×HOR20.0546 0.0827 0.0181 0.0129 0.0721 0.0705 0.0237 0.0201 0.0053 0.0195 0.0276 0.0433
(0.0633) (0.0546) (0.0634) (0.0599) (0.0647) (0.0569) (0.0653) (0.0638) (0.0688) (0.0603) (0.0603) (0.0627)
M/GDP (in %) -0.0106*** -0.0098*** -0.0111*** -0.0126*** -0.0102*** -0.0100*** -0.0111*** -0.0122*** -0.0117*** -0.0089*** -0.0103*** -0.0081***
(country sample) (0.0036) (0.0033) (0.0035) (0.0036) (0.0036) (0.0032) (0.0036) (0.0037) (0.0027) (0.0020) (0.0022) (0.0024)
Constant 0.3513 -0.9314 0.2252 -1.7794** 0.1498 5.0804*** 1.3216 4.7000*** 0.9649*** 0.8979*** 0.6943*** 0.5753***
(0.4023) (0.6158) (0.3325) (0.7415) (0.8827) (1.5367) (0.9543) (1.6219) (0.1640) (0.0996) (0.1904) (0.1918)
Country fixed effects ××××××
Observations 2,250 2,250 2,044 2,044 2,250 2,250 2,044 2,044 794 794 905 905
R-squared 0.2692 0.3348 0.2731 0.3046 0.2677 0.3409 0.2677 0.2935 0.3386 0.4336 0.3512 0.4211
Notes: *** p<0.01, ** p<0.05, * p<0.1. Standard errors are clustered by study. The expenditure-to-GDP ratio (columns 1-4) and the Fraser economic freedom index
(columns 5-8) correspond to the year of publication.
Table A3: National imprint & individual market orientation – alternative WLS
(1) (2) (3) (4) (5) (6)
Variable Fiscal multiplier estimate
Expenditure/GDP 2.2324
(workplace) (1.3729)
Expenditure/GDP 3.5407**
(education) (1.4050)
Economic freedom -0.4910*
(workplace) (0.2582)
Economic freedom -0.6138**
(education) (0.2455)
Dummy: market orientation -0.0420
(survey responses) (0.1109)
Dummy: market orientation -0.3077***
(survey & other sources) (0.0947)
Multiplier type controls × × × × × ×
Other controls × × × × × ×
Model controls × × × × × ×
Country coverage × × × × × ×
Country fixed effects × × × × × ×
Observations 2,250 2,044 2,250 2,044 794 905
R-squared 0.3031 0.2660 0.3097 0.2699 0.3944 0.3708
Notes: *** p<0.01, ** p<0.05, * p<0.1. Standard errors are clustered by study. The
expenditure-to-GDP ratio (columns 1-2) and the Fraser economic freedom index
(columns 3-4) correspond to the year of publication.
Table A4: National imprint & individual market orientation – excluding obser-
vations with the US as a work/education country
(1) (2) (3) (4) (5)
Variable Fiscal multiplier estimate
Expenditure/GDP (workplace) 1.4468
Expenditure/GDP (education) 2.8569**
Economic freedom (workplace) -0.7364*
Economic freedom (education) -0.1388
Dummy: market orientation -0.0880
(survey & other sources) (0.0923)
Multiplier type controls × × × × ×
Other controls × × × × ×
Model controls × × × × ×
Country coverage × × × × ×
Country fixed effects × × × × ×
Observations 1,395 1,099 1,395 1,099 636
R-squared 0.3691 0.3502 0.3761 0.3435 0.5084
Notes: *** p<0.01, ** p<0.05, * p<0.1. Standard errors are clustered by study. The
expenditure-to-GDP ratio (columns 1-2) and the Fraser economic freedom index
(columns 3-4) correspond to the year of publication.
Table A5: Coding of project grants into categories
National science
funding agency
Commission Government Central Bank
Foundation /
National Science
Foundation European Commission
Pierre Werner Chair
Programme on
Monetary Union
Banco D’Espania Stanford Center for
Economic Policy
Social Sciences and
Humanities Research
Council of Canada
World Bank
(Knowledge for World
Fondation Banque de
Barcelona GSE
Research Network
German Research
Foundation Arbeitskammer Wien Sloan Foundation
Irish Research Council
for the Humanities
and Social Sciences
Spanish Ministry of
Education and Science
Centre for
Spanish Ministry of
Science and
Institute for New
Economic Thinking
Table A6: Coding of workplaces into categories
Government institution Private institution International
organization Central Bank
Belgian Federal Planning Bureau Goldman Sachs IMF National central banks (ITA,
Economic Bureau of Spanish Prime
Foundation OECD ECB
European Commission Moody’s Analytics World Bank Federal Reserve System
French Ministry of the Economy and
Federal Bank of Chicago,
Kansas City, Minneapolis,
New York, Chicago
INSEE France
Office of the (US) Vice President
Table A7: Interaction co-authorship with national imprint, multiple authors
(1) (2) (3) (4) (5) (6) (7) (8)
Variable Fiscal multiplier estimate
Expenditure/GDP (workplace) 0.7818 3.3666**
(0.9789) (1.3391)
Expenditure/GDP (education) 3.0650** 6.3738***
(1.4585) (1.7367)
Exp/GDP ×mult. authors -0.1074 -0.3336 -2.6202* -2.4609
(1.2693) (1.1355) (1.5013) (1.5543)
Fraser index (workplace) 0.0496 -0.6543**
(0.2605) (0.2651)
Fraser index (education) -0.4641 -0.8280**
(0.2895) (0.3152)
Fraser index ×mult. authors 0.0204 0.0374 0.5044* 0.4232
(0.2758) (0.2194) (0.3020) (0.2971)
Multiple authors 0.0675 0.2043 1.1202* 1.0672* -0.1448 -0.2681 -3.9124 -3.2801
(0.5520) (0.4809) (0.6223) (0.6277) (2.1529) (1.7029) (2.3947) (2.3584)
Multiplier type controls × × × × × × × ×
Other controls × × × × × × × ×
Model controls × × × × × × × ×
Country coverage × × × × × × × ×
Country fixed effects × × × ×
Observations 2,250 2,250 2,044 2,044 2,250 2,250 2,044 2,044
R-squared 0.2694 0.3359 0.2810 0.3111 0.2678 0.3411 0.2777 0.2998
Notes: *** p<0.01, ** p<0.05, * p<0.1. Standard errors are clustered by study.
... A large review of over a hundred studies on the fiscal multiplier shows that there is a large observed variance in the estimates. However, in most studies the public investment multiplier varies from 1 to 2.5, while the multipliers of tax reliefs and transfers to the private sector vary between 0 and 0.5 (Asatryan et al. 2020). Fiscal multipliers do not apply when the economy is stopped due to a period of lockdown. ...
The novel nature of the economic and social crisis, due to the spread of COVID-19, requires new rules and a drastic change in the economic measures to be adopted. The pandemic has caused a spiral of supply–demand shocks that brings about several market failures that make necessary a public intervention to assure the return to health security in the first place and then to restore economic growth and reduce unemployment. Fiscal policy has to intervene both to cover health expenses and sustain families’ income and firms’ fixed costs, and to create the basis for a future recovery through investment. To guarantee the stability of the resulting higher public debt, expansionary monetary policies have been implemented even in a non-conventional way. In the euro area, among other measures, a pandemic emergency purchase programme (PEPP) has been adopted, through which the European Central Bank (ECB) has been buying temporarily existing public debt on the secondary market. We ask whether the ECB should go even further monetising permanently on the primary markets the public debt of euro area countries or relieving a fraction of the debt that they are currently holding. An alternative possibility would be the issuance of perpetual bonds to be bought by the central bank. Needless to say, such measures are rather controversial, not only considering their political feasibility, but also for the feared effects on the central bank’s anti-inflationary credibility and on moral hazard. We argue that a sufficiently strong political will might help overcoming the objection regarding feasibility. Central bank’s anti-inflationary credibility may not necessarily be a concern: according to the credibility theory, in the case of unexpected shocks, no credibility can be gained by following policies that are not credible themselves. On the other hand, moral hazard cannot be an issue when a given measure is fully justified by the occurrence of extraordinary circumstances. Finally, we show that when the economy is hit by a stochastic shock, a moderate inflation might well be optimal, as it would allow a reduction of the unemployment rate.
This paper investigates recent developments in meta-analysis, the tool to quantitatively synthesize research in a certain body of literature. After providing a brief overview on how to do a meta-analysis and discussing recent methodological advancements, I review applied contributions to the field of macroeconomics. It turns out that meta-analyses have often questioned the conventional wisdom and established new consensuses in fiscal, monetary, and labor market policies by uncovering substantial publication bias and unexpected determining factors in many bodies of literature—in particular those dominated by policy conclusions in the neoclassical tradition like minimum wages, central bank strategies, financial regulation and the relative effects of tax and spending policies.
We scale theoretical/ideological positions of economic research institutes over debates. Using only parts of German research institutes’ business cycle reports that deal with economic policy commentary and advice as an example, we extract sections from these reports dealing with monetary and fiscal policy issues from 1999 to 2020. To these corpora we apply methods of unsupervised text scaling (Slapin and Proksch, 2008; Lauderdale and Herzog, 2016), namely Wordfish and Wordshoal. Roughly, results are in line with the common sense in the public policy discourse. For monetary policy texts, we observe a strong, but short-lived consensus in debate-specific positions at the height of the financial crisis in 2008 and a larger polarization thereafter compared to the sample period before. For the fiscal policy textual corpus, the polarization was similarly high before and after the crisis and decreases somewhat during the COVID-19 pandemic. For both policy areas, the German Institute of Economic Research (DIW), Berlin, and the Institute for World Economics (IfW), Kiel, tend to be the most diverse institutes within the spectrum of latent ideological positions. We argue that text-mining techniques might be useful to scale underlying ideological positions in policy-related publications.
We compare the findings of central bank researchers and academic economists regarding the macroeconomic effects of quantitative easing (QE). We find that central bank papers find QE to be more effective than academic papers do. Central bank papers report larger effects of QE on output and inflation. They also report QE effects on output that are more significant, both statistically and economically, and they use more positive language in the abstract. Central bank researchers who report larger QE effects on output experience more favorable career outcomes. A survey of central banks reveals substantial involvement of bank management in research production.
Full-text available
This study explores the positions of economic experts from Central and Eastern European (CEE) Member States in the euro reform debate. Given the dominant voices from French and German politicians and academics in the European discourse, there is an obvious neglect for the positions of CEE countries. Our study tries to fill this gap with a large survey among economic expert communities in all CEE countries conducted in spring 2019. We compare euro reform preferences to benchmarks of surveyed experts in France, Germany, and Italy. We discuss implications for the ongoing euro area reform with a particular focus on several non-euro members’ growing reluctance to introduce the common currency. We argue that only a balanced reform package that combines solidarity with debt self-responsibility could foster the euro’s appeal in the CEE region.
Full-text available
I use Monte Carlo simulations, the jackknife and multiple forms of the bootstrap to study a comprehensive sample of 1309 instrumental variables regressions in 30 papers published in the journals of the American Economic Association. Monte Carlo simulations based upon published regressions show that non-iid error processes in highly leveraged regressions, both prominent features of published work, adversely affect the size and power of IV tests, while increasing the bias and mean squared error of IV relative to OLS. Weak instrument pre-tests based upon F-statistics are found to be largely uninformative of both size and bias. In published papers IV has little power as, despite producing substantively different estimates, it rarely rejects the OLS point estimate or the null that OLS is unbiased, while the statistical significance of excluded instruments is exaggerated.
Full-text available
Background Concern is widespread about potential sponsorship influence on research, especially in pharmacoeconomic studies. Quantitative analysis of possible bias in such studies is limited. Aims To determine whether there is an association between sponsorship and quantitative outcomes in pharmacoeconomic studies of antidepressants. Method Using all identifiable articles with original comparative quantitative cost or cost-effectiveness outcomes for antidepressants, we performed contingency table analyses of study sponsorship and design v . study outcome. Results Studies sponsored by selective serotonin reuptake inhibitor (SSR1) manufacturers favoured SSRIs over tricyclic antidepressants more than non-industry-sponsored studies. Studies sponsored by manufacturers of newer antidepressants favoured these drugs more than did non-industry-sponsored studies. Among industry-sponsored studies, modelling studies favoured the sponsor's drug more than did administrative studies. Industry-sponsored modelling studies were more favourable to industry than were non-industry-sponsored ones. Conclusions Pharmacoeconomic studies of antidepressants reveal clear associations of study sponsorship with quantitative outcome.
Full-text available
It is well-known in statistics (e.g., Gelman & Carlin, 2014) that treating a result as publishable just because the p-value is less than 0.05 leads to overoptimistic expectations of replicability. These effects get published, leading to an overconfident belief in replicability. We demonstrate the adverse consequences of this statistical significance filter by conducting seven direct replication attempts (268 participants in total) of a recent paper (Levy & Keller, 2013). We show that the published claims are so noisy that even non-significant results are fully compatible with them. We also demonstrate the contrast between such small-sample studies and a larger-sample study; the latter generally yields a less noisy estimate but also a smaller effect magnitude, which looks less compelling but is more realistic. We reiterate several suggestions from the methodology literature for improving current practices.
Full-text available
We investigate how often replication studies are published in empirical economics and what types of journal articles are replicated. We find that between 1974 and 2014 0.1% of publications in the top 50 economics journals were replication studies. We consider the results of published formal replication studies (whether they are negating or reinforcing) and their extent: Narrow replication studies are typically devoted to mere replication of prior work, while scientific replication studies provide a broader analysis. We find evidence that higher-impact articles and articles by authors from leading institutions are more likely to be replicated, whereas the replication probability is lower for articles that appeared in top 5 economics journals. Our analysis also suggests that mandatory data disclosure policies may have a positive effect on the incidence of replication.
This paper takes stock of what we have learned from the “Renaissance” in fiscal research in the ten years since the financial crisis. I first discuss the new innovations in methodology and various strengths and weaknesses of the main approaches to estimating fiscal multipliers. Reviewing the estimates, I come to the surprising conclusion that the bulk of the estimates for average spending and tax change multipliers lie in a fairly narrow range, 0.6 to 1 for spending multipliers and -2 to -3 for tax change multipliers. However, I identify economic circumstances in which multipliers lie outside those ranges. Finally, I review the debate on whether multipliers were higher for the 2009 Obama stimulus spending in the United States or for fiscal consolidations in Europe.
We use confidential data on value added tax payments at the sector level, in two large Italian cities, to estimate the effect of audit publicity on tax compliance of local sellers. By employing a difference‐in‐differences identification strategy, we find that such publicity has a positive effect on fiscal declarations made shortly after. The results suggest that increasing awareness on future audits via the media can be an important instrument in the hands of tax authorities. This article is protected by copyright. All rights reserved.