Thread by @SGechert

Some personal good news in hard times: Joined with wonderful coauthors @Chris_ptz @pvillenueve we just published our paper „The macroeconomic effects of social security contributions and benefits” in the Journal of Monetary Economics doi.org/10.1016/j.jmon… #econtwitter Thread:

The gist of the paper is to estimate #macroeconomic #multiplier effects of spending on #socialsecurity vs. cutting contributions. The literature shows it is hard to estimate such effects, because spending and revenues are highly endogenous to the business cycle./2

We did a lot of nitpicking work conducting a time series of timing+size+circumstances of major legislations of social security in GER 1970q1-2018q4 (inspired by seminal work of Cristina+David Romer 2010 AER). This shall identify exogenous changes for causal analysis./3

To do so we filed through tons (or actually gigabytes) of legislative texts, protocols (and also some press coverage). These are well documented at pdok.bundestag.de @BundestagFragen. The following figure summarizes the budgetary impact of these laws over time./4

We hope that our dataset is useful for other researchers as well. The publication includes #replication files and an extensive documentation of the dataset law by law at (raw GSOEP data have to be requested separately) doi.org/10.1016/j.jmon… /5

A site note: We learned a lot about the history+conventions of social security in Germany. You get much more insights into the #institutional underpinnings than just doing some time series number crunching. Moreover, @pvillenueve internalized lots of German officialese :)/6

Ultimately, we regress macro variables (GDP, consumption, investment, production, employment, wages etc.) on these shock series, testing the relative effectiveness of contributions vs. benefits using dynamic IV regression tools (inspired by proxy SVAR of Mertens+ @MortenORavn)./7

Here are the central findings:The GDP response to a cut in contributions yields a fiscal multiplier of about 0.4 on impact that fades relatively quickly. For benefit increases, the impact multiplier is 1.1 and much more persistent. Here are IRFs to 1% of GDP expansionary shock /8

Why are these effects so different? The response of other macro variables suggests benefits work through a strong demand-side channel (consumption), while contributions exhibit supply-side effects. In particular, consumption expenditures react very differently as shown above/9

Why is this so? We go into details of the consumption channel by attributing our social security shocks to micro panel data from #GSOEP according to employment status. Beneficiaries (unemp, pensioners, marginal empl) show much stronger consumption response than contributors./10

This can be rationalized e.g. by credit/liquidity constraints of these HH. We show that beneficiaries have higher probability of credit/liquidity constraints than contributors do. They cannot smooth out income losses and will cut back consumption after benefit cut./11

Our JME paper has implications for #Corona #covid19: current extension of social safety nets will be important to cushion macroeconomic outfalls. + If you want to kickstart aggregate demand after the crisis, transfers to poor HH will be very efficient. doi.org/10.1016/j.jmon… /12

Our paper also speaks to the new #HANK literature. These models often find stronger aggregate effects of fiscal or monetary policy via redistribution channels (towards HH with high marginal propensities to consume #MPC) @GregWKaplan @glviolante doi.org/10.3982/ECTA10… /13

Here is a fresh and interesting #HANK model that has a very similar story to our empirical analysis with very similar multiplier effects by @christianbaye13 @bornecon @RalphLuet

https://twitter.com/RalphLuet/status/1249665104808943616

/14

on #pluralism: #PostKeynesian models stressed redistributive channels long before they became fashionable in #HANK. PK models have own weaknesses, but featured #heterogenous agents in their core.This idea coulda been popularized earlier if people talked+listened to each other/15

A note on the publishing process: “The long road is finally over” as @APeichl put it once

https://twitter.com/APeichl/status/911159277053759488

. We started out with this project in 2013, went through multiple revisions, refocusing, updates of the dataset. First submission was in 2016 /16

Looking back: incredible how much time we spend on it (I’d be scared of budgeting this, but luckily we did not count hours). Nevertheless, given that this is my first high rank publication: Yes, I would do it again! @PHuenermund

https://twitter.com/PHuenermund/status/1248676719885783040

/17

We tried multiple outlets and failed (often with 1st round rejecting referee reports).They were mostly nice, but skeptical re focus of the paper. In hindsight, I see their points, but I was often desperate why they didn’t give us a second round. Finally, JME was nicer to us./18

The process at JME was amazingly fast,fair+constructive (+tough). We had reports <1 month in each revision step + they really improved the paper.E.g. the idea to combine shock series with micro data came from a referee. The editor Yuriy Gorodnichenko also reviewed thoroughly./19

CC: @AchimTruger @SDullien @Sigginho @BachmannRudi @christianbaye13 @jsuedekum @CR_Proano @FlorinBilbiie @Matze_Weber @bornecon @MH_Wolters @KatjaRietzler @MiriamRehm @SvenjaFl @ClaudiusGrabner @PhilippaSigl @boeckler_de @IMKFlash @FMM_macro /end

/end+1: lest we forget to CC @heimbergecon who discovered the publication early on

https://twitter.com/heimbergecon/status/1252820382957989888?s=20