# Lars AhnlandStockholm University | SU · Department of Economic History

Lars Ahnland

Doctor of Philosophy

## About

4

Publications

133,584

Reads

**How we measure 'reads'**

A 'read' is counted each time someone views a publication summary (such as the title, abstract, and list of authors), clicks on a figure, or views or downloads the full-text. Learn more

5

Citations

Citations since 2016

## Publications

Publications (4)

Financialisation has become a new buzz word in social sciences, but, although some of the earliest usages of the concept can be found with economic historians, the recent fad has largely been ignored by economic history. This is true also for the Nordic region. This survey article highlights a handful of studies on financialisation in the Nordic co...

This investigation explores the long-run relationship between the wage share in the non-construction private sector and government efforts to create jobs in public services and construction of infrastructure and houses, in Sweden in 1900 to 2016. In the present article, it is argued that the creation of employment with generous wages by the Swedish...

This study demonstrates a long-run relationship between inequality and the bank debt to GDP ratio in Sweden in 1919-2012. The findings suggest that much of the impact of the top income share on the debt ratio comes from changes in the profit share. Earlier research claims that the rich, via the banks, have lent their savings to the poor as a substi...

This study presents new time series data for private debt in Sweden in 1900–2013, including credit from banks, mortgage institutes and credit companies. The reconstruction of the data is a scientific task by itself, and is complicated by changed definitions, breaks in the series, and the need for occasional interpolation and cross-reference of sour...

## Questions

Questions (68)

Hi! How do I fix both serial and cross-section auto correlation in panel date at the same time? I run a fixed effects model with PSCE errors (takes care of cross-sectional autocorrelation), and one lag of the dependent variable (not significant).

But my Durbin-Watson test is about 1.7, which is too low accordning to the Bhargava, Franzini, and Narendranathan test statistics tables (which are very tight at about 1.9 for the lower bound). Likewise the Wooldridge test for autocorrelation in panel data is significant.

I try to apply a dynamic panel data with two-step estimation and an AR model, but regardless of the number of lags I get significant results for a Pesaran CD test for cross-sectional dependence.

Bottom line: It seems like I can´t address serial and cross-section auto correlation at the same time. Or can I? How?

I only have 28 observations for a regression, while the minimum is often said to be 30. Is that ok anyway, and are there any extra security measures I can take, such as more stern application of significance (1% level)? I use an ARDL model.

Like the title says - where do I put exogenous variables in a panel ARDL model - in the long-run or short-run equation? I want to control for world war one, which means either impulse dummies for each year, or a step dummy for the whole 1914-1918 period.

Best,

Lars

Hi!

I have used VEC models with two variables and have estimated impulse-response functions (IRFs)from them. I am testing two theories, and each variable is the dependent variable for a corresponding theory. Also, I test both theories with different time periods. Naturally, in setups, the error correction term (ECT)/speed of adjustment coefficient is significant and negative, as it should be in a correctly specified model, and in the opposing model (when the other variable is normalized to one in the long-run equation) it is significant and positive (which signifies an ill-specified model, with explosive behavior). Now, I want to present both models for each period, and sometimes the ECT is significant and negative with one of the variables, and in another it is so with the other variable. What IRFs should I present for display in the paper? Should I only use IRFs from the models where the ECTs are significant and negative or can I also present the IRFs that have significant and positive ECTs?

Thanks!

Can I calculate impulse-response function in Excel in some way, by extracting the coefficients I get from a model estimation done with other software? I wish to make an impulse-response function from both a regular ARDL EC model, and from a panel ARDL EC model. Is that doable and how if so?

What are the equivalent cases of cointegration trend spec. (restricted constant, constant, and restricted trend) in an panel ARDL PMG ECM setting?

Hi! I want to mix stationary and non stationary panel data in an error correction and for cointegration. In a non-panel data setting, this is done with ARDL ECM, where a significant EC and long-run coefficient indicate cointegration, and where a bounds test can confirm this. Now, in a panel setting, is the corresponding ARDL ECM with PMG (Pesaran et al. 1998) valid? To test for cointegarion, I can use the Westerlund test in the depvar is non-stationary but indepvars are statioary, right?

Why is there no constant in the long-run in panel ARDL (MPG) regression? It is the same in Eviews and Stata.

With ARDL cointegration (the bounds tests approach), why is the dependent variable stated in first differences in software such as Eviews? Please explain as simply as possible - my algebra is terrible.

Hi! I have a number of variables which are non-stationary in the long-run (1900-2016), but stationary during some sub-sets of that period (1900-1930, 1946-1973, 1980-2016). Can I use cointegration tests with those variables (target dependent variable is always non-stationary) on the sub-sets (shorter periods)?

Hi! I an estimating a relationship between two variables among 13 countries. In some countries, the variables are stationary, and in some, they are non-stationary and cointegrated. For single-equation cointegration, it is common to estimate long-run coefficients with either DOLS or FMOLS. For the sake of comparability, can I use DOLS/FMOLS also for the countries where the variables are stationary - that is in a non-cointegration framework?

How do I interpret impulse-response functions (IRFs) in relation to beta and alpha coefficients obtained from a Johansen cointegration test? For instance, my target (normalized) variable Y has a speed of adjustment (aplha) coefficient of -0.2, meaning that it will return to the long-run relationship after ca five years (data is annual). I have two independent variables (and no constant): X and Z. Both have significant long-run (beta) coefficients. Now, the IRFs show that a change in X leads to a significant effect of 0.006 Std deviations in Y. With a Std deviation of ca 0.1, this equals an effect of 0.06 = 6 per cent. Z does not have a significant effect according to IRF analysis. This seems really small when compared to the alpha coefficient of Y, and looks like a conflict to me. Is this a correct interpretation?

Hi! If an unstable cumulative sum of recursive residuals squared (CUSUMSQ) test indicate instability in the residuals of, say an ARDL model, can this be remedied by the use of heteroscedasticity robust (HAC) standard errors?

Hi! I have a question regarding VECM and IRF interpretation in Gretl. According to the speed of adjustment coefficient (error correction term) in the VEC, the target variable adjusts at a speed of 15,5% per year to equilibrium. But the IRF says that the cumulative change in the target variable is only 2% from year four trough the time horizon of the IRF, which is nine years. How can I interpret this?

Look for ECT short run equation 1, and compare it to the IRF in the files below please.

Best regards,

Lars

## Projects

Project (1)

To determine the interactions between banking debt, inequality and asset prices.