Comments and Questions

The Authors study an agent-based New Keynesian DSGE model with different social network structures to investigate the significance of network topologies to macroeconomic stability. I have a list of major comments and also some minor comments that I suggest the Authors to take into account in their revision. Major Comments: The Authors should emphasize a little bit more the importance of the representative-agent assumption in standard DSGE models and the relevance of heterogeneity in agent-based approaches. It is not entirely true that, as the Authors state at p. 6, "the difference between the stylized New Keynesian DSGE model and the agent-based DSGE models is the difference between the expectations of the output gap and inflation." Heterogeneity is another key factor that differentiate the two approaches. Another one that the Authors overlook is out-of-equilibrium dynamics in ABM, whereas in stadard DSGE everything happens in equilibrium. I would suggest the Authors to consider these points in their revision. The assumption of a non-directed network is crucial, but is not fully justified. Why expectation formation and interactions in the real-world should be always reciprocated? I am not totally convinced that the distinctions that the Authors make between small-world (SW) networks and scale-free (SF) networks is entirely sound. A SW network exhibits low average path length and high clustering, together with a bell-shaped degree distribution. A SF network displays a power-law degree distribution, but it can also have low average path length and high clustering, as it happens for example with the preferential attachment model. Therefore, the two classes are actually overlapping. Eq. 17: Why is the interaction strength inversely related to the number of neighbors? Please justify with an economic argument. Some more economic justifications for the chosen network parameters should be given. Results in tables 3-6 are puzzling. Whereas there are differences in variances across different network topologies and parameters, these differences often occur at the 3rd digit. No indication on whether these differences are statistically significant are given. The Authors should provide a more in depth statistical analysis of these figures. Hypothesis 2 should be rephrased. A higher clustering may simply be due to a higher connectivity. Therefore, the net effect of a higher clustering is not "per se" clear. One should analyze what happens when clustering increases conditional to the same level of degree. Therefore results should be compared with null network models that take as given e.g. the level of local connectivity. The discussion at page 19 below table 7 should be generalized by doing an experiment that checks wthere the conclusions are true when one changes the rewiring probability, etc. Why the Authors have chosen the parameters in Table 7? Are regression results robust to changes in these parameters? Regression analyses are performed by drawing at random a network with different structural properties, computing its topological properties and see if they affect output gap and inflation in the DSGE model. I have two concerns with this strategy: (i) the analysis should be also performed for each given network class seprately, to understand if apart from its topological properties, the structural features of the graphs, such as the shape of the degree distributions, affect the results; (2) we already know from the model the economic covariates that influence the two dependent variables, so why not controlling for them in the regressions? In fact the results obtained in table 8 may suffer from omitted-variable biases. The Authors should discuss their results about the effect of connectivity on volatility in terms of the recent literature on interbank networks and their robust-yet-fragile properties, see for example Gai, P. and Kapadia, S. (2010), “Contagion in financial networks”, Proceedings of the Royal Society A: Mathematical, Physical and Engineering Science , Vol. 466, The Royal Society, pp. 2401–2423. Minor comments: There is a lot of work in agent-based macroeconomics on Keynesian macro models (without an explicit network microfoundation) that is not cited in the paper. For example: Dosi, G., Fagiolo, G., Napoletano, M. and Roventini, A. (2012), “Income Distribution, Credit and Fiscal Policies in an Agent-Based Keynesian Model”, Journal of Economic Dynamics and Control. Dosi, G., Fagiolo, G. and Roventini, A. (2010), "Schumpeter Meeting Keynes: A Policy-Friendly Model of Endogenous Growth and Business Cycles", Journal of Economics Dynamics and Control, 34: 1748–1767. I urge the Authors to consider the following works and cite them consistently in the paper. Typos: - page 4: "lagged" instead of "logged"