Housing Markets and Resource Sector Fluctuations: A Cross-Border Comparative Analysis
Round 1
Reviewer 1 Report
Please see attached file for comments.
Author Response
We thank the reviewers for their comments and have uploaded our replies to these.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
1.- In several figures (for ex. 1, 2, 3, there is not information about the unit of measurement used on the Y-axis; it is necessary to add them.
2.- In point 4.1 it is stated that there are similarities between the two Swedish towns, andless with Australian one; could some support data be added? What is the production in each region? The reader should have an idea about the size of the mining markets in each region.
3.- In table 2 the ADF test is applied, but it is not clear to which time series, and their number of data. Also, what is the statistics value for each test and its p-value? A graph ot these series could be more informative, about the possible stationarity of the time series involved, that should be clearly specified for each performed test.
4.- In page 10, lhere is a possible model for house prices (HP) or its natural logarith (LHP; it is not stated that this is ln(HP), and the reader has to guess it) as a function of demand (D), or or ln(D) (LD, again the reader has to guess that this is it); then, there is another variable TREND that it is not clear that it is just the series t = 1,2,3,... or ln(t). This should all be clarified as for the reader has just to guess what model is applied.
5.- Continuing with last point, there is a term MA so called 'moving average'; but what is this? Is it an MA term in an ARMA structure for the errors? Or a transformation of the original data? As it is, it is not understable.
6.- In tables 4, 5 there is no clue about what is been estimated- It seems that these are several regression models. But what is the endogenous variable: HP in each town; what are the variables exogenous? For example, what are NCO, AVER, SCO? The R2-adjusted should be replaced by R2, as,there is no comparaison between alternative models and R2 is more informative. What are the dummy variables used for? There is no interpretation about them, and this is necessary to be able to understand what has been done.
7.- The models used are (as I understood) to estimate trends in HP and their relation to demand. But where is the explanations about the volatility in prices?
Author Response
We thank the reviewers for their comments and have uploaded our replies to these.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
I have a few additional minor changes to suggest after reading this greatly improved revised version.
1) Table 1 includes "SGU" as a data source, but does not define acronym in text, although it is included as a reference. Either define in text or add citation to table for clarity.
2) Table 2 denotes the sample size as 2000-2018, which I believe instead is the sample period. This should be corrected and the sample size added.
3) Line 453 "are" vs. "is". Line 592, misspelling. Line 753, seems to be a word or two missing, etc. (There seem to be a few instances throughout of minor errors in spelling, grammar, or punctuation, particularly in the text that was added or edited for this version. I'd encourage the authors to comb through these sections carefully.)
4) Line 401 (Figure 1). It appears to me that the Swedish mining town prices have tripled over the sample period, not quadrupled.
Author Response
Attached you'll find our replies to the questions raised by reviewer 1 & 2. Thank you!
Author Response File: Author Response.pdf
Reviewer 2 Report
The models are not specified; por example in page 12, there is LHP = f(LD); are the models estimated linear? It should be stated: log(P) = a + b log (D) + error; the next one, surely used log(t) as an explanatory variable, and that the error term follows and MA model to try to model the autocorrelation. In lines 509-515 there is a try to explain this; the estimated model should be put clearly and omit expression with no sense (like 'Ma= a simple moving average' or 'MA1 = e(t-1).
In any case, the modeling of the autocorrelation has not been achieved, as can be seen directly in figure 3.
Comparing figures 1 and 2, in the first the average housing prices in Sweden is included, showing that in the mining towns the trend is similar than in the mining town, while in figure 2, there is not a global Australian house prices, that surely is quite different from prices in the mining zones.
Table 2 reflects the results of the ADF test for different time series; but, if it has been applied to the series in figure 1 and 2, it is superfluous: these series are clearly non stationary, and there is no use of employing a statistical test for something evident; on the other hand, if these series are integrated of order 1, and the ADF text has been applied to the first differences, it is not stated. And, another detail: if the original time series are of 18-20 data, the ADF test use auxiliary models on differences of the series and lags of these differences, loosing data as these auxiliary models are estimated; for example, if you use four lags in the test, you lose five cases, so the sample size reduces from 18-20 to 13-15 data and with these you use a model with up to six parameters; this could be put on hold, at least; no information about this is explained
Table 3: the results are confusing; the y variables are expressed in log, but in table the variables NHP, CHP, SHP are prices; followed by a = sign. The explanatory variables are also in log scale? Why use the adjusted determination coefficient, and not plainly R2? (these is no comparison with sub-models. What is the economic justification of introducing a dummy variable D9 for 2009? (in a model estimated with 12 cases and just 9 degrees of freedom, as there are no autocorrelations terms included); and it is stated in line 540 that 'D10 a dummy for D10'
Author Response
Attached you'll find our replies to the questions raised by reviewer 1 & 2. Thank you!
Author Response File: Author Response.pdf
