Discussion Paper

No. 2012-55 | October 25, 2012
Polarization, Growth and Social Policy in the Case of Israel, 1997–2008


In this paper we apply two statistical models to the measurement of polarization to Israeli income data over the past decade in order to empirically detect income classes as sub-populations of incomes concentrated around an optimal number of poles. The statistical models compared are a multi-resolution analysis (MRA) and a log-normal approach (LNA). We find the MRA to be superior to the LNA, by providing a more efficient allocation of households into each of the classes, reducing the overlap between the classes around the cut-values for each class. We then study polarization by use of the MRA in a multinomial logit-analysis by including ethnic-cultural, individual, family and other characteristics. We use a multiplicative normalized polarization measure developed by Palacios and Garcia (2010) which consists of presenting the interaction of three components, consistent with the axioms spelled out by Esteban and Ray (1994): alienation and identification, the number of income classes and the size distribution of the groups. The strong cultural heterogeneity of Israeli society, the sharp shifts in social policy during the observation period and the generally high quality of yearly Israeli income data render this dataset particularly useful for analyzing polarization. We find polarization to be significantly affected by cultural classes, by social policy and by standard demographic and individual characteristics. A comparison of our results with those of Esteban and Ray and Zhang and Kanbur reveals some similarity with our normalized version of Zangh and Kanbur (2001).

JEL Classification:

O15, I21, I3, J1, H54, O53



Cite As

Rosa María García-Fernández, Daniel Gottlieb, and Federico Palacios-González (2012). Polarization, Growth and Social Policy in the Case of Israel, 1997–2008. Economics Discussion Papers, No 2012-55, Kiel Institute for the World Economy. http://www.economics-ejournal.org/economics/discussionpapers/2012-55

Comments and Questions

Anonymous - Referee Report 1
December 04, 2012 - 10:13

The manuscript provides an analysis of polarization for Israel in the last decade. First it is analyzed the composition and definition of population groups, then an analysis of the evolution of polarization, along with its three main components (identification/alienation, number of groups, distribution of groups), is carried out.


I found the paper of relevant interest for publication, both for the case study - indeed, Israel is a really interesting country to analyze-, and for the technique used - a recently developed polarization measure that seems robust to various kind of comparisons.

That said, I found some choices a bit questionable so I would recommend either to reconsider some of them or to motivate the choices more thoroughly. These points are detailed below.

The author exclude year 1997 from the analysis. This group analysis for this year generated 4 groups in the population and the authors attribute it to a possible lower quality of the data. However, if I understand correctly, the polarization index they use can account for different number of groups, so, in principle, this should not be a problem. I suggest either to include 1997 in the analysis or to give a stronger motivation for its exclusion.

Jerusalem-Arabs were excluded from the analysis because they were not surveyed in years 2000 and 2001, compromising somewhat the comparability of the results. However, from footnote 24, it appears that including them reduces the number of groups to 2 for 2008, implying a stronger evidence of the vanishing of the middle class. To me this is a strong and interesting result, especially if coupled with 1997 result of having 4 classes. Maybe it would be sufficient evidence that differences in polarization found for years 2000 and 2001 (on which the analysis is never focused) may be due to this sampling error.

In this analysis the authors chose to analyze membership only for three specific years: 1998, 2004, 2008. The choice is motivated by the introduction of some "harsh economic policies" just before 2004 and a period of stable growth from 2005 to 2008. Why not repeating the analysis also for the other years and check whether in those chosen year there was some substantial change? This, if not a real contribution to the analysis, would provide a robustness check (maybe by a further table in the Appendix).

Similarly to the previous comment, the authors choose to make the analysis only for 1 year:2005. Again, they motivate the choice, but it would be interesting check what happens for the other years under investigation. If they confirm what found for 2005, it would just be a robustness check that may juts be worth a footnote mention, otherwise there could be something interesting to comment.

1) The abstract would benefit from a slight reduction. It is a bit overlong and sometime gives non-necessary information
2) The legend in the figures are sometime not readable.
3) Figure 3 is a bit confusing. It could be a good idea to remove the black and the dashed lines, and (maybe) provide a linear fit for the tendency (and maybe testing for structural breaks)
4) Table 4 should be completed with standard errors and significance level of the parameters, together with the indication of the reference group.
5) Table 5 is not well explained. What are the X?

Anonymous - Referee Report 2
December 04, 2012 - 10:35

The manuscript deals with the evolution of polarization of the Israeli society from 1997 to 2008. To this aim the authors use a multi-resolution analysis to identify income classes and a multiplicative normalized polarization index.

The paper is very interesting and worth publication with some minor ...[more]

... revisions, mainly in the form in which the paper is presented, rather than its technical content.
The abstract, for example, would need almost a complete rewrite. The aim of the abstract is to allow the reader to know what he will find in the paper in a concise text. I find that the present abstract is both vague and at the same time overlong. I will give more suggestions below.
The introduction suffers of some lack of clearness. It is somewhat difficult to explain, but the introduction becomes clear after you read the paper, but at a first read it does not give a clear idea of what the reader will find in the paper. A slight revision would greatly benefit the paper.
For the rest, I am substantially fine with the choice and presentation of the statistical instruments to measure polarization and find that the results are interesting and well presented, except for some graphs that have problems with the legend.


The first sentence needs to be rewritten: it should contain the aim of the paper, without referring to how many, or which statistical models are used.

The details on the techniques used to analyse polarization in Israel should follow. This sentence should not include the sentence "we find MRA to be superior ..."

A very short presentation of the main findings should follow.

Overall, the length should be reduced, avoiding unnecessary details (for example I would drop the sentence "A comparison of our results...")


Page 2: the first sentence of the second paragraph is somewhat ambiguous. It may seem that you pool data from 1997 to 2008. This needs to be clearer.

Page 3: the sentence that starts with "we show that the values of the new measure proposed..." comes out of the blue. There seems to be no link with the previous sentence. You should maybe include this sentence in the subsequent paragraph about polarization measures. (you may need to rearrange it a bit)

Page 4 last paragraph: I would avoid mentioning that the analysis of the allocation of households to each class is carried out for three years only. It may generate confusion on the whole analysis (one may think that all the analysis is carried out in these years only) and would need more explanation.

Empirical results

Page 20: The two figures reported in the text fog the mean net equivalized income could not be found in Table 3

Page 20 footnote 22-23: the fact that that Jerusalem Arabs were excluded is repeated.

Page 20: Table 3 is calculated without using sampling weights... why? Even though you needed to exclude Jerusalem Arabs, at least rest of the population would have been correctly weighted.

Page 22 Figure 2: the legend is unreadable (the second and third entry).

Page 23 Figure 3: the legend is unreadable and the graph overpopulated. The continuous black line and the dashed line could be avoided. Are confidence intervals based on a linear fit of the series? If so it should be specified, also the PG line might not be smoothed (this is personal styling preference, though opinable).

Page 24: The first sentence should be rewritten (in part already present in page. 23). So I would drop the first part of the sentence and leave only the part that refers to the three components of the PG index.

Page 25 line 8: remove the word "appendix"

Page 26 Table 4: the coefficients are reported without s.e., p-value or any other significance indicator (asterisks). A note should indicate whether the baseline group is the poor or the rich.

The authors may want to compare their findings of the impact of "harsh social policies" with what found by Olivier, Piccoli and Spadaro (2010), RIW, Vol 56(1), pp 186-214.


Page 28 "we propose and alternative measure of polarization...": I am not sure whether this sentence is true. The measure used in the manuscript is proposed by Palacio and Garcia (2010), apparently without modifications.

In several points the use of terms "Positive/Normative" to describe the type of analysis are used in a somewhat unusual way. It should be somewhat restated clarifying the meaning.

A quite substantial part of the conclusion is spent commenting the consequences of not having the "super rich" in the sample. This seems not the right place for the discussion. It should go in the sample description or in a footnote.


Tables A3, A4 and A5: P should be replaced by PG.

Anonymous - review
January 09, 2013 - 00:08

review report: Polarization, Growth and Social Policy in the Case of
Israel, 1997–2008

The paper propose an alternative measure of polarization to the main existing ones. Then applying this index on Israeli data which provide an interesting ground for the analysis. The paper provides a better understanding of the ...[more]

... polarization in Israel.
In general the paper is well written and provides a nice contribution to the existing literature.

I have some minor suggestion that might improve the paper.
Minor suggestions
1. The abstract should be rewritten so that it would better reflect the contribution of the paper
2. When presenting data in the tables its would be much more convenient if the number will be presented with two/three digits after the dot

Daniel Gottlieb - Reply to Referee Reports
February 25, 2013 - 13:14

See attached file