# Results of the analysis

**1. «rpt-fazit» sheet : explanation of rows.**

When the analysis is completed, the «rpt_fazit» sheet will appear, containing the key results. This page will tell you:

- The extent of the gender wage-gap (row 3)
- Whether or not wage equality in the stricter sense is respected (row 4)
- Whether wage equality with regard to the procurement tolerance threshold of 5% is respected (row 5)
- R2 and the interpretation of R2: The closer R2 is to 1, the more wage variability in the company can be explained by the factors taken into account in the model (rows 6 and 7, see also What is R2?).
- The number of valid observations is taken into account for the analysis (row 12). To obtain statistically significant results, 50 valid observations are generally required. The greater the number of employees, the more reliable the results will be. An equal proportion of men and women also increases reliability. (row 13 indicates the number of women).

**2. The «rpt-fazit» sheet: explanation of columns**

Two regression analyses are performed

- Regresion based on characteristics related to personal qualifications only (Regression pq) (first result column), which only considers personal characteristics: training, professional experience and years of service
- Standard regression, which also considers job related factors: The level of qualifications required for the job and professional position.

If the discriminatory wage differential shown in regression pq is significantly higher than the difference in the standard regression, this means that with equal qualifications, women are under-represented in higher positions and very demanding jobs. This additional information can help you to identify potential occupational discrimination, which is also prohibited under the Gender Equality Act.

**3. «data_form» sheet: explanation of column Y**

Column Y in the «data_form» sheet indicates the percentage deviations of actual wages compared with the wages calculated by the Logib model. A positive value means that the actual wage is higher than the calculated wage, while a negative value means that the actual wage is lower than the calculated wage. Values which deviate by more than 20% are marked in red. Should you find any errors in the data entered, you can correct them and restart the analysis.

# Results in the form of tables and figures

- The upper table on the «rpt_regr2_1» sheet shows an overview of the distribution of women and men according to the characteristics used in the regression analysis. Row 10 displays the total number of men and women on whom the analysis is based. Row 26 shows the average standardised full-time gross wage and the percentage wage differential between women and men.
- In the figure below, the wages of each employee are represented by dots and the straight line represents the wages predicted by the model. In an ideal scenario, all of the dots would be on the straight line, i.e. the predicted wage would always correspond exactly to the actual wage. Small deviations are normal; greater deviations suggest either data errors, outliers or discrimination.
- The following sheets («rpt_regr2_2» to «rpt_regr2_7» ) show descriptive results and figures for personal characteristics : age, training, professional experience, years of service, level of qualifications required for the job and professional position.

# FAQ - Interpreting the analysis

#### I have obtained a clear result, but I believe it to be incorrect. What should I do?

- Check whether the data are correct. Column Y in the data sheet «data_form» shows you the percentage deviations of actual wages compared with calculated wages. Marked deviations may be an indication of errors. (Result of the analysis, point 3).
- Information on employee qualifications often has to be entered manually from personnel files and is therefore not always up to date. Update this information.
- Logib automatically calculates years of training and professional experience. This calculation is based on assumptions regarding the length of the specified training. If you think that there is a considerable difference between the professional experience calculated and actual professional experience, you can ascertain and enter the actual number of years of professional experience of your employees. However, you will no longer be able to work with Logib and will have to use a statistical software package to process the data. The data entered can, however, be exported from Logib (see the question below Can I export data?).
- You may have made a mistake when defining the level of qualifications required for the job, the professional situation, or both, on the basis of the wage categories. If this is the case, correct the columns «Required level of skills» and «Professional position» by referring to the job description for the required level of skills and to the organization chart for professional position. By using the four levels for the required level of skills and the five grades for the professional position, the explanatory value of these characteristics will be increased.
- The result can be incorrect if wage components, for example the 13th monthly wage figure has not been entered correctly (1/12 of the 13th monthly wage amount), if there are several different categories of working time, or if hourly and monthly wages are not converted correctly (see Can I include employees paid monthly and those paid hourly in the same analysis?).
- Other factors influencing wages may not have been considered in the models (e.g. location-dependent allowances). If such a factor is not potentially discriminatory, you have two options:
- Remove this information (in the case of allowances) from the wage details
- Or include an additional explanatory variable in the regression calculation. In order to this, you will have to use a statistical software package.
- If you are some doubts, do not hesitate to ask a specialist.

#### What additional conclusions do the regressions calculated by Logib provide?

If the discriminatory wage differential shown in the regresion pq is significantly higher than the difference in the standard regression, this means that with equal qualifications, women are under-represented in higher positions and very demanding jobs. This additional information can help you to identify occupational discrimination, which is also prohibited under the Gender Equality Act.

#### Why does the number of valid observations differ between the regression pq and the standard regression?

If the number of valid observations in the regression pq differs from the one of the standard regression, it is likely that information on the professional position or the level of qualifications required for the job has not been correctly entered (or not entered at all) for some employees. You can check and correct this information on the «data_form» sheet.

#### Why am I getting almost the same analysis on two different sheets?

If the number of valid observations in the regression pq differs from the one of the standard regression, an analysis will be provided once for the base regression and – on a separate sheet – once for the extended regression.

#### What is the purpose of the tolerance threshold?

The 5% tolerance threshold was introduced to use the tools for controlling the public sector within the Confederation. It is possible that the standardised control conducted by Logib does not consider all factors influencing a company’s wages. The tolerance threshold enables this fact to be taken into consideration and to avoid companies being wrongly sanctioned.

**Important**: The Swiss Federal Constitution and the Gender Equality Act prescribe that women and men shall receive equal pay for work of equal value. Here, there is no tolerance threshold. If you suspect systematic or individual discrimination in your company, please contact a specialist in order to conduct an in-depth analysis and to introduce the appropriate corrections to wages.

#### If a wage gap of more than 5% is calculated, why does Logib confirm that the tolerance threshold of 5% is nevertheless respected?

The wage gap calculated by Logib is situated within a statistical range (area of imprecision) as it can be affected by random factors – just like any statistical calculation. The measure for this distribution is the standard error (conclusion sheet, row 10).

Based on this standard error, Logib calculates whether the wage gap lies above the 5% tolerance threshold with a probability of 95%. If this is the case, the tolerance threshold is deemed to have been significantly exceeded. If this probability is not detected, the tolerance threshold is deemed to have been respected – although the absolute value is higher than 5%.

For statistical reasons, the lower the number of employees and the smaller the R2, the greater the standard error is(i.e. the quality of the model, see next question What is R2?).

#### What is R2?

R2 is the coefficient of determination (= R²). You will find the R2 value on row 6 of the «rpt_fazit» conclusion sheet.

R2 shows how well the model is able to reproduce the wage structure in a company. R2 can only accept values between 0 and 1. The higher the R2 value, the better wage variability in the company is explained by the factors considered in the model.

#### Can I export the data?

Use the «file/save as» option to save data in the data format of your choice supported by Excel. You can also select data and then paste them in a new Excel file.

**TIP 1**: Do not forget to save the Logib file before saving your data with «save as» in a different format. If you do not do this you may lose all your work in Logib so far.

**TIP 2**: If you choose to use the second method and your data selections include calculated columns (e.g. the standardised total earnings), you must use the «Edit – Paste Special – Values» menu option. If you do not do this you will not obtain the correct values for columns containing formulas.

### Contact

Patric Aeberhard

Tel. 058 462 68 42

Mail Patric Aeberhard

Logib Helpline

Tel. 0800 55 99 00

Mail Logib Support

### Link

### Documents

Printable version Logib step-by-step

Last modification: 28.07.2015

Size: 856 KB | Type: PDF