Register based staff turnover statistics : An evaluation of the Employer declaration at the individual level

(1)

Register based staff turnover statistics

An evaluation of the Employer declaration at the individual level

Author:

Viktor Törnqvist (970403)

Fall 2020

Statistics, Independent Project I, Second Cycle 15.0 credits, ST433A Subject: Statistics

Örebro University School of Business Supervisor: Ann-Marie Flygare Examiner: Thomas Laitila

(2)

Abstract

The purpose of the thesis is to estimate register based staff turnover statistics for businesses in the private sector in Sweden with the new register; the Employer declaration at individual level (AGI- Arbetsgivardeklarationen på individnivå) as a base and present the estimates for different domains/ groups of study. The estimates are then compared with estimates from the Short-term employment statistic survey. Methods such as linkage, to connect different registers, and derivation of new necessary variables is used to create the final version of the AGI-register (called the KS-register) with the new variables that is used to make subsets of the data. The staff turnover was estimated with estimators described by Wallgren & Wallgren (2014) and Särndal, Swensson & Wretman (2003). The conclusion that is made is that estimated staff turnover for different sector of operations can be extracted from the KS-register with the AGI-register as the base. However the staff turnover results for the months investigated differ between the KS-register and the Short-term employment statistics. These differences might be explained by the fact that the KS-register contain all sorts of employments, while the staff turnover results in the Short-term employment statistics do not.

(3)

(4)

List of tables

Table 1. Description of letter group and sector of operation (SCB 2020c). 12

Table 2. Number of employees each month, by gender in the KS-register. 15

Table 3. Proportion (in percent) of hires and separations in relation to the total number of employees

(permanent employees) presented as confidence intervals by sector, second quarter of 2020. 17

Table 4. Proportion (in percent) of hires and separations in relation to the total number of employees

(permanent employees) presented as confidence intervals by sector, third quarter of 2020. 18

Table 5. Number of permanent employees, contracted/hourly paid employees and total number of

employees presented as confidence intervals by sector, second quarter of 2020. 19

Table 6. Number of permanent employees, contracted/hourly paid employees and total number of

employees presented as confidence intervals by sector, third quarter of 2020. 20

Table 7. Proportion (in percent) of hires and separations in relation to the total amount of employees (permanent employees, fixed-term employees/contracted employees and hourly paid employees) by

sector, April, May and June 2020. 23

Table 8. Total number of employees by sector of operation, April, May and June 2020. 24

Table 9. Proportion (in percent) of hires and separations (estimates in parenthesis from the second quarter of 2020 from Short-term employment statistics survey) in relation to the total amount of employees (permanent employees, fixed-term employees/contracted employees and hourly paid

employees) by sector, second quarter of 2020. 25

Table 10. Proportion (in percent) of hires and separations in relation to the total amount of employees (permanent employees, fixed-term employees/contracted employees and hourly paid

employees) by sector, July, August and September 2020. 26

Table 11. Total number of employees by sector of operation, July, August and September 2020. 27 Table 12. Proportion (in percent) of hires and separations (estimates in parenthesis from the third quarter of 2020 from Short-term employment statistics survey) in relation to the total amount of employees (permanent employees, fixed-term employees/contracted employees and hourly paid

employees) by sector, third quarter of 2020. 28

Table 13. Proportion (in percent) of hires and separations in relation to the total number of employees by month (permanent employees, fixed-term employees/contracted employees and

hourly paid employees) by month and gender, second quarter of 2020. 29

Table 14. Proportion (in percent) of hires and separations in relation to the total number of employees by month (permanent employees, fixed-term employees/contracted employees and

(7)

List of figures

Figure 1. Number of employees, divided by month of work in the KS-register. 13

Figure 2. Number of workplaces by month in the KS-register. 14

Figure 3. Histogram of differences in number of employees for each workplace between the

(8)

(9)

1 Introduction

The usage of administrative data is on the rise and has become increasingly more essential for academic analysis and in the field of official statistics (Bakker & Daas 2012). Large numbers of countries are looking at the potentiality of using administrative data for statistical purposes (Wallgren & Wallgren 2014). Statistics Sweden is the main statistical producer in Sweden. One of their surveys is the Short-term employment statistics (KS- Kortperiodisk sysselsättningsstatistik) and it has been done every quarter for over thirty years. This survey is done through a web questionnaire (the questionnaire may also come by regular mail) that the sampled workplaces from the sampling frame answers (SCB 2020b). When new data, received from the tax agency, that may contain useful information became available 2019, thoughts about improving the survey appeared. One thought where if it is possible to use the information from the new register to make some parts of the Short-term employment statistics survey based solely on registers instead of the sample survey. This will reduce the burden of the informant at each workplace, reduce the cost and the precision of the statistics may improve. The precision may also improve because when drawing a sample, the problem of not using the entire population gives an uncertainty and there will always be a few that doesn’t answer the questionnaire which lead to non-response errors. However if there is a difference in the definition of a variable, measurement errors will occur.

A broader perspective may also be taken into account. A survey is done and carried out because of the fact that the users are in need of the statistics. When a new survey is executed, there is a dialog between the users and the producers so that the statistics that are desired by the users are produced, the statistics are for the users and the users may use the statistics for important decision making or in research projects. The new register that is given by the tax agency might have other information that make it possible to extract new statistics which may be desired by the users.

In Netherlands, a traditional census could cost several hundred millions euros, but with the register based survey approach, the cost stayed around 1.4 million euros. The register based census was proved to be rewarding by the fact that the cost shrunk, the response problem was almost non-existing and the time of production became a lot shorter (Nordholt 2018). However, this comparison was between a traditional census where data about all objets in the population was collected for a specific period of time and a register-based census where they do not use the traditional way of data collection and just rely on registers that where in possession. This might explain the huge financial difference but it is a good illustration of the impact of using a register based survey instead of a sample survey.

Another strong argument of doing a register based survey is, as described above by Nordholt, the production time. One of the advantages that Statistics Netherlands had compared to other countries that used a traditional census was that they did not have to check and correct any census forms. However, the alternative to do the low cost register based survey census approach, is only an alternative for nations who are in possession of sufficient registers (Nordholt 2018).

The first register-based census in Sweden was done in 2011 and these registers that is now in place shall be of great importance in the production of official statistics in the future. Registers that are up to date and coherent are going to make it possible to improve the quality and keep the cost down of the official statistics (Andersson, Holmberg, Jansson, Lindgren & Werner 2013).

The workload between working with register based surveys and sample surveys may also be to the favor of register based surveys.No sample has to be drawn and questioned, no variance calculations

(10)

study variables that is wanted and received from a questionnaire is not always available in the register that is received from another authority. Then only the variables that exist in the register, are the ones that the statistician can work with. On the other hand, there are different methods available to derive new variables by using links with other registers or by combining different variables that exist in the register (Wallgren & Wallgren 2014).

Register based surveys are becoming more and more popular due to the fact that more countries around the world are beginning to use and create administrative registers of their population as the nordic countries has done since the 1960’s (Wallgren & Wallgren 2014). To be able to create high quality statistics with little burden of the respondents with the aid of high quality registers is a luxury that the nordic countries has been having for quite some time. Although the administrative registers that is available at different authorities should be further explored so that more surveys that is carried out by Statistics Sweden can be based or partially based on registers.

1.2 Purpose

The purpose of this thesis is to estimate register based staff turnover statistics for businesses in the private sector in Sweden with the new register; the Employer declaration at individual level (AGI- Arbetsgivardeklarationen på individnivå) as a base and present the estimates for different domains/ groups of study. These estimates should then be compared with estimates from the Short-term employment statistic survey. The aim is to extract register based staff turnover statistics from the new register and present them in different domains.

1.3 Disposition

The thesis begins with the section background, here the Short-term employment statistics survey and the employer declaration at the individual level are explained and described. This section is followed by the method section where the methods that is applied in the work are presented. The next section is data, here the collection, description, processing and editing of the data is described along with the derived variables. This sections also present results from the Short-term employment statistics. The section methodology describes the estimators that will be used to calculate the totals and proportions. This section is followed by the result section where the result of the study is presented. The discussion of the results of the study along with the possible errors of the study and references are shown at the end.

2 Background

This sections describes the Short-term employment statistics survey, the study variables and domains/groups of study. The AGI-register that is delivered from the tax agency are also describes along with the study variables.

2.1 Short-term employment statistics (KS)

The main purpose of the Short-term employment statistics survey is to (in detail) quickly find changes in the number of employees at the industrial level and regional level as well as report business affiliated variables such as absence from work and staff turnover (SCB 2020a). The survey is done four times a year (every quarter) and the target parameters are the total number of employees, total number of fixed-term/contracted/hourly paid employees, proportion of absence from work, proportion of newly hired and proportion of newly separated employees. The number of employees refers to the total number of individuals who are employed during the time period. The

(11)

proportion of hires and separations refers to the total number of individuals who has begun their employment and finished their employment during the time period, as a proportion of the total number of employed individuals during the same time period. The target population is all workplaces that conduct business in Sweden. The survey uses two different sampling frames, the first being for businesses within trade and industry and consist of the workplaces (arbetställen). The second sampling frame are for legal entities within the public administrations and the household non-profit organizations also called HIO (SCB 2020b).

2.1.1 Study variables

There are five study variables that are investigated in the Short-term employment statistics. Number of employees

Number of fixed-term/contracted/hourly paid employees Absence from work

New hires New separations

The first being total number of employees for the workplace. The number of employees at the time of the data collection includes permanent, part time and fixed-term/contracted/hourly paid employees that where employed one specific day during the time period of which the survey was carried out. An employed person could also include an individual who at the time of measurement were not on duty due to the arrangement of working hours, illness, holidays or leave. Active owners of stock corporations are also included. An employee who has more than one workplace are being counted several times. If, on the other hand, an employee has several positions within the same workplace, he or she is only counted once (SCB 2020b).

The number of fixed-term/contracted/hourly paid employees at the time of measurement refers to persons (full-time and part time employees) that has been hired for a specific period of time or for a specific assignment or is an hourly paid employee who received payment from the workplace that time period. The third variable is absence from work. Absence are reported for three reasons of absence. The first being illness, the second being vacation and the third being other absence (SCB 2020b).

For the businesses within trade and industry the number of new hires and separations are also collected. New hires is described as the total number of new hires who has been hired at the workplace during the month in question. Separations are described as the number of employees that left the workplace due to termination or at own request during the month in question. The statistics consist of calculated values as totals and proportions. This section is called staff turnover statistics (SCB 2020b).

2.1.2 Domains of study

The estimated totals and proportions are also presented in different domains in the Short-term employment statistics. The estimates are reported for every sector of operation. For the businesses within trade and industry, the data is reported by industry, size class and county. The number of employees and proportion of absence of work is divided into gender. The proportion of newly hired and separated is reported only by industry group and size class (SCB 2020b).

y1=

y2=

y₃=

y₄=

(12)

The public administration is divided into state, municipality and region. The number of employees is reported by gender and form of employment for every sub-sector. The proportion of absence from work are reported by gender and sub-sector. For the household non-profit organizations only the total number of employees and total proportion of absence are reported for this sector (SCB 2020b).

2.1.3 Sample Survey

In the sampling frame for the businesses within trade and industry there are two auxiliary variables that are used as stratification variables. The first being industry sector (63 sectors) and the second being size class by number of employees (5 classes). The sampling frame for the public administration and HIO also has two auxiliary variables that are used as stratification variables which is sector of operation (4 sectors) and size class by number of employees (5 classes). The stratification divides the sampling units into 313 respectively and 13 strata from which a sample of objects are selected. From the businesses within trade and industry and HIO, all businesses with more than 100 employees are selected to the sample, in other words, these businesses have inclusion probability one. Within the public administration, a sample is only drawn from strata where there are fewer than 50 employees within the municipal administration meanwhile the businesses with more than 50 employees have inclusion probability one. A sample is also drawn from strata where there are fewer than 100 employees within the state administration meanwhile the businesses within the state administration with more than 100 employees has inclusion probability one. The selection procedure results in a probability sample of the type stratified simple random sampling (SCB 2020b).

The parameters are estimated with a GREG-estimator (general regression estimator) where information about the number of employees in the business register at statistics Sweden is the auxiliary information (SCB 2020b).

2.2 Employer declaration at the individual level (AGI)

The employer declaration are delivered to the tax agency each month from companies in Sweden. The declaration contains information about tax deduction, gross salary, different types of benefits for the employee, along with other information for each workplace. If a business has more than one workplace, they should report this information for each and everyone of the workplaces. The declaration is a way for the tax agency to receive information about the companies in Sweden and their workplaces and also receive knowledge about each employee (as of 2019), and now, each month. As of the year of 2019, the companies should also report the gross salary, benefits, tax deductions and deducted preliminary taxes per payee (every month) instead of reporting the total for every employee of every variable the end of the year as prior to 2019 (among other information) (Skatteverket 2020). The declaration consist of one part that has information about the individuals (individuals are the observation object) and the second part of the register has information about the workplaces (the workplace is the observation object). The first register, about the individual is called as mention, the AGI-register. The other register, which contains information about the workplace itself is called AKU_AGI_13manader (AA13).

(13)

2.2.1 Parameters

The parameters that will be estimated in this thesis are totals and proportions

. The variables that will be described by the parameters are newly hired and separations and the number of employees in each sector. The estimated parameters will be presented for the months of April, May and June in the second quarter of 2020 and July, August and September for the third quarter of 2020. The target population are all the employees that has received payment/ benefits from a workplace one specific month.

2.3 Sources of uncertainty

In a register-based survey, sampling errors is not a problem because of the fact that a sample is not drawn from a population. The issues that the statisticians may face are the non sampling errors. These are error such as coverage errors, measurement errors, processing errors and non-response. For the Short-term employment statistics survey one of the sources of uncertainty that are present, is because of the fact that a sample is drawn. Another uncertainty is non-response, meaning the companies who do not respond to the survey. Measurement errors are also an uncertainty, here, the question or the answers may be misunderstood. The first source of uncertainty (the fact that a sample is drawn) contributes with random uncertainty meanwhile the other two sources of uncertainty contributes to systematic uncertainty (SCB 2020b). However, the errors mentioned above about the fact that a sample is drawn and non-response might not contribute to a lot of uncertainty. Below, the error for register-based statistics are presented.

2.3.1 Coverage errors

Coverage errors can be divided into two different kinds, over coverage and under coverage. Over coverage is when the sampling frame contains objects that do not belong to the target population, for example if a person has emigrated to another country but has not given notice to the tax agency, then the person still belongs to the population register. Under coverage is when the sampling frame do not contain every object that belongs to the target population, as an example if a person has immigrated to the country but has not given notice about their stay in the new country, then they do not belong to the population register but is a part of the population (Särndal, Swensson & Wretman 2003).

2.3.2 Measurement errors

Measurement errors may occur if the individuals who is interviewed/answering the questionnaire gives incorrect responses. This may happen if the individuals misunderstands the questions from the questionnaire, the interviewer ask the question in an impartial way or measurement errors might occur due to the mode of the interview (Särndal, Swensson & Wretman 2003). In sample surveys, the agency might work with improving the questionnaire to somehow prevent major measurement errors. In register-based surveys, it is not always the agencies who are responsible for the questionnaires and instructions, however the agencies should try to influence the authorities that are in possessions of the administrative data to improve their instructions and questionnaires (Wallgren & Wallgren 2014). Other errors might exist because of the fact that the statistical concepts do not entirely meet the administrative, because the administrative ones may have been defined with different purposes. The statistician/researchers might not have any influence of the production of

ty = ∑ k

yk

p = ∑_∑kyk

(14)

the administrative data nor the definitions, this kind of error is called validation errors (Bakker & Daas 2012).

2.3.3 Processing errors

Processing errors happens during the preparation of the collected data (the data processing prepares the data for analysis and estimation). It might be coding errors and/or transcription errors. Errors in any imputed values also belongs to the category processing errors. (Särndal, Swensson & Wretman 2003).

2.3.4 Non-response errors

This is a problem that register based survey are more or less saved from. However the non-response are not entirely gone when having a register as a base for the survey. There might occur some item non-responses if the information giver do not understand a question and leaves a blank space instead of an answer for parts of the questionnaire (Särndal, Swensson & Wretman 2003). However the non-response problem in register based survey do not induce the same issue as for a census or a sample survey (Wallgren & Wallgren 2014).

2.3.5 Linkage errors

Administrative data contained in various registers may be connected by linkage. There can occur two types of errors when performing linkage techniques. These two errors are missed links and miss links. A missed link occur when a matching is not found. This error are alike the non-response problem in surveys. Miss links arise if objects that should not match, are being matched together (Bakker & Daas 2012).

2.4 Differences between KS and AGI

The main difference between the Short-term employment statistics survey and the AGI-register is the sampling method. The Short-term employment statistics uses stratified simple random sampling as sampling design and a sample is drawn from each strata (except the stratums who contains workplaces with inclusion probability one) containing workplaces in Sweden. The AGI-register contains information received from each of the workplaces that are active in Sweden. Here the information is given to the tax agency by the workplaces and a sampling design is not required to collect the data (it is mandatory by law to deliver this information). The Short-term employment statistics micro data consist of workplaces and variables who describes the workplaces such as total number of employees, newly hired employees and how many individuals who has left the business the month that is investigated meaning that the workplace is the sampling unit and observation unit. The target population for the Short-term employment statistics survey is all the workplaces in Sweden and the frame population is the workplaces in Sweden.

The AGI-register contains information about each employee along with variables such as gross salary, which workplace the individual work at, which organization number that workplace is tied to and personal identification number for the individual meaning that the observation unit is the individual and the sampling unit is the individual as well. The target population is all employees in Sweden and the frame population contains all legal employees. The frame population do not contain individuals who does not receive any kind of payment/benefit for their work or work illegally.

(15)

3 Method

In the book Register-based Statistics, Statistical Methods for Administrative Data by Anders Wallgren and Britt Wallgren they talk about register based surveys. They mention that when a statistical office receive data from an authority, processing of the data and selection of variables are done to meet the requirements of the office. They also mention that linking different registers may be required to make use of the administrative data that is received from the authority. Matching keys between different registers may be personal identification number or organization number just to mention two possible keys. These kinds of links are called deterministic linkage and is described as when two records matching keys are exactly the same, the two records are linked together (Wallgren & Wallgren 2014).

These kinds of methods/techniques are the ones that will be used in this thesis. The possible matching keys in the AGI-register is personal identification number for individual, organization number for the business, CfarNr (a unique code for each workplace) and also the workplace number, these last three might work as a link combination as well. The personal identification number may connect where a person work month one compared to month two, this also helps to identify if a person belongs to the labour force one month compared to another. The organization number and CfarNr may also be used to see how many individuals that are employed at the workplace/business during one specific month.

The two available registers will be linked (AGI-register and AA13-register) to receive information about sector of operation for each of the columns in the AGI-register. This makes it possible to divide the new linked register for each month into subset of individuals who belongs to the same sector and compare the sector subset with the corresponding sector subset for other months. This newly created register will be called the KS-register. These comparison may be done with help from the R package ”compareDF”.

3.1 Derived variables for register based statistics

Derived variables is described by Wallgren and Wallgren to be an important part of register surveys. Instead of asking questions and receive information about the variables that are desired (as in sample surveys), it is possible to derive new variables by using the data from the registers. There are four different types of derived variables in register based statistics. The first being variables "that is derived by grouping and dividing into class intervals” (Wallgren & Wallgren 2014, p.62). Age is variable that can be divided into age groups or age intervals or origin such as country of birth might be divided into much broader groups such as continent. The second type is ”variables derived

by arithmetic operations using variables in the same register” (Wallgren & Wallgren 2014, p.62).

These are variables that are made up of one or several other variables in the register, for example the number of adults in the household is the new variable that is wanted. Then the number of individuals in a household that is equal or over the age of 18 is subtracted by the number of individuals below the age of 18 in the household. To be able to describe the third and fourth type, source register and target register should be described. The target register is the register that the statistician uses for statistical analysis meanwhile the source registers is other registers that the statistician may use to retrieve usable information from. The third type is ”variables derived by

adjoining” (Wallgren & Wallgren 2014, p.62). This type of derived variables are a variable that are

created by deriving a new variable by using variables that exist in other registers. A object from the source register may be linked to other objects in the target register with a one-to-one relation or

(16)

one-to-many relation. This third type can be summarized as that every object in the source register may be linked to one or several objects in the target register.

The fourth type is ”variables derived by aggregation” (Wallgren & Wallgren 2014, p.63). This type is also a method of combining two registers. In the source register, one or several objects may be linked to only one object in the target register. Meaning that one new variable in the target register is a combination of two or several variables from the source registers. These kinds of methods, to create new derived variables is the ones that will be used in this study (Wallgren & Wallgren 2014).

3.2 CompareDF package

The ”CompareDF” package compare two data sets with help from one or more variables and it is also possible to code which variables that should be ignored in the comparison. Specify which months that should be compared (month one and month two). Code which variables that should be ignored and which variable(s) that should be compared between the months. Then the reference variable needs to be chosen, often a name or a PIN code and the output will give the information ”additions”, ”changes” and ”removals”. The additions is a number that shows which objects that was not in month one compared to month two in relation to the PIN code/name that is compared. Changes gives the number of objects in month one that has changed the value of the variables that should be compared from month one to month two and removals is the number of objects that exists in month one but not in month two (CompareDF 2020).

An example may illustrate how CompareDF works. The months that should be compared are July of 2020 and June of 2020. The staff turnover for Sector A is wanted. Divide the data in such a way that the first data only contains objects that worked in sector A in June 2020, and the same is done for July 2020. Then put June and July as the data that should be compared in the compareDF package and specify which variables that should be compared. We want to see if an individual has changed their workplace between the month of June and July or/and if a person has begun their employment at a workplace in July. The identification variables to be compared are organization number and CfarNr together and the variable that should be the reference variable is personal identification number. The rest of the variables are ignored. Then the package will look at every personal identification number and see if they belongs to both registers. If a new personal identification number is detected in July, then it is called an addition. This means that new individuals has begun their employment in sector A in July. If an individual exists in both data sets but has changed either their CfarNr or organization number, a change is detected. This means that the person had begun its employment at a new workplace in July. If adding these two (additions and changes) the total of newly hired individuals is found for the month of July. The data material will be analyzed and processed in R-studio.

3.3 Adjustment

In the Short-term employment statistics, results are presented for permanent employees and fixed-term/contracted/hourly paid employees. A fixed-term, contracted or an hourly paid employer is an employee who works at a workplace during one specific time period or as an hourly paid employee. A permanent employee is an employee who work permanently at a workplace. In the KS-register, the terms of employment is not given. This gives a register containing employees with all kinds of term of employment. This will give a higher proportion of newly hired and separations than the tables produced within the framework of the Short-term employment statistics since the fixed-term/ contracted/hourly paid employees are more likely to be newly hired during a month than permanent employees. This can be seen in table ten in the Short-term employment statistics report from the

(17)

second quarter of 2020. It can be read that 37.5 percent of the contracted employees are newly hired that quarter compared to only 2.4 percent of the permanent employees (SCB 2020a).

One way towards being able to present staff turnover estimates for only the permanent employees, is to estimate the total number of fixed-term/contracted/hourly paid employees. The Short-term employment statistics survey use the sampling design stratified simple random sampling, and the combined ratio estimator as seen below can be used.

(1)

Where is the total number of employees from the KS-register in sector , is the

total number of fixed-term/contracted/hourly paid employees from the Short-term employment

statistics in sector and is an estimate of the total number of employees from the

Short-term employment statistics in sector . This gives an estimate of the total number of fixed-Short-term/ contracted/hourly paid employees in the KS-register in sector (Särndal, Swensson & Wretman 2003). If it is reasonable to assume that the ratio (data from previous month) is quite stable, then this estimator can be used. The estimated total number of fixed-term/contracted/hourly paid may be estimated by totals from the Short-term employment statistics report from month to estimate the

number of fixed-term/contracted/hourly paid employees month . To estimate the number of

fixed-term/contracted/hourly paid employees month , the ratio may consist of estimates from

month .

The variance of a ratio estimator is seen in equation two.

(2)

Where is the estimated total number of employees in sector in the Short-term employment statistics and is the estimated number of fixed-term/contracted/hourly paid employees in sector

in the Short-term employment statistics. is the estimated ratio and is the

covariance between and (Särndal, Swensson & Wretman 2003).

When having estimated total number of fixed-term/contracted/hourly paid employees in each sector along with the corresponding variance, the estimated total must be multiplied with the estimated

proportion of newly hired fixed-term/contracted/hourly paid employees ( ) for that quarter (if the

proportion for every sector is available, then it is preferred). These figures can be found in the reports from the Short-term employment statistics. This gives the estimated total number of newly hired fixed-term/contracted/hourly paid employees for sector (see equation three below). When the estimated number of newly hired fixed-term/contracted/hourly paid employees and the total number of fixed-term/contracted/hourly paid employees are estimated, these figures are

̂t yrai = ∑ U x_kKS_i⋅ ∑ H h=1Nh¯ysh ∑H_h=1Nh¯xsh ∑ U xkKSi i H ∑ h=1 Nh¯ysh i _∑H h=1 Nh¯xsh i i t t + 1 t + 2 t + 1 ̂V( ̂R) = 1_̂t 2x[ ̂V( ̂ty) + ̂R2 ̂V( ̂tx) − 2 ̂R ̂C( ̂ty, ̂tx)] ̂t x i ̂t y i ̂R ∑Hh=1Nh¯ysh ∑H_h=1Nh¯xsh ̂C ̂t x ŷt ̂p ks ̂t nhi i

(18)

subtracted from the calculations of the staff turnover to only receive estimated staff turnover for permanent employees (equation four).

(3)

(4)

These estimates should be estimated for every sector separately. The proportion of newly hired fixed-term/contracted/hourly paid employees should also have a variance estimator to estimate the corresponding variance of the ratio. This should be further explored if this method of estimating the staff turnover for only permanent employees is to be done.

Due to lack of Short-term employment statistics micro data, the variance estimate for the total of newly hired fixed-term/contracted/hourly paid employees is not possible to estimate in this thesis. However, if the desire is to receive figures that might be more comparable with the staff turnover result for permanent employees and micro data from the Short-term employment statistics is available. Then this estimator (equation number one) mentioned above is one step towards being able to estimate the proportion of newly hired permanent employees which are the proportions that are presented in the staff turnover tables in the Short-term employment statistics.

4 Data

This section contains information about the collection, processing, editing, description of the data along with the most important variables and derived variables. It also contains results from the Short-term employment statistics.

4.1 Data collection

The data is collected from two data sources available and received from Statistics Sweden. The main source of information is the employer declaration at the individual level (AGI) which are given by the tax agency (Skatteverket) to Statistics Sweden every month. The AGI-register is contained in a Microsoft SQL Server database and the data are transferred with the help from the odbc package to R-studio to be further analyzed (ODBC 2020). The received data only contains information for companies in Sweden who belongs to the private sector. The other data source is called AKU_AGI_13manader (AA13). This contains information about the workplaces that has been active during the time periods and this data frame are received in the same way as the AGI-register.

4.2 Data description

This part contains information about the different registers and the variables used in this study.

4.2.1 Identification variables

In the employer declaration at the individual level a lot of variables are accounted for. The identification variables that will be used in this study is workplace number, the personal

identification number of the payee, time period , organization number and the CfarNr. The

workplace number is a number between 0 and 99 999, this variable is reported in the AGI-register if the business/organization has more than one workplace. The personal identification number is a

̂t

yra_i⋅ ̂pks= ̂tnh_i

(Changes + additions − ̂tnhi)

(19)

unique number for each person in the register, however some of the individuals in the register has the value 0000000 as personal identification number. Time period refers to the month in question. The organization number is a ten digit code that each business is linked to (Skatteverket 2020). The CfarNr is an eight digit identity code for the workplace (SCB 2020e).

The second part of the declaration is called AKU_AGI_13manader (AA13) and consist of variables that describes the workplace with information such as organization number, workplace number,

period of time, sector of operation for the business and sector of operation for the workplace.

(Skatteverket 2020). One important matter that should be pointed out is that the AGI-register do not contain information about number of employees or if a person is employed at a workplace or not, it contains information about salary payments to individuals who has worked at the workplace, this is a difference that needs to be acknowledged. However, in this study, an employed person will be an individual who belongs to the AGI-register and that is tied to a specific organization number and workplace (Skatteverket 2020).

4.2.2 Study variables

There are three study variables in the KS-register and two of them (newly hired and newly separated) are derived variables:

Number of employees Newly hires

Newly separated

The definition of a new hire when using the KS-register is when an individual has received payment

from a workplace month but did not receive payment from that workplace month .

Interpreted as that the individual began their work at the workplace month . A separation is when an individual received payment from a workplace month but did not receive payment from that workplace month . Interpreted as that the individual finished their work at the workplace month . The number of employees for workplace is all of the employees that receive payment from workplace during month . The above variable can be broken down to subsets by gender as in Short-term employment statistics.

y1= y₂= y3= t t − 1 t t t + 1 t k k t

(20)

4.2.3 Employer declaration at the individual level data

The variables that are given and used in this study from the AA13-register is organization number,

workplace number, period of time, sector of operation for the business and sector of operation for the workplace. The AA13-register has 4 880 568 objects and as described, each object in the

register is a workplace during one month of the time period. The sector variables JE_Ng1 and

AE_Ng1 are the sector of operation for the business and for the workplace respectively. These two

sector variables consists of six digits. The first two digits gives the two digit group which is divided into letter sections. This is described in table one.

Table 1. Description of letter group and sector of operation (SCB 2020c).

Above, the digit code and letter sections is shown along with its corresponding sector of operation. In the other register, an individual who receives payment or another form of benefit from a workplace the specific period of time belongs to the register. The variables from the AGI-register

Letter section (digit group) Sector of operation (description)

A (01-03) Agriculture, forestry and fishing

B (05-09) Extraction of materials

C (10-33) Manufacturing

D (35) Supply of electricity, gas, heating, cooling.

E (36-39) Water supply, sewage treatment, waste management and _remediation

F (41-43) Construction activities

G (45-47) Trade, repair of motor vehicles and motorcycles

H (49-53) Transport and storage

I (55-56) Hotel and restaurant business

J (58-63) Information and communication activities

K (64-66) Finance and insurance business

L (68) Real estate activities

M (69-75) Activities in law, economics, science and technology

N (77-82) Rental, real estate services, travel services and other _{support services}

P (85) Education

Q (86-88) Health and social care; social services

R (90-93) Culture, entertainment and leisure

(21)

that are received in this study are the variable period of time, which describes at which time period (month) the information is received, astnr, which is the workplace number, personnr is the individuals personal identification number, peorgnr is the organization number and CfarNr who is the workplace identity code. (Skatteverket 2020). The AGI-register contains 46 374 414 observations. Each row consist of a personal identification number, an organization number, a workplace number, CfarNr and period of time. For the variable period, there are 13 different months ranging from October in 2019 to October in 2020.

4.2.4 Descriptive statistics

This section present descriptive statistics for the KS-register data with the aid from figures and tables describing and illustrating the data.

Figure 1. Number of employees, divided by month of work in the KS-register.

In the above figure, the number of employees each month can be seen between October 2019 and October 2020. In October 2019, there are approximate 3.7 million employees in the private sector. This amount of employees continues until January of 2020 when the number of employees decrease to around 3.2 millions. This is due to a drop out of information that occurred in January 2020 which makes the figures from January above to be considered with caution. After January 2020 (February-September), the number of employees lies stable around 3.5 million employees.

(22)

Figure 2. Number of workplaces by month in the KS-register.

The figure above illustrates the number of workplaces each month. The number of workplaces ranges between 320 000 and 400 000. It may be noticed that the number of workplaces is almost constant between the time periods if excluding the month of January.

(23)

The table below shows the number of employees each month by gender. The unknown gender column shows the quantity individuals in the KS-register that did not have a personal identification number.

Table 2. Number of employees each month, by gender in the KS-register.

In the above table, the number of employees can be seen each month, divided into males, females and the individuals with unknown gender. It may be noticed that the majority of employees are men. There are also a number of individuals with unknown personal identification numbers, hence impossible to extract gender from the register. The total amount of objects in the KS-register that has no personal identification number are counted to 51 700 objects of the total amount of 46 374 414 objects. These 51 700 can be the same individuals during different time periods meaning that the same individual exist 13 times, in every time period.

4.3 Data processing and editing

The KS-register and AA13-register are divided into subsets by period of time (month). This makes 13 different subsets containing information about one specific month. This is done to make it possible to compare each month. Packages such as dplyr and tidyr are used in the processing and editing of the data in the KS-register (Dplyr 2020) (Tidyr 2020).

Some of the objects in the AA13-register only had values for the variable JE_Ng1 (305 441 workplaces) meaning that the register only has information about the sector of operation for the

Number of

employees Males: Females: Unknown gender: Total:

Oktober 2019 2 237 262 1 456 921 3 839 3 698 022 November 2019 2 233 076 1 454 319 3 681 3 691 076 December 2019 2 294 280 1 513 068 4 559 3 811 907 January 2020 1 923 193 1 271 673 3 466 3 198 332 February 2020 2 156 509 1 407 422 4 657 3 568 588 Marsch 2020 2 186 015 1 421 035 3 838 3 610 888 April 2020 2 168 148 1 401 719 3 763 3 573 630 May 2020 2 126 967 1 351 492 3 209 3 481 668 June 2020 2 180 489 1 388 087 3 830 3 572 406 July 2020 2 166 489 1 387 778 3 658 3 557 925 August 2020 2 157 056 1 382 922 4 009 3 543 987 September 2020 2 166 394 1 396 785 4 155 3 567 334 October 2020 2 125 702 1 367 913 5 036 3 498 651 Total: 28 121 580 18 201 134 51 700 46 374 414

(24)

business, not the workplace. For these workplaces, the sector of operation for the business is imputed as the sector of operation for the workplace.

4.4 Derived variables

The variable gender is derived in the KS-register from the variable PIN. This is done to make it possible to divide the data frame into subsets by genders as seen in the descriptive statistics section. If the individual is a male the object is coded 1, if the individuals is a female the object is coded 2 and 0 if gender is unknown.

The variable males and females are also derived in to the AA13-register. From the KS-register, the number of male and female employees each month for each workplace is extracted by giving each workplace each month a new PIN code in the AA13-register. Then the AA13-register is linked to the KS-register by the matching keys organization number and workplace number. When every individual in the KS-register has a PIN code that is tied to a specific workplace, then a data set is created for the month in question which shows the new PIN code for the workplace and the total amount of employees that is tied to that new pin code divided by females and males. This table is linked using methods described by Wallgren and Wallgren by the new PIN code to the AA13-register and the new variables males and females are derived which shows the number of male and female employees at each workplace the month in question. The variable nworkers is then derived to the AA13-register by adding the number of male and female employees at each workplace, which gives to total number of employees at each workplace.

To make it possible to create subsets of the data in relation to sector of operation, a new variable called sector is derived in the KS-register. This new variable is generated by the variables AE_Ng1 and JE_Ng1 from the AA13-register. The two registers are linked together by matching keys; organization number and workplace number, since the KS-register contain the variables AE_Ng1 and JE_Ng1, the first two digits in the six digit code that AE_Ng1 and JE_Ng1 consist of is coded to be the new variable sector. Now sector consist of two digits. If the workplace did not have a sector, the business’s sector of operation was imputed. Then the two digit code can be decoded (from table two) to find the sector of operation for the workplace (SCB 2020c).

The variables newly hired and newly separated are also derived variables, they are described in section 4.2.2.

To summarize, the KS-register and the AA13-register is divided into 13 months, each of the columns in the KS-register containing information about PIN, organization number, workplace number, sector of operation for the workplace, gender, CfarNr and period of time. This is all that is required to be able to calculate the staff turnover for different sectors in different months using the package ”CompareDF”. For the AA13-register each row has variables containing information about organization number, workplace number, sector of operation for business and workplace, number of female employees, number of male employees, total number of employees and period of time. The new PIN code that was introduced to find information about number of employees for each workplace is dropped from the register.

(25)

4.5 Short-term employment statistics, results

This section shows results from the Short-term employment statistics survey. The quarters that will be presented here is the second quarter of 2020 and the third quarter of 2020. These results are shown because this makes is possible to compare the staff turnover results along with results about total amount of employees in each sector from the KS-register.

Table 3. Proportion (in percent) of hires and separations in relation to the total number of

employees (permanent employees) presented as confidence intervals by sector, second quarter of 2020.1

The table above shows estimated newly hired proportions and estimated proportions of separated in the form of confidence intervals from the Short-term employment statistics the second quarter of the year 2020. One thing that may be noticed once again is that the proportion of separations is higher than the proportion of newly hired for the second quarter of 2020. In the report from the second quarter of 2020 it can be found that 2.4 percent of the total amount of permanent employees are newly hired and 37.5 percent of the fixed-term/contracted/ hourly paid employees are newly hired. 4.1 percent of the total amount of permanent employees are separated from their work while 25.6 percent of the fixed-term/contracted/ hourly paid employees are separated from their work (SCB 2020a).

Sector of operation Newly hired Separations

A: Agriculture, forestry and fishing. [1.7 ; 7.5] [0.5 ; 4.7]

B+C: Extraction of materials and

manufacturing. [1.1 ; 1.5] [1.9 ; 2.5]

D+E: Supply of electricity, gas, heating, cooling and Water supply, sewage treatment,

waste management and remediation. [1.9 ; 3.3] [1.6 ; 2.4]

F: Construction activities. [2.5 ; 4.9] [2.9 ; 6.5]

G: Trade, repair of motor vehicles and

motorcycles. [1.4 ; 2.6] [2.6 ; 4.0]

H: Transport and storage. [1.3 ; 3.0] [3.3 ; 6.5]

I: Hotel and restaurant business. [0.9 ; 6.1] [8.9 ; 18.5]

J: Information and communication activities. [2.1 ; 3.9] [2.3 ; 4.1]

K: Finance and insurance business. [1.4 ; 2.2] [1.9 ; 3.7]

L: Real estate activities. [1.3 ; 4.9] [0.5 ; 3.7]

M+N: Activities in law, economics, science, technology, rental, real estate services, travel

services and other support services. [2.1 ; 3.3] [4.5 ; 6.3]

P: Education. [0.4 ; 1.6] [2.3 ; 5.1]

Q: Health and social care; social services. [2.4 ; 5.2] [3.0 ; 4.8]

R+S: Culture, entertainment, leisure and other

service activities. [1.0 ; 4.0] [2.5 ; 4.9]

TOTAL [2.1 ; 2.7] [3.7 ; 4.5]

SCB 2020a

(26)

Table 4. Proportion (in percent) of hires and separations in relation to the total number of

employees (permanent employees) presented as confidence intervals by sector, third quarter of 2020. 2

The table above shows estimated newly hired proportions and estimated proportions of separated in the form of confidence intervals from the Short-term employment statistics the third quarter of the year 2020. This table shows the same as table number three and four, that the proportion of separations is higher than for newly hired. In the report from the third quarter of 2020 it can be found that 2.8 percent of the total amount of permanent employees are newly hired and 32.5 percent of the fixed-term/contracted/ hourly paid employees are newly hired. 3.8 percent of the total amount of permanent employees are separated from their work while 31.4 percent of the fixed-term/ contracted/ hourly paid employees are separated from their work (SCB 2020d).

Sector of operation Newly hired Separations

A: Agriculture, forestry and fishing. [1.0 ; 7.0] [0.8 ; 4.2]

manufacturing. [1.1 ; 1.5] [1.8 ; 2.2]

D+E: Supply of electricity, gas, heating, cooling and Water supply, sewage treatment,

waste management and remediation. [1.9 ; 3.7] [1.8 ; 2.6]

F: Construction activities. [2.0 ; 4.2] [2.4 ; 5.0]

G: Trade, repair of motor vehicles and

motorcycles. [1.8 ; 3.6] [2.9 ; 4.7]

H: Transport and storage. [2.3 ; 5.7] [3.1 ; 5.9]

I: Hotel and restaurant business. [2.6 ; 7.6] [4.7 ; 12.9]

J: Information and communication activities. [2.1 ; 3.3] [2.4 ; 4.4]

K: Finance and insurance business. [1.5 ; 2.3] [1.9 ; 2.7]

L: Real estate activities. [1.1 ; 4.7] [1.6 ; 4.0]

M+N: Activities in law, economics, science, technology, rental, real estate services, travel

services and other support services. [2.5 ; 4.1] [4.2 ; 6.0]

P: Education. [3.2 ; 6.2] [2.8 ; 5.2]

Q: Health and social care; social services. [2.0 ; 3.6] [2.4 ; 3.8]

R+S: Culture, entertainment, leisure and other

service activities. [1.5 ; 4.1] [1.9 ; 4.1]

(27)

Table 5. Number of permanent employees, contracted/hourly paid employees and total number of

employees presented as confidence intervals by sector, second quarter of 2020.3

In the private sector, it was estimated to be 3 251 200 employees the second quarter of 2020. 2 802 209 individuals are estimated to be permanent employees and 449 034 individuals are estimated to be fixed-term/contracted/ hourly paid employees (SCB 2020a).

Sector of operation Total number of employees Permanent employees Contracted/hourly paid A: Agriculture, forestry and

fishing. [42 183 ; 51 051] [33 300 ; 37 228] [7 276 ; 15 430]

B: Extraction of materials. [9 079 ; 9 647] [8 207 ; 8 725] [836 ; 958]

C: Manufacturing. [525 548 ; 534 992] [492 185 ; 500 471] [31 395 ; 36 491]

D: Supply of electricity, gas,

heating, cooling. [26 983 ; 30 275] [26 288 ; 29 534] [527 ; 907]

E: Water supply, sewage treatment, waste management and

remediatio [18 934 ; 21 428] [16 854 ; 19 140] [1 741 ; 2 627]

F: Construction activities. [319 523 ; 339 229] [306 842 ; 325 336] [9 957 ; 16 617]

G: Trade, repair of motor vehicles

and motorcycles. [571 587 ; 612 727] [496 249 ; 522 847] [69 526 ; 95 692]

H: Transport and storage. [210 087 ; 225 309] [181 421 ; 195 377] [26 495 ; 32 105]

I: Hotel and restaurant business. [135 607 ; 158 187] [104 560 ; 119 252] [25 873 ; 44 109]

J: Information and communication

activities. [203 612 ; 216 148] [193 411 ; 205 771] [8 574 ; 12 004]

K: Finance and insurance

business. [94 577 ; 99 505] [87 133 ; 91 979] [6 811 ; 8 159]

L: Real estate activities. [70 054 ; 76 942] [64 081 ; 70 483] [5099 ; 7 333]

M: Activities in law, economics,

science, technology. [288 612 ; 300 066] [272 632 ; 283 808] [14 104 ; 18 134]

N: Rental, real estate services, travel services and other support

services. [255 540 ; 274 908] [182 383 ; 198 611] [67 424 ; 82 030]

P: Education. [85 618 ; 92 412] [66 148 ; 71 318] [17 814 ; 22 750]

Q: Health and social care; social

services. [219 607 ; 239 051] [131 270 ; 145 644] [83 432 ; 98 312]

R: Culture, entertainment, leisure. [31 602 ; 37 090] [23 432 ; 26 756] [7 108 ; 11 396]

S: Other service activities. [35 254 ; 38 734] [30 916 ; 34 216] [3 413 ; 5 443]

Total: [3 218 769 ; 3 283 719] [2 777 609 ; 2 826 809] [428 824 ; 469 646]

SCB 2020a

(28)

Table 6. Number of permanent employees, contracted/hourly paid employees and total number of

employees presented as confidence intervals by sector, third quarter of 2020.4

In the private sector, it was estimated to be 3 248 800 employees the third quarter of 2020. 2 791 256 individuals are estimated to be permanent employees and 457 550 individuals are estimated to be fixed-term/contracted/ hourly paid employees (SCB 2020d). It may be noticed that the estimated confidence intervals are somewhat wider in table eight than in table six and seven.

Sector of operation Total number of employees Permanent employees Contracted/hourly paid A: Agriculture, forestry and

fishing. [41 400 ; 48 198] [33 310 ; 37 186] [6 439 ; 12 663]

B: Extraction of materials. [9 343 ; 10 001] [8 212 ; 8 828] [1 048 ; 1 256]

C: Manufacturing. [521 011 ; 530 0741] [486 830 ; 494 972] [32 407 ; 37 543]

heating, cooling. [26 886 ; 30 170] [26 129 ; 29 367] [599 ; 963]

remediatio [19 207 ; 21 109] [16 818 ; 18 538] [2 072 ; 2 888]

F: Construction activities. [322 450 ; 342 982] [306 981 ; 326 565] [11 668 ; 20 218]

and motorcycles. [578 133 ; 618 341] [492 257 ; 520 359] [79 763 ; 104 095]

H: Transport and storage. [212 049 ; 229 191] [180 437 ; 195 329] [28 141 ; 37 333]

I: Hotel and restaurant business. [133 665 ; 160 231] [102 012 ; 119 722] [25 325 ; 46 837]

activities. [205 150 ; 217 930] [195 486 ; 207 602] [8 271 ; 11 721]

business. [93 785 ; 98 733] [87 018 ; 91 860] [6 246 ; 7 394]

L: Real estate activities. [70 669 ; 77 667] [63 492 ; 69 882] [6 223 ; 8 739]

science, technology. [285 837 ; 297 675] [271 060 ; 282 482] [13 047 ; 16 923]

N: Rental, real estate services, travel services and other support

services. [253 574 ; 273 338] [181 389 ; 198 339] [67 042 ; 80 142]

P: Education. [81 456 ; 87 968] [66 527 ; 71 497] [13 432 ; 17 968]

services. [217 881 ; 236 371] [131 727 ; 145 553] [82 377 ; 94 595]

R: Culture, entertainment, leisure. [32 412 ; 38 292] [23 515 ; 27 035] [7 750 ; 12 404]

S: Other service activities. [34 635 ; 38 335] [30 152 ; 33 414] [3 642 ; 5 762]

(29)

5 Methodology

The estimator (or mathematical formula) that will be used to calculate the totals are found in the book Register-based Statistics, Statistical Methods for Administrative Data by Anders Wallgren and Britt Wallgren on page 202 and is shown in equation three.

(5)

Where R is the number of objects in one specific domain and is an object that belongs to the domain. The estimates that is produced by this estimator depends on which method(s) that was used to create the register. If the work of creating the register is different, the values estimated from this estimator with the same base register, may be different (Wallgren & Wallgren 2014).

To estimate proportions, equation four is used.

(6)

Where and are estimated totals from equation three. In the book by Särndal, Swensson and Wretman, the totals are estimated with the Horvitz-Thompson estimator but in this case, the totals are estimated with equation three (Särndal, Swensson & Wretman 2003).

The estimates are made from values on aggregate. The number of ”additions” the investigated month is summed together with the number of individuals who ”changed” their workplace and/or organization. This value is divided by the total amount of employees in the KS-register in sector ( ) and multiplied by 100 to receive the proportions (in percent) of newly hired (equation 7). The same are done to calculate the proportion of separations. The only change is that the number of ”removals” the investigated month is counted compared to the next coming month and this is summed with the individuals who ”changed” their workplace and/or organization between the investigated month and the next coming month (equation 8).

(7) (8) ̂t y = R ∑ i=1 y_i i ̂R = ŷt ̂t x ̂t y x̂t i t_y_i (Changes + additions) ty_i ⋅ 100 = pnewlyhired (Changes + removals) tyi ⋅ 100 = pnewlysepareted

(30)

6 Results

This section covers results from this study, containing tables, descriptions and comparisons with the Short-term employment statistics. The result section is divided into four parts. The first two parts contains results from the six selected months along with the staff turnover estimates for the whole quarters, the third part describes the staff turnover for each month divided by gender and the last part is a comparison on workplace level between the Short-term employment statistic and the KS-register. April, May and June 2020 will be compared against the second quarter of 2020 and July, August and September 2020 will be compared against the third quarter of 2020.

(31)

6.1 Second quarter of 2020

Here, results are shown for April, May and June 2020 separately and combined (for the entire quarter).

Table 7. Proportion (in percent) of hires and separations in relation to the total amount of

employees (permanent employees, fixed-term employees/contracted employees and hourly paid employees) by sector, April, May and June 2020.

Proportion of

employees New hires Separations

Sector of

operation April May June April May June

A: Agriculture,

forestry and fishing. 12.6 12.8 18.0 8.98 8.15 13.5

manufacturing. 3.59 3.00 4.40 3.90 3.06 3.73

D+E: Supply of electricity, gas, heating, cooling and Water supply, sewage treatment, waste management and remediation. 5.39 5.18 6.47 5.65 4.77 6.03 F: Construction activities. 5.70 5.32 5.86 5.08 6.41 6.41 G: Trade, repair of motor vehicles and

motorcycles. 6.82 6.76 6.47 7.29 6.40 5.48

H: Transport and

storage. 6.85 6.10 7.19 8.84 6.82 5.98

I: Hotel and restaurant

business. 10.2 10.2 16.6 26.1 13.2 11.4 J: Information and communication activities. 7.64 4.81 5.62 8.50 5.09 6.26 K: Finance and insurance business. 7.00 8.16 20.1 7.13 7.91 19.7 L: Real estate activities. 19.5 24.3 37.8 19.8 23.3 37.6 M+N: Activities in law, economics, science, technology, rental, real estate services, travel services and other support services.

8.28 7.48 7.87 9.99 7.97 8.60

P: Education. 8.52 7.63 11.1 11.7 9.67 12.3

Q: Health and social

care; social services. 8.91 8.14 8.80 10.1 8.06 7.95

R+S: Culture, entertainment, leisure and other service activities.

14.9 13.9 22.2 20.1 14.3 23.6

(32)

Above the staff turnover for the months of April, May and June of 2020 It may be noticed that the proportions are quite high compared to the estimates from the Short-term employment statistics. However six of the estimated proportions falls into the confidence interval (confidence intervals are seen in table three) that is produced in the Short-term employment statistics survey for the second quarter.

Table 8. Total number of employees by sector of operation, April, May and June 2020.

It might be noticed that the largest sector in the table above is sector G (trade, repair of motor vehicles and motorcycles) and the sector with the least amount of employees are sector B (extraction of material).

Total number of employees

Sector of operation April May June

A: Agriculture, forestry and

fishing. 41 799 43 716 49 256

B: Extraction of materials. 5 273 5 331 5 388

C: Manufacturing. 515 449 510 428 517 726

heating, cooling. 19 077 18 979 19 314

remediation. 11 855 11 866 11 996

F: Construction activities. 345 327 345 943 350 692

and motorcycles. 577 637 574 091 574 709

H: Transport and storage. 211 938 205 437 206 763

I: Hotel and restaurant business. 181 998 149 686 155 896

activities. 226 144 217 177 218 457

business. 137 879 139 881 160 835

L: Real estate activities. 69 818 74 101 91 697

science, technology, rental. 391 380 384 673 389 342

N: Real estate services, travel services and other support

services. 260 674 249 328 244 449

P: Education. 119 750 114 500 116 361

services. 248 304 242 968 245 106

R: Culture, entertainment, leisure. 87 561 77 639 84 947

Register based staff turnover statistics : An evaluation of the Employer declaration at the individual level